University of Southern California Dissertations and Theses

Intermedia agenda setting in an era of fragmentation: applications of network science in the study of mass communication

(USC Thesis Other)

Intermedia agenda setting in an era of fragmentation: applications of network science in the study of mass communication

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INTERMEDIA AGENDA SETTING IN AN ERA OF FRAGMENTATION: APPLICATIONS OF NETWORK SCIENCE IN THE STUDY OF MASS COMMUNICATION By Katherine Ognyanova ________________________________________________ A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (COMMUNICATION) May 2013 Copyright 2013 Katherine Ognyanova ii ACKNOWLEDGEMENTS I owe a debt of gratitude to more people than I can name on this page. The following list mentions only a few of the great friends and scholars who have contributed to my work over the last five years. Thanks go out to:  My family, for their endless love and support. Special thanks are due to my parents who gave me the best of both nature and nurture, my kind and caring aunt, and my sister who threatened to get her Ph.D. first if I did not hurry up.  Omri, who almost never complained about our severely impaired social life during the long months of dissertation writing. He deserves high praise for enduing most graciously countless discussions of media fragmentation and network models.  My advisor, Sandra Ball-Rokeach – a brilliant academic mind and an exceptional mentor without whom this intellectual journey would have been impossible.  Peter Monge, a consummate scholar who shared with me the wonderful world of networks and inspired the WWPD (What Would Peter Do) approach to data analysis.  Geneva Overholser and Ann Crigler, who provided invaluable insights from their respective fields of journalism and political science.  My friends back at home, for putting up with my prolonged absence.  My friends in Los Angeles, for putting up with my prolonged presence.  My dissertation support group, for all the academic discussions and bottles of wine we shared.  The Metamorphosis research group and the Annenberg Networks Network – two teams I am proud to be a member of.  The Annenberg School and its good people who helped every step of the way: Larry Gross, Tom Goodnight, Imre Meszaros, Carola Weil, Ernest J. Wilson III, and of course the infinitely patient and sympathetic Anne Marie Campian. iii TABLE OF CONTENTS ACKNOWLEDGEMENTS ................................................................................................ ii LIST OF TABLES ............................................................................................................ vii LIST OF FIGURES ......................................................................................................... viii ABSTRACT ....................................................................................................................... ix INTRODUCTION .............................................................................................................. 1 CHAPTER 1: THE AGENDA-SETTING TRADITION: BACKGROUND AND HISTORY .................................................................................... 5 Intellectual origins of the theory ................................................................................... 7 Development and early studies ..................................................................................... 9 Definitions: issue, agenda, and salience ..................................................................... 10 Typology of agenda-setting research .......................................................................... 13 Setting the public agenda ............................................................................................ 17 Building the media agenda .......................................................................................... 23 News values and journalistic standards ...................................................................... 26 Gatekeeping: levels of analysis ................................................................................... 31 CHAPTER 2: CURRENT TRENDS IN AGENDA-SETTING RESEARCH ................. 33 iv CHAPTER 3: CONTEMPORARY CHALLENGES TO THE AGENDA-SETTING PERSPECTIVE .................................................................... 45 Conceptual Challenges: Agenda Fragmentation ......................................................... 46 Fragmented public agenda: selective exposure ..................................................... 50 Reversed direction of influence: bottom-up effects .............................................. 55 Fragmented media agenda: more voices, more diversity? .................................... 59 Policy Challenges: Media Diversity ........................................................................... 64 CHAPTER 4: A NETWORK MODEL OF AGENDA-SETTING. ADDRESSING EXISTING CHALLENGES ................................................................... 70 A network approach to agenda-setting ....................................................................... 73 Relationship typology ........................................................................................... 74 Issue adoption (issue – information source/ audience member) ........................... 74 Media use (audience member – information source) ............................................ 74 Interorganizational ties (information source – information source) ..................... 75 Social connections (audience member – audience member) ................................ 75 Concept associations (issue – issue) ..................................................................... 76 Link direction and agency ..................................................................................... 77 Individual and dyadic attributes ............................................................................ 77 Network mechanisms ............................................................................................ 78 Reducing complexity .................................................................................................. 81 v Research design and methodological challenges ........................................................ 84 CHAPTER 5: FORMULATING HYPOTHESES............................................................ 89 News Values and Selection Mechanisms: Drivers of Agenda Convergence ............. 90 CHAPTER 6: RESEARCH DESIGN ............................................................................... 97 Dataset Description ..................................................................................................... 97 Data cleaning procedures ............................................................................................ 99 Measurement ............................................................................................................. 100 Ownership ........................................................................................................... 101 Channel ............................................................................................................... 101 Sector .................................................................................................................. 101 Audience size ...................................................................................................... 101 Audience demographics ...................................................................................... 102 Party identification / Political ideology .............................................................. 103 Agenda convergence network ............................................................................. 104 Analysis..................................................................................................................... 106 CHAPTER 7: AGENDA CONVERGENCE: STUDY RESULTS ................................ 112 RQ1: Fragmentation over time ................................................................................. 112 RQ2: Levels of similarity across time points ............................................................ 116 H1-H7: Actor-based stochastic model estimates ...................................................... 117 Post-Hoc Analysis ..................................................................................................... 123 vi CHAPTER 8: DISCUSSION AND CONCLUSION ..................................................... 125 Limitations and further research: data collection and methodological considerations ................................................................................................................................... 133 REFERENCES ............................................................................................................... 138 APPENDIX A: NEWS OUTLET LIST ........................................................................ 176 APPENDIX B: NEWS OUTLET ATTRIBUTES ......................................................... 178 APPENDIX C: AGENDA CONVERGENCE NETWORK METRICS I ..................... 181 APPENDIX D: AGENDA CONVERGENCE NETWORK METRICS II ................... 183 APPENDIX E: NEWS OUTLET DEGREE DISTRIBUTIONS .................................. 184 vii LIST OF TABLES Table 1. Agenda-setting articles by journal, 2000-2011. .................................................. 36 Table 2. Agenda Convergence Model 1: parameter estimates. ...................................... 119 Table 3. Agenda Convergence Model 2: accounting for time heterogeneity in density. 120 viii LIST OF FIGURES Figure 1. The Acapulco Typology .................................................................................... 15 Figure 2. Mentions of agenda setting in books, 1940-2008. Source: Google NGram. .... 33 Figure 3. Number of agenda-setting articles per journal. ................................................. 37 Figure 4. Number of agenda-setting articles published yearly. ........................................ 37 Figure 5. The focus of agenda-setting articles (media, public, policy, activist agendas). 38 Figure 6. Formats examined by agenda-setting articles focusing on the media agenda. . 39 Figure 7. Theoretical and empirical agenda-setting articles. ............................................ 39 Figure 8. Data collection strategies used in the empirical agenda-setting articles. .......... 40 Figure 9. Data analysis methods used by empirical quantitative agenda-setting articles. 41 Figure 10. Major issues examined in agenda-setting articles. .......................................... 42 Figure 11. Average number of citations to agenda-setting articles per journal. ............... 43 Figure 12. Conceptual challenges to the agenda-setting process. .................................... 48 Figure 13. Agenda-setting: A network model example. ................................................... 73 Figure 14. Network mechanisms underlying agenda-setting processes. .......................... 79 Figure 15. Reducing the complexity of networked agenda-setting models. .................... 82 Figure 16. Agenda convergence network: Fragmentation over time (2008) .................. 112 Figure 17. Agenda-convergence: a network snapshot for Nov. 3rd to 9th of 2008. ...... 114 Figure 18. Agenda convergence network: Density over time (2008) ............................ 116 Figure 19. Agenda convergence network: Average correlations (2008). ....................... 117 Figure 20. Goodness of fit: In-degree distribution. ........................................................ 121 Figure 21. Goodness of fit: Out-degree distribution ...................................................... 121 ix ABSTRACT This work proposes a relational approach to the study of news agendas and media fragmentation in a digital age. It examines the origins, development, and current status of mass communication theories dealing with news selection and impact. Recent conceptual and methodological challenges facing research in a changing information environment are discussed. A broad analytical strategy exploring the media system as a dynamic multilevel, multidimensional network is outlined. Implementing this approach, the study sets out to evaluate fragmentation levels and factors predicting agenda convergence in a longitudinal network of news outlets. The analysis is based on secondary data collected by the Pew Center’s Project for Excellence in Journalism (2008) and the 2008 National Annenberg Election Survey. The study sample contains U.S. news outlets from five industry sectors: newspapers, online sources, radio, cable and network TV stations. The results uncover a decrease in fragmentation over time, with a minimum reached during the U.S. presidential elections in November 2008. The dynamics of agenda convergence are found to be shaped by the story selections of popular outlets and driven by similarities in format, audience demographics, and political ideology. Audience size does not significantly influence the correspondence in news source agendas over time. The analysis also shows that co-ownership relations lead to lower agenda convergence for the outlets in the sample. 1 INTRODUCTION This work proposes a relational approach to the study of news agendas and media fragmentation in a digital age. It examines the origins, development, and current status of mass communication theories dealing with news selection and impact. Recent conceptual and methodological challenges facing research in a changing information environment are discussed. A broad analytical strategy exploring the media system as a dynamic multilevel, multidimensional network is outlined. Implementing this framework, the study sets out to evaluate fragmentation levels and factors predicting agenda convergence in a longitudinal network of news outlets. Chapter 1 presents a broad overview of the agenda-setting paradigm and related theoretical perspectives. It describes the emergence of the tradition and traces its development over time. Key premises and concept definitions are examined. The chapter includes a typology of major research designs employed by studies within the theoretical perspective. The media, audience, and policy aspects of agenda setting, as well as its civic and political implications are also discussed. Chapter 2 contains a descriptive quantitative overview of agenda-setting works published in major communication journals over the past twelve years. This section of the thesis provides a snapshot of recent studies in the area, noting their key characteristics, dominant themes, data collection and analytical strategies. The results of this preliminary study are used to identify gaps in the literature that would need to be addressed in the research design employed here. 2 While the first chapter of this work presents a classic take on media content and public priorities, Chapter 3 focuses on recent conceptual and policy challenges. Agenda- setting theory, first developed in the context of a relatively homogeneous, centralized media system, has since been challenged by technological, social, and economic shifts. New digital media formats, user-generated content, proliferation of distribution channels and fragmentation of audiences have called into question the cohesion of media agendas and their effects on the priorities of the public. Chapter 3 outlines related academic debates, examining claims of prevalent bottom-up effects and fragmentation in media production and consumption. The study finds virtually no empirical evidence in academic literature supporting the fragmentation hypotheses. On the contrary, recent research findings indicate that agenda setting processes are still taking place – and mainstream media sources continue to influence public opinion. The final section of the chapter provides an overview of regulatory objectives related to the evaluation and promotion of media diversity (including source, viewpoint, and exposure diversity). The potential contributions of agenda-setting studies to policy-oriented research are discussed. Pressures to rethink the simple model of media influence proposed in the 1970s emerge from parallel developments in theory and society. In order to capture agenda- setting mechanisms in a changing information environment, an integrated framework should reflect the dynamic flow of issues through public and media agendas. Accordingly, Chapter 4 outlines a model constructed as a dynamic multidimensional network of issues, individuals, and information sources. The text provides a rationale for the inclusion of five relation types incorporated in the framework (issue adoption, media use, interorganizational, social, and concept connections). It also sketches relevant 3 structures and actor characteristics. Agenda-setting processes and effects are mapped onto corresponding network mechanisms. The study describes several strategies aimed at reducing the complexity of this approach, both in terms of data collection and analysis. Implementing a simplified version of the proposed framework, Chapter 5 describes a network study of media sources. Avoiding some of the methodological limitations of previous works, the analysis sets out to test the fragmentation hypothesis and identify drivers of intermedia agenda-setting effects. The media agenda is examined through an agenda convergence network of news sources. A tie between two outlets in this network indicates their propensity to select and prioritize content in a similar fashion. The strength of the connection is evaluated based on the overlap in prominent issues covered by media outlets over a period of time. Key factors hypothesized to influence the convergence of media agendas include media sector, ownership patterns, audience size, demographics, and dominant political ideology. The research questions and hypotheses of the study are tested using secondary data collected by the Pew Center’s Project for Excellence in Journalism (2008). The dataset contains U.S. news outlets of five different types: newspapers, radio, cable and network TV stations, and online services. The data include a full year of news coverage classified into several hundred different topic categories. Aggregated audience attributes are obtained from the 2008 National Annenberg Election Survey. Details about the sample and methods used in the analysis are presented in Chapter 6. A network measure of media fragmentation is employed to examine potential trends towards agenda disintegration over time. Agenda convergence drivers are tested through stochastic actor- based modeling (Snijders, Van de Bunt, & Steglich, 2010). 4 Chapter 7 presents the results of the analysis. Agenda fragmentation is found to decrease over time, reaching its minimum during the U.S. presidential elections in November 2008. The dynamics of agenda convergence are shaped by preferential attachment to popular outlets, and driven by similarity in format, audience demographics, and political preferences. Audience size does not significantly influence the correspondence in news source agendas over time. The analysis also shows that co- ownership relations lead to lower agenda convergence for the outlets in the sample. The theoretical, methodological, and policy implications of these research findings are discussed in Chapter 8, which also addresses limitations and future research directions. 5 CHAPTER 1: THE AGENDA-SETTING TRADITION: BACKGROUND AND HISTORY One of the most influential media effects theories, agenda-setting, remains popular even as its premises are challenged by the technological and social transformations of the last two decades. The framework suggests that news media influence the way we perceive the world – a notion proposed as early as 1922 by Walter Lippmann in his book Public Opinion. This assumption later evolved into the understanding that at any given time a limited number of issues occupies the attention of journalists, citizens, and politicians. The focus of public and political attention on a narrow range of topics facilitates a shared perception of community priorities, allowing social mobilization and collective action to take place. Both the content and the format of news stories provide cues about the social relevance of objects and events. This mechanism is particularly well-studied in the context of political communication. Agenda-setting theory in its present form was first articulated by Maxwell McCombs and Donald Shaw (1972) in their classic Chapel Hill study. The study laid the groundwork for a new research paradigm. A host of scholars followed the lead of McCombs and Shaw, often adopting a similar approach of matching content analysis and audience research (Rogers, Dearing, & Bregman, 1993). Agenda-setting emerged as a broad and complex tradition with both cognitive and social dimensions, encompassing aspects of media production, content, and audience research. 6 Academics have since extended the scope of this framework to study the formation of media agendas, investigating factors that influence the salience of items in the news. Work in that area involved exploring key external news sources (extramedia level), the influence of media on each other (intermedia level), and the internal newsroom dynamics affecting editorial decisions (Dearing & Rogers, 1996a). As research largely focused on the interplay between news coverage and public opinion, the intermedia level received relatively less attention. Presently, the agenda-setting framework has also moved beyond its initial focus on issue salience. A set of studies within the tradition has examined the role of media in making particular attributes or aspects of an object more prominent than others (Weaver, McCombs, & Shaw, 2004). This line of research often compares key characteristics of political figures captured by media coverage with those recorded by opinion polls (Coleman, McCombs, Shaw, & Weaver, 2008). Scholars are still investigating the specific mechanisms underlying this second level of the agenda-setting process. Another set of studies worth mentioning here have looked into relevant micro- level patterns. Those include the individual characteristics predicting the strength of media effects, as well as the consequences of agenda-setting for personal attitudes and behavior (McCombs & Reynolds, 2009). This chapter provides a broad overview of the origins and development of agenda-setting research. It summarizes the major premises and concepts of the theory. The following sections describe the emergence of the tradition as a response to limited 7 effects paradigms of mass communication. The media, audience, and policy aspects of agenda studies, as well as their civic and political implications are also discussed here. Important to note, Chapter 1 presents a classic version of the theory developed in the context of a homogeneous, centralized media system. Today, this view has been challenged by technological and economic processes taking place in media and society as a whole. Chapter 3 of this work provides a detailed overview of the resulting theoretical shifts. Intellectual origins of the theory In its early decades, mass communication scholarship was under the influence of propaganda research positing strong and immediate effects of media (particularly broadcast) messages on individuals. In the 1940s and 1950s, however, multiple studies reported little or no impact of mass communication on attitudes and behavior, undermining the idea of all-powerful media. As a result, a limited effects model of mass communication emerged. A paradigm shift in the 1960s and 1970s moved the field away from both the initial magic bullet theories and the following law of minimal consequences research. More nuanced conceptualizations of media influence were developed during that period (Ball-Rokeach & DeFleur, 1976; Gerbner & Gross, 1976; McCombs & Shaw, 1972). Agenda-setting was one research tradition that set out to demonstrate the social significance of media, contrary to the premise of the limited effects model. In a series of overviews of the research perspective, Maxwell McCombs reviews the intellectual origins of agenda-setting (McCombs, 2004; McCombs & Bell, 1996; McCombs & Reynolds, 2009). The earliest work guiding theory development is Public 8 Opinion, a book authored by the American journalist Walter Lippmann (1922). In it he observes that media provide us with cognitive maps to a vast universe of people, places, and events that we could not experience directly. A central theme discussed by Lippmann is the emergence of public opinion, largely based on a version of reality constructed by the press. The research of Paul Lazarsfeld and colleagues (1944) was also important, although in a different way. It established a general approach often used in the field of mass communication in later years. The 1940 presidential election study was the first large-scale systematic social science analysis of media influence on public opinion. The work was based on a panel study of 2400 voters in Erie County, Ohio. It motivated the development of the two-step flow model (E. Katz & Lazarsfeld, 1955; Lazarsfeld et al., 1944) emphasizing the role of interpersonal communication in opinion formation. This perspective came to be one of the major theoretical frameworks associated with the limited effects model of mass communication. In 1948, Harold Lasswell published a canonical book chapter containing his five- part summary of the communication process: "Who Says What in Which Channel to Whom with What Effect”. According to Lasswell, mass communication served three major social functions: surveillance of the environment, coordination of priorities in response to that environment, and transmission of culture between generations. He observed that media had a crucial role in directing the attention of the public and policy- makers to a set of issues. 9 In a book that directly influenced the development of agenda-setting theory, Bernard Cohen (1963) claimed that the press “may not be successful much of the time in telling people what to think, but it is stunningly successful in telling its readers what to think about” (p.13). A decade later, McCombs & Shaw would call this “the agenda- setting function of mass media”. Development and early studies Although grounded in earlier work, agenda-setting theory was first articulated in McCombs and Shaw’s classic Chapel Hill study of media influence on political reality (McCombs & Shaw, 1972). The study analyzed data collected during the 1968 presidential election campaign. One hundred undecided voters were recruited and asked to list key issues facing the country. This particular sample was selected in view of the dominant paradigm at the time. The minimal-consequences perspective suggested that media had little direct effect on individuals, mostly reinforcing existing predispositions and complementing stronger social influences. Undecided voters were presumed to be more susceptible to media messages as they sought information to help them in the selection of a candidate. Investigating the relative salience of items emerging from the responses, McCombs and Shaw uncovered a remarkable similarity between the priorities of news media and those of the voters. The study operationalized the public agenda as a list of the most important problems facing the U.S. at the time, ranked based on the proportion of survey respondents who mentioned each one. The media agenda – perhaps easier to identify in the 1970s than it is today – was compiled based on the number of news stories 10 devoted to each issue. Nine news sources were selected for the project – the CBS and NBC evening broadcasts, Time and Newsweek, The New York Times, and four local newspapers. A systematic content analysis of the coverage produced by those outlets over a period of 26 days was conducted. The correlations between item emphasis on the two agendas were extremely high, both for issues that had received a lot of media attention and for ones that were less prominent. In addition, the study found no evidence of selective perception: participants did not exhibit stronger preference for issues introduced by the political party they belonged to. While its results were suggestive, the Chapel Hill study only demonstrated a correlation between media coverage and public opinion. To examine the direction of influence, the authors conducted a follow-up longitudinal study (Shaw & McCombs, 1977) in Charlotte, North Carolina, during the 1972 presidential election. An analysis of panel data from a sample of registered voters combined with media monitoring confirmed that news coverage of issues preceded voter interest. This supported the major premise of agenda-setting that media stories influence public priorities. The project also examined differences in the capacity of media to set the public agenda: in this particular case, television news seemed to have a greater short-term effect on voters compared to newspapers. Definitions: issue, agenda, and salience Early works (McCombs & Shaw, 1972) put forward a fairly limited conceptualization of the agenda-setting mechanism. It was defined as a process in which 11 the frequent and prominent coverage of certain issues in the news increased the perceived importance of those issues among large segments of the public. In the forty years following the initial Chapel Hill study, the agenda-setting perspective was theoretically extended and applied in a number of different contexts. The definitions of major theoretical concepts evolved to be both more nuanced and wider in scope. In their comprehensive overview of the theory, Dearing and Rogers (1996a) define an agenda as “a set of issues that are communicated in a hierarchy of importance at a point in time” (p.2). The process of agenda-setting is seen as a dynamic interplay among issues competing for the attention of media professionals, policy elites, and the general public. Each individual issue represents a social problem – often, though not always, a controversial one – that has received some media coverage, policy attention, or public recognition. Expanding the definition outside the realm of civic and political participation, McCombs (2004, 2010) suggests thinking of issues on the agenda as attitude objects: anything that draws attention, or any item about which one may hold an opinion. This broader conceptualization goes beyond a focus on public issues to include a virtually unlimited number of entities: topics, public figures, organizations, countries, etc. The process of agenda-setting in this case is broadly defined as the transfer of salient elements from one agenda to another (McCombs & Valenzuela, 2007). This definition is fairly inclusive, as it does not limit the agendas under study to those of media, policy-makers, and the public. The issue priorities of businesses or non- governmental organizations, for instance, fall into the wider scope of this conceptual 12 framework. Research in the area has found that media attention influences perceptions of corporate reputation and product qualities among consumers, employees and investors (C. E. Carroll & McCombs, 2003). Each of the attitude objects on the agenda, furthermore, is associated with numerous attributes. Those item characteristics or traits can also vary in salience. In recent works, McCombs goes on to propose a second level of the agenda-setting theory dealing with the attention paid to object properties (McCombs, 2004, 2005; Weaver et al., 2004). Similarly to the mechanisms defined for object-level media effects, the salience of attributes in news coverage is said to influence public priorities. Thus media are expected not only to focus public attention on particular topics, but also to influence interpretations making some issue features more prominent than others (Merritt & McCombs, 2004). The concept of salience plays a central role in the agenda-setting perspective (Kosicki, 1993). Yet many works do not address it specifically, using it interchangeably with importance, prominence, and attention. Issue salience, both in the news and on the public agenda, has been defined and operationalized in a variety of ways (McCombs, 2005). While media salience is traditionally considered a unidimensional construct, Kiousis (2004) suggests that it consists of three core components: attention, prominence, and valence. Most studies focus on attention, looking at the number of stories dedicated to a particular topic. Occasionally, research also examines prominence measured through indicators like page placement, time or space dedicated to the issue. Valence is more applicable in studies of second-level agenda-setting as it refers to affective attributes of the news (e.g. positive/negative tone, or level of conflict in the story). Conducting an exploratory factor analysis on political issues in The New York Times, Kiousis (2004) 13 found only two dimensions of salience: visibility and valence. Visibility included the combined attention and prominence indicators and accounted for the larger proportion of variance. Typology of agenda-setting research One relatively straightforward categorization of existing agenda-setting research is based on the dependent variables under examination. Dearing and Rogers (Dearing & Rogers, 1996a; Rogers & Dearing, 1988; Rogers et al., 1993) note that the theoretical perspective incorporates studies of public, media, and policy agendas – as well as the relationships among the three. The tradition of public agenda-setting research starts with the Chapel Hill study (McCombs & Shaw, 1972). Originally introduced in the fields of journalism and mass communication, it has been adopted by a wide range of social sciences. This set of studies examines the effect of news coverage on the issue priorities of audience members. The main dependent variable is, accordingly, the public agenda. Its operationalization relies almost exclusively on opinion polls, often incorporating Gallup’s well-known MIP question (“What is the most important problem facing the country today?”). As its name indicates, media agenda-setting research examines the salience of issues in news coverage. The media agenda is typically indexed through content analysis recording the volume and prominence of stories about a certain issue. Studies of this type come out of both mass communication and sociology and investigate a wide range of factors influencing the selection of media content (Kosicki, 1993). While Dearing and Rogers (1996b) refer to this subfield as media agenda-setting, it is also known as agenda- 14 building (Kiousis, Mitrook, Wu, & Seltzer, 2006; Lang & Lang, 1981), or intermedia agenda-setting (Ragas & Kiousis, 2010; Roberts & McCombs, 1994) when the influence of news outlets on each other is examined. Policy agenda-setting studies investigate how issues appear on the political agenda. Mass media coverage is often included as an independent variable. Conducted primarily by political scientists, this type of research seeks to identify factors that shape the priorities of public officials, political parties, and governmental bodies. Typical indicators used to evaluate the policy agenda include introductions of laws, budget appropriation, time spent debating a subject in congress, or the salience of an issue in political speeches and documents. The bulk of agenda-setting scholarship produced since the 1970s has focused on the relationship between media agendas and public perceptions of issue importance (Weaver et al., 2004). The research area now encompasses five distinct aspects: “basic agenda-setting effects, contingent conditions for those effects, attribute agenda setting, origins of the media agenda, and consequences of the agenda-setting process for people’s opinions, attitudes, and behavior” (Valenzuela & McCombs, 2009, p. 90). One helpful typology of existing studies based on their research design was first proposed by McCombs (1981) during an Acapulco conference of the International Communication Association. Under that categorization, agenda-setting scholarship is grouped into four sets which differ along two dimensions. The first differentiation is between studies that focus on a single issue (e.g. tracking the salience of immigration in news coverage and public interest) and research examining multiple issues (e.g. the 15 Chapel Hill study). The second dimension has to do with the type of data and analysis of salience employed by a study. The distinction here is between inquiries using aggregate vs. individual measures of issue priorities. In later works, the resulting four categories of studies are further elaborated and given names: mass persuasion/competition, automaton, natural history, and cognitive portrait (McCombs, Danielian, & Wanta, 1995; McCombs, Holbert, Kiousis, & Wanta, 2011; Valenzuela & McCombs, 2009). One important note here is that the Acapulco typology (Figure 1) is mostly applicable to studies comparing media and public priorities (i.e. what Dearing and Rogers would call public agenda-setting research). The Acapulco Typology Aggregate measures Individual measures Single issue Type I: Mass Persuasion Type II: Automaton Multiple issues Type III: Natural History Type IV: Cognitive Portrait Figure 1. The Acapulco Typology Type I or the mass persuasion is the most traditional among the four groups. It compares the salience of items in the news with an aggregate measure of the public agenda. In more recent works, this category is also known as the competition perspective (Valenzuela & McCombs, 2009) since it examines a number of issues competing for attention. This is the simplest, most straightforward type of research. Much like the 16 Chapel Hill study, the analysis here often involves calculating rank-ordered correlations between the number of news items covering certain topics and aggregated survey responses to the MIP question. Type II or the automaton perspective also examines a set of issues, but the item priorities here are constructed for each individual in the study, not for the public as an aggregate. Studies of this kind (e.g. McLeod, Becker, & Byrnes, 1974) typically find little correspondence between the agendas of individuals and that of the mass media (McCombs et al., 2011). Contrary to the name of this group, research suggests that respondents, when examined at the individual level, do not automatically and completely reproduce the priorities of news outlets (Weaver, 1984). Type III or the natural history perspective includes studies that focus on a single issue and evaluate its aggregated perceived importance. Like the mass persuasion category, research here tends to match the volume of news coverage about a problem and its position on the public agenda. The perspective is named natural history because many studies of this type are longitudinal, examining the surges and waning of attention to a problem (McCombs & Reynolds, 2009). Type IV or the cognitive portrait perspective matches the salience of a single issue on the media agenda with its importance for individual members of the public. For instance, Meijer and Kleinnijenhuis (2006) used content analysis of business news and a nationally representative Dutch panel of respondents to demonstrate the influence of media agenda on individual perceptions of corporate reputation. Other studies in this category are experimental, with individual priorities measured prior to and after 17 controlled exposure to news content. Cognitive portrait research has been instrumental in providing methodologically rigorous support of agenda-setting hypotheses (Wanta & Ghanem, 2007). Setting the public agenda After the initial wave of scholarship confirming the existence of agenda-setting effects, academics started elaborating and expanding the framework. Research examined the contingent conditions determining the strength of the influence news coverage has on individual and group priorities. One of the first questions asked was which media formats were more successful in changing public perceptions of issue importance (Wanta, 1997a). Early on, scholars were interested in the comparative effects of newspapers and television (Shaw & McCombs, 1977). Over time, it became clear that there was no simple, unequivocal result of the comparison between broadcast and print influences (McCombs, 2004). Intermedia processes, issues specifics, social and individual factors could all lead to different findings in that respect. Examining specifically political content, Stromback and Kiousis (2010) found that overall news consumption was a significantly better predictor of issue salience than exposure to any specific source or format. More recently, another important question emerged (one discussed in more detail later in this work): do agenda-setting mechanisms operate in the same way for online and traditional media? A key work in this area was authored by Althaus and Tewksbury (2002) who aimed to fill the gap in existing literature. Their study used an experimental design to evaluate whether readers of the print and Web versions of the New York Times would form different perceptions about the importance of political issues. As expected, 18 there was a consistent main effect: both online and offline readers’ agendas differed significantly from a control group with no exposure to the Times. The effect, however, was stronger for the group reading the print newspaper. The authors concluded that new media had subtle but potentially consequential effects on agenda-setting. Another source of variation in media effects comes from individual and group differences. Those were largely ignored in the early works of the tradition which focused on an aggregated public opinion. As the agenda-setting perspective matured, it incorporated more nuanced micro-level considerations, taking into account psychological characteristics and allowing for individual agency (Shaw, McCombs, Weaver, & Hamm, 1999). It was acknowledged, for instance, that issues which were not perceived as personally relevant could remain low on one’s agenda regardless of their salience in news coverage (Matthes, 2008). Influences from other theoretical approaches were also important. Incorporating elements from the two-step flow of communication model (Lazarsfeld et al., 1944) resulted in works examining the key role of interpersonal discussion in agenda-setting (Brosius & Weimann, 1996; Wanta & Wu, 1992; Yang & Stone, 2003). Drawing on the theory of media system dependency (Ball-Rokeach, 1985; Ball-Rokeach & DeFleur, 1976) allowed researchers to position the agenda-setting process in the context of larger social systems and power relations (Matsaganis & Payne, 2005). Yet other studies elaborated the psychological processes behind agenda-setting and the related priming and framing effects (Scheufele & Tewksbury, 2007). Information processing and cognitive perspectives highlighted the role of learning and affective components in the individual 19 evaluation of political issues (Bulkow, Urban, & Schweiger, 2012; Neuman, Marcus, Crigler, & MacKuen, 2007). While many studies continue to use aggregate measures of “the public opinion”, it is clear that agenda-setting effects are far from uniform across individuals. Demographic, psychological and behavioral variables could influence the process (Wanta, 1997b; Wanta & Ghanem, 2007). Higher exposure to – and reliance on – media were found to enhance the impact of the news agenda on individual priorities (McCombs & Reynolds, 2009; Wanta & Hu, 1994). Political knowledge, which is known to predict media exposure (Neuman, Just, & Crigler, 1992), also seems to increase the magnitude of agenda-setting effects. In two experimental studies, Miller and Krosnick (2000) examined the impact of political expertise and trust in media. While they found that agenda-setting occurred among people at every level of knowledge and trust, the effect was much stronger among the most trusting and knowledgeable. Miller and Krosnick concluded that agenda-setting was a thoughtful, deliberate process, even though weaker effects could also occur automatically and with little cognitive effort. Building on that analysis, several recent studies (Bulkow et al., 2012) set out to identify the conditions under which individuals would make a deliberate attempt to evaluate the importance of issues in the news, as opposed to perceiving it subconsciously based on language and design cues. The results suggested that one’s involvement with the presented issue predicted the level of cognitive effort invested in the process. High issue involvement led to a more thoughtful agenda-setting process, while low involvement resulted in a more automatic one. 20 Perceptions of media credibility are another relevant factor (Dearing & Rogers, 1996a; Wanta, 1997b). Wanta and Hu (1994) found that exposure to, reliance on, and perceived credibility of media, as well as respondent education levels, influenced the magnitude of agenda-setting effects. Formal education was also the demographic variable most consistently found to positively influence individual adoption of media issue priorities across formats and locations (McCombs, 2004). In spite of the recognized importance of personal characteristics in the agenda- setting process, the theory has retained its emphasis on the role of mass communication in achieving consensus among the members of a public. Building a sense of community across social groups is still seen as a key function of the media (McCombs, 1997; McCombs et al., 2011). A number of studies have found evidence to support that claim. Based on a state-wide poll in North Carolina, Shaw and Martin (1992) conducted a comparative analysis of different demographic groups defined by gender, age, race, education, and income. Their study examined the issue agendas of respondents across three levels of media exposure and found that higher news exposure was associated with higher levels of consensus among social groups. For gender, for instance, the correlation between the issue priorities of men and women with low exposure to print news was +.55. The medium-exposure groups had a correlation of +.80, while frequent newspaper readers demonstrated an almost perfect +1.0 agreement between genders. In a more recent study, Coleman and McCombs (2007) found consistent agenda-setting effects across age groups, even though the younger generation relied more heavily on new media, while older respondents preferred traditional news sources. 21 One key psychological variable predicting individual differences in agenda- setting effects is the need for orientation. First introduced in the 1972 Charlotte study (Shaw & McCombs, 1977), the concept refers to the innate desire of people to understand their social environment. Similar ideas such as uncertainty reduction and dependence on media for social cues have been tested and confirmed in other research traditions (Ball- Rokeach, 1998; Ball-Rokeach, Rokeach, & Grube, 1984). The need for orientation construct has two dimensions: relevance and uncertainty. Relevance is an initial condition: it is necessary but not sufficient. The reasons for perceiving an issue as relevant may include personal interest, civic duty, emotional involvement, or peer influence (McCombs, 2004). If a subject is not perceived as personally relevant, the lack of information about it is not likely to cause much psychological discomfort. In such situations, the need for orientation is low, resulting in little attention to news on the topic and weak agenda-setting effects. When an issue is perceived as relevant, the need for orientation depends on one’s level of uncertainty about the matter. High uncertainty about pertinent topics prompts people to monitor closely relevant media content. Thus the combination of high relevance and uncertainty results in the closest match between individual and media priorities. Empirical tests have confirmed that the need for orientation predicts both media exposure and agenda-setting outcomes (Merritt & McCombs, 2004). In more recent work, increasingly sophisticated measures of the construct have been developed and validated (Matthes, 2006, 2008). 22 In retrospect, the near-perfect (+.97) correlation between individual and media priorities found in the initial Chapel Hill study is likely to be partly due to the high need for orientation of the survey respondents. As noted earlier, the sample in that case included only undecided voters polled in the months before a presidential election (McCombs & Shaw, 1972). One important aspect to note here is that the need for orientation is conceptualized as a psychological trait premised on more than just the nature of the issue, personal experience, and circumstances. Individuals may have an innate propensity to a very high or very low need for orientation in any situation (McCombs, 2000). Another construct within the theoretical perspective – that of obtrusive and unobtrusive issues – is specifically designed to reflect personal experience (Coleman et al., 2008). Obtrusive issues are ones that people encounter in their everyday life. For instance, an individual may have a first-hand experience with unemployment, residential burglary, or senior services. Conversely, foreign affairs matters are rarely encountered directly and many people learn about international events exclusively from the news (Wanta & Ghanem, 2007). The agenda-setting perspective predicts a higher degree of correspondence between the media and individual priorities for unobtrusive as opposed to obtrusive topics. When obtrusive issues are high on the public and news agendas (e.g. the economic crisis), this is likely due to exogenous factors rather than media influence. One reason for the sustained academic interest in the agenda-setting process is that it has important social and political consequences. In addition to fostering a shared perception of community priorities, media salience influences the formation and strength of opinions, as well as consequent observable behavior (McCombs, 2010). Press 23 coverage of political candidates, for instance, helps people form an opinion about them which in turn informs voting decisions (Kiousis & McCombs, 2004; Moon, 2011). Using panel data on adolescents in three U. S. states, Kiousis and McDevitt (2008; 2005) examined the impact of agenda-setting on opinion strength, political ideology, and voter turnout. Their results indicated that the process was critical for political socialization, contributing to the formation of political predispositions which in turn increased the likelihood of electoral participation. Another recent study (Kiousis, 2011) found that the media salience of presidential candidates predicted not only their public salience, but also the strength of attitudes held towards them. The priming effect, seen as another outcome of media agenda-setting, refers to the increased accessibility of certain objects or attributes in memory (Roessler, 2008; Scheufele & Tewksbury, 2007). When faced with a necessity to make a judgment, individuals use cognitive heuristics, drawing upon salient objects and attributes instead of examining all available information. The selective attention media dedicate to issues influences the criteria citizens use to evaluate political figures (Iyengar & Kinder, 2010; Sheafer & Weimann, 2005). Building the media agenda In its early years, agenda-setting research focused almost exclusively on the impact of news coverage on public opinion. The media agenda was recognized as a key driver of citizen priorities, but not examined in its own right (Dearing & Rogers, 1996a). In the 1980s, scholars started looking into the factors shaping mass media issue coverage. 24 Media agenda-setting studies – also known as agenda-building research – took object or attribute salience in the news as their dependent variable. McCombs and his colleagues view media agendas as shaped by three major factors: external sources of information, interactions among news organizations, and journalistic norms (McCombs, 2004; McCombs & Valenzuela, 2007). Public officials and communication professionals are among the key external sources capable of increasing the salience of issues in the news. As journalists have finite resources, they often rely on information subsidies (Gandy, 1982): materials obtained from news releases, public announcements, press conferences, briefings, and organized events. Political candidates and press releases, for instance, have been found to significantly influence both object and attribute salience in media coverage (Kiousis et al., 2006; Tedesco, 2001, 2005). In the U.S., the President plays a particularly important role in shaping the news agenda and is generally considered to be the nation’s lead newsmaker (Fahmy, Wanta, Johnson, & Zhang, 2011; Wanta & Foote, 1994). The ability of the state to promote and suppress issues in the news shifts over time in response to social and economic factors. Media historian Dan Hallin (1989) for instance discusses the media-government relationship in the context of U.S. news coverage of the Vietnam War. According to him, in times of crisis media will inevitably follow the government agenda until a collapse of public and political consensus legitimizes a deviation from the official party line. One aspect of the agenda-building process that is particularly relevant here involves the patterns of influence and content exchange among news outlets. Agenda- 25 setting articles often talk about the agenda of media in general terms, even when the analysis examines a single news source. To some extent, this is justified: multiple studies have registered a high degree of similarity in the topics covered by different news outlets (Weaver et al., 2004). A redundancy in source agendas was in fact reported very early on in the foundational Chapel Hill study (McCombs & Shaw, 1972). In it, the issue priorities of five daily newspapers and two television stations had a correlation of +.81. Dearing and Rogers (1996a) go as far as to suggest that the agendas of different outlets can be construed as separate measurements of an underlying general “media agenda” construct. Interestingly, major online newspapers also exhibit agreement in topic salience, both in terms of story volume and placement prominence (Boczkowski & De Santos, 2007; Lim, 2010). The apparent homogeneity of issue priorities is one of the major drivers prompting scholars to study the social and institutional forces shaping news coverage (Reese, 1991). A wide range of factors contribute to the uniformity of story selection across news sources. Among them are (1) the shared professional values and practices of reporters and editors, (2) the tendency of journalists to monitor other media, and (3) exogenous variables (e.g. real world events). The standards instilled by journalism schools, combined with a high job mobility within the industry, mean that news selection is guided by a set of criteria shared by most traditional media. Keeping a close watch on the content of rival news outlets helps validate one’s own news judgment. It can be seen as part of the information surveillance routines of journalism – or as a response to competitive pressures (Shoemaker & Vos, 2009). Mirroring coverage from other sources is also a common way to cope with financial limitations hindering the production of 26 original reporting. This practice is particularly widespread in expensive and resource- intensive areas like international news and investigative journalism. Research studying patterns of intermedia agenda-setting has often focused on the influence wire agencies and elite outlets have over the priorities of other news media. The New York Times, regarded by many as the national newspaper of record, is considered to be particularly important in that process (Reese & Danielian, 1989). This is also why scholars often use its content in as a way to represent the general media agenda. The influence of the nation’s flagship newspaper extends beyond print across media formats. Testing intermedia effects in the area of foreign affairs reporting, Golan (2006) found significant correlations between the international news in the morning New York Times and those of three evening television news programs. Elite newspapers in general are known to set the agendas of local print and television journalism (Protess & McCombs, 1991). An important recent development in intermedia studies is the investigation of interactions between traditional and online news agendas. Yu and Aikat (2005), for instance, found convergence between offline and Web media priorities. Belt, Just and Crigler (2012) uncovered similarities in election coverage across formats (newspaper, radio, television, cable, Internet) both in terms of topics and tone. News values and journalistic standards A number of theoretical frameworks developed in the last forty years have examined the news selection process. In many of its aspects, agenda-building research converges with work coming from sociology, mass communication, and journalism. 27 The literature on news values and newsworthiness investigates the characteristics that make certain stories more likely to be covered by media outlets (Gans, 1979; Tuchman, 1978a). It emphasizes the idea that news is a social construct and its composition depends on more than just the intrinsic properties of a particular event (O'Neill & Harcup, 2009). Stories are, in themselves, not inherently newsworthy – they are perceived as such within a certain context based on a particular set of values. The news values research tradition gained traction in the 1970s. As the limited effects model of mass communication started to lose ground, scholars once again turned their attention to the social significance of news content. This also highlighted the importance of studying newswork. In 1972, Gans wrote an article calling for more sociological research of news production, lamenting what he saw as a "famine" in American media research (Gans, 1972). The works that followed provided a critical examination of journalistic practices, rejecting the idea of news as an objective reflection of reality. Even before Gans made his appeal, Tuchman had launched her investigation into strategic rituals – a set of journalistic routines aimed at reducing bias that often had the opposite effect (Tuchman, 1972). Later on, she coined a metaphor describing story selection as a news net – intended to catch some events, while making sure that others will slip through (Tuchman, 1978a, 1978b). Newsworthy stories (the "big fish" that get caught in the net) are identified by journalists based on assumptions about the public interest. 28 The need to evaluate newsworthiness stems from the physical constraints that journalistic products are subject to – the availability of reporters and editors, the space on a newspaper page, the time allotted to a newscast, as well as the limited attention span of audience members. Newsroom routines and procedures (deadlines, formats, editorial process, etc.) limit the number of stories an organization can produce (Kosicki, 1993). Among the key determinants of news value are characteristics like controversy, novelty, interest, importance, sensationalism, timeliness and proximity (Shoemaker, Chang, & Brendlinger, 1987; Shoemaker & Reese, 1996). Revising a popular taxonomy of news values proposed by Galtung and Ruge (1965), UK scholars Harcup and O’Neill (2001) compiled a list of rules guiding the selection of news. According to them, stories are more likely to be chosen if they: ‐ Involve power elites or celebrities; ‐ Are entertaining or surprising; ‐ Are particularly negative (tragedies, conflict); ‐ Are particularly positive (rescues, cures); ‐ Are significant or impactful; ‐ Are seen as relevant to the audience; ‐ Are follow-up stories on topics already in the news; ‐ Fit the news organization’s own agenda. More recently, Shoemaker and colleagues (Shoemaker, 1991, 2006; Shoemaker & Vos, 2009) generalize the large number of relevant factors and boil it down to two components: social significance and deviance. Social significance measures the level of 29 importance that a society attributes to a particular issue or event. The construct has four dimensions: political, economic, public, and cultural significance. Presidential elections, for example, have a high political significance, while unemployment ranks high on the economic dimension. Health and education have high public significance, and matters of religion are an example of a culturally significant issue. The second component of newsworthiness, deviance, is conceptualized as having three dimensions: statistical, normative, and social change (Shoemaker, Danielian, & Brendlinger, 1991). Statistical deviance is a property similar to the qualities of novelty or oddity identified in earlier studies. It is high for events that are unexpected or unusual. Normative deviance describes the extent to which laws or social norms have been violated. Finally, social change deviance is related to the probability that an event could change the status quo. One explanation for the perceived newsworthiness of deviant events comes from evolutionary biology (Shoemaker, 1996). People are hardwired to observe their environment looking for irregularities which may signal the presence of danger. Monitoring one's surroundings for potential threats increases the likelihood of detecting relevant changes on time and avoiding incidents that may disrupt the status quo. This is why, for instance, one major factor that makes foreign affairs stories newsworthy is the extent to which a potential threat to the United States is perceived (Golan, 2008). At the individual level, two more factors emerged as important aspects of a story’s news value (Shoemaker & Cohen, 2006; Shoemaker, Johnson, Seo, & Wang, 2010). One was complexity, with more complex events getting more attention than simple 30 ones. The other was personal relevance, or the extent to which a story is perceived as engaging or having significance to the reader. A number of contextual predictors related specifically to the presence of foreign countries in the U.S. news have also been identified in existing research. Those include traits of the nation (population, GDP, political importance and others), geographic and cultural proximity to the U.S., as well as trade relationships, frequency and volume of economic interactions (H. D. Wu, 2007). Other culture-related variables include dominant religion, common ancestry and level of press freedom in the country (Wanta, Golan, & Lee, 2004). The deviance dimension is also very relevant in this case: conflicts, wars, and disasters abroad tend to receive media attention (Semetko, 2009). An often-overlooked aspect of news selection involves the competition for attention among issues (Rogers & Dearing, 1988). An evolutionary approach to agenda- building (Monge, Heiss, & Margolin, 2008) would construe the news value of stories as dependent on traits allowing some items to proliferate while others are filtered out in a variation-selection-retention process. As a consequence, whether an issue will be covered depends not only on its own newsworthiness, but also on that of the stories it is competing against. Djerf-Pierre (2012) calls this a crowding-out effect. In a longitudinal study of Swedish television content, she demonstrates that in times of crises, economic stories and war coverage absorb media attention, “crowding out” environmental news. This displacement of issues is likely to be less pronounced on the Web, where stories still compete for attention, but without the added restrictions of limited page space or broadcast time. 31 Gatekeeping: levels of analysis Gatekeeping is another key paradigm studying the construction of media content. The theory, as conceptualized in the field of mass communication, examines “the process by which the vast array of potential news messages are winnowed, shaped, and prodded into those few that are actually transmitted by the news media.” (Shoemaker, Kim, & Wrigley, 2001). Like agenda-setting, this perspective deals with the way social reality is presented in media coverage. A major contribution of gatekeeping research is the identification of levels – or gates – where news selection happens. Bennett (2004) points to the need to specify a gatekeeping model capturing the complex interactions between personal, organizational and economic factors in a rapidly changing social and technological environment. The four main gates he proposes include (1) individual news judgment and values, (2) bureaucratic or organizational news- gathering routines, (3) economic constraints on news production, and (4) information and communication technologies. Bennett puts emphasis on the fact that the four factors do not operate separately but rather interact with each other to shape news coverage. Moreover, both the nature of its influence and the relative importance of each factor are changing over time. In recent years, for instance, economic pressures on journalism and technological developments have been two particularly powerful forces driving the construction of news. In their most recent book on the subject, Shoemaker and Vos (2009) describe five levels of analysis where gatekeeping processes may occur. The individual level examines demographic and personal characteristics of news workers – for instance their political 32 leanings. The routine level deals with practices and standards typical of journalistic work across individuals and media companies. This includes common procedures such as following deadlines, as well as key principles such as the news values determining the prominence of objects in coverage. Routines are expected to influence the selection of news more than individual characteristics of reporters (Shoemaker & Reese, 1996). News values and practices explain the homogeneity of media agendas produced through the work of individuals with potentially very different personal priorities. The organizational level captures properties of media companies: ownership, structure and size, among others. The social institutions level looks into external factors relevant to the media industry – audiences, advertisers, political institutions, and interest groups. The last, macro-scale area of analysis examines features shaping news selection at the social system level. Studies at this level explore gatekeeping controls imposed by a country’s economic, political, or cultural system – including the impact of prevailing ideology on news selection. Like many conceptual frameworks in the field of media studies, the gatekeeping perspective is changing in order to account for the social impact of new technologies (Dimitrova, Connolly-Ahern, Williams, Kaid, & Reid, 2003; Shoemaker et al., 2010). The nature and consequences of this shift are discussed in Chapter 3. The same chapter contains an important related discussion of a key new influence on the media agenda: bottom-up effects coming from the public. 33 CHAPTER 2: CURRENT TRENDS IN AGENDA-SETTING RESEARCH As this work attempts to address the conceptual and methodological challenges facing the agenda-setting perspective, one important preliminary task is to capture the current state of the field. Beginning with McCombs and Shaw’s (1972) classic work, the number of studies referring to the theoretical tradition has grown steadily, reaching a peak in the late 1990s (see Figure 2). During that time, several academic articles provided quantitative overviews of the literature in the field, as well as meta-analysis of key relationships predicted by the theory. Figure 2. Mentions of agenda setting in books, 1940-2008. Source: Google NGram. Twenty years ago, Rogers, Dearing and Bergman (1993) examined academic publications including one or more variables pertinent to agenda-setting. They collected and reviewed 223 articles published in the 70-year period between 1922 and 1992. From those, 131 (59%) papers studied the relationship between media and public priorities, and 34 65 (29%) examined policy agenda-setting. Only 15 (7%) of the works were found to focus specifically on media agendas. Investigating trends in mass communication literature over time, Bryant and Miron (2004) collected a probability sample of articles from three academic journals: Journalism & Mass Communication Quarterly, Journal of Communication, and Journal of Broadcasting & Electronic Media. A total of 1,806 papers published between 1956 and 2000 were analyzed. Agenda-setting tied with the uses and gratifications approach (E. Katz, Blumler, & Gurevitch, 1973) for the first place as the most cited mass communication theory. In a meta-analysis of agenda-setting effects, Wanta and Ghanem (2007) identified 90 relevant works published between 1972 and 1996. The mean correlation between object salience in the news and public opinion across the studies was significant, r = .53. More recently, Tai (2009) conducted a bibliographic analysis of the field, constructing a citation network of agenda articles. He examined 56 works published between 1996 and 2005 and used co-citation analysis to group them in area clusters. While their results were revealing, none of these studies provided a recent snapshot of all parameters of agenda-setting research relevant to this work. Consequently, a preliminary descriptive analysis of the literature in the field was completed here. The results are presented in this chapter. The main purpose of this quantitative overview is to summarize characteristics of recent agenda-setting works, including major analytical approaches and emerging themes. This was a necessary step prior to addressing key conceptual and methodological issues of the theory. 35 The academic journals included in the analysis were selected with a view to capturing the most prominent research in the field of communication studies. They included the six major journals published by the International Communication Association (ICA): Communication, Culture & Critique, Communication Theory, Human Communication Research, Journal of Communication, Journal of Computer-Mediated Communication, and Political Communication. The second set of publications selected for examination contained the five highest-ranking journals in the communication field, as determined by their Web of Science impact factor: Communication Monographs, Journal of Communication, Journal of Computer-Mediated Communication, Public Opinion Quarterly, and Public Understanding of Science. As two of those are also published by ICA, the total number of journals reviewed here was nine (see Table 1). To get a sufficiently comprehensive sample reflecting recent shifts in the field, the review included works published in the twelve years between January 2000 and December 2011. During that time, a total of 710 articles containing the phrase “agenda setting” or “agenda-setting” appeared in the nine journals. Upon examination, 116 of those studies were found to substantively employ the theory or some of its aspects. This final set of works was examined and the main focus, data collection strategy, analytical methods, themes and other attributes of each article were recorded. 36 Table 1. Agenda-setting articles by journal, 2000-2011. Journal Articles mentioning agenda-setting Articles focusing on agenda-setting Communication Monographs 55 2 Communication, Culture & Critique 41 7 Communication Theory 112 11 Human Communication Research 48 1 Journal of Communication 202 43 Journal of Computer-Mediated Communication 104 5 Political Communication 63 24 Public Opinion Quarterly 33 5 Public Understanding of Science 52 18 TOTAL 710 116 The largest number of agenda-setting works (n = 43, or 37% of all articles) came from the Journal of Communication, followed by Political Communication (n = 24, or 21%). 37 Figure 3. Number of agenda-setting articles per journal. The yearly number of articles ranged between one in 2001 and 19 in 2007. No consistent trend towards a decline in the volume of published works employing the agenda-setting perspective was recorded. Figure 4. Number of agenda-setting articles published yearly. 43 24 18 11 7 5 5 2 1 Agenda‐Setting Articles by Journal (2000‐2011) Journal of Communication Political Communication Public Understanding of Science Communiction Theory Communication, Culture & Critique Journal of Computer‐Mediated Communication Public Opinion Quarterly Communication Monographs Human Communication Research 12 1 6 8 9 8 5 19 8 16 15 9 0 4 8 12 16 20 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Agenda‐Setting Articles by Year (2000‐2011) Number of Articles 38 Out of the 116 articles, 82% examined a media agenda – either on its own, or studying its interplay with public, policy, or activist agendas. The average number of media outlets included in those studies was 3.4. Over a third (35%) of all articles examined the relationship between media and citizen priorities, while 13% studied the link between media and policy agendas. Another 32% focused exclusively on media. Close to a third of the works in that media-only group (11% of all articles) investigated inter-media agenda-setting patterns, or the influence of some news outlets on others. Figure 5. The focus of agenda-setting articles (media, public, policy, activist agendas). Out of the subset of articles that discussed media agendas (alone or in combination), 59% examined newspapers, 34% studied television programs, and 15% looked at online news sites or blogs. Those categories were not exclusive since some works explored multiple media formats. Media + Public Agenda 35% Media Agenda Only 32% Media + Policy Agenda 13% Other combinations (media,public, policy,activist) 20% Agendas under Examination (All Articles, 2000‐2011) 39 Figure 6. Formats examined by agenda-setting articles focusing on the media agenda. About a quarter (26%) of the 116 agenda-setting works included here were purely theoretical and contained no data analysis. The other 74% were empirical studies: 55% quantitative, 17% qualitative and 2% using mixed methods. Figure 7. Theoretical and empirical agenda-setting articles. 59% 34% 15% 4% 3% 0% 20% 40% 60% 80% Newspapers Television New Media, Blogs Movies Wire Agencies Types of Media Examined in Articles Discussing Media Agendas (2000‐2011) Percentage of articles (Base: Articles examining media agenda) Empirical, Qualitative 17% Empirical, Quantitative 55% Empirical, Mixed Method 2% Theoretical Articles 26% Theoretical vs. Empirical Agenda‐Setting Research (2000‐2011) 40 The most popular data collection procedure used in empirical papers was content analysis alone (43%), followed by surveys (25%). Another 13% of the articles followed the model of McCombs and Shaw (1972) and used a combination of survey and content analysis. Figure 8. Data collection strategies used in the empirical agenda-setting articles. Content Analysis 43% Survey 25% Survey + Content Analysis 13% Interviews 8% Citation Analysis 4% Case study 4% Other 3% Data Collection Strategy (Empirical Agenda‐Setting Papers 2000‐2010) 41 Almost half (46%) of the 67 quantitative studies in the sample employed regression as their main analytical tool. The next most commonly used method was correlation analysis (16%), followed by difference tests (10%) and purely descriptive approaches (10%). Figure 9. Data analysis methods used by empirical quantitative agenda-setting articles. Most of the articles in the sample (n = 90 or 76%) focused on a specific theme or a subset of issues rather than examining the full list of all media, public or policy Regression 46% Correlation 16% Difference tests 10% Descriptives 10% ANOVA 6% Correspondence Analysis 3% Structural Equations 3% Time Series 3% Other 3% Data Analyis Method (Quantitative Agenda‐Setting Papers 2000‐2010) 42 priorities. The areas under study included politics (38%), environmental issues (9%), crime and justice (8%), and science (7%), among others. Figure 10. Major issues examined in agenda-setting articles. The last piece of information about each article in the sample – number of citations – was collected from Google Scholar (Google Inc, 2012). The Journal of Communication had the highest average number of citations per article (70 citations/article), followed by Communication Monographs (64 citations/article). Theoretical articles had a higher number of citations (83 citations/article), compared to empirical quantitative (36 citations/article) and qualitative (21 citations/article) research. Articles that examined both the public and media agendas (74 citations/article) were also cited more often than ones focusing on media and policy agendas (56 citations/article ) or media only (35 citations/article). 38% 9% 8% 7% 3% 3% 3% 3% 3% 2% Politics Environment / Environmental Hazards Crime / Justice Science Foreign Affairs War / Terrorism Youth / Education / Literacy Health Economy / Business / Employment Same sex marriage Main Issue Examined in the Article (2000‐2010) 43 Figure 11. Average number of citations to agenda-setting articles per journal. The overview of agenda-setting works published in the last 12 years reveals a number of gaps in existing literature, further discussed and addressed in this work. The results suggest that the agenda-setting perspective is still quite relevant in the field of communication studies. There is, furthermore, no pronounced trend towards a decline in the number of articles on the subject over time. Within the research tradition, however, relatively few empirical works deal with the formation of media agendas and the factors predicting diversity or homogeneity of content across outlets. The majority of those articles examine a small number of outlets, often limited to a single sector (typically print or television). Media-level studies are also less cited than research following the classic formula of comparing news and public priorities. The dominant methods employed in empirical works (difference tests, correlations, regression analysis) are not too helpful in testing the patterns of influence among media. Assumptions of independence among cases preclude the examination of 71.2 64.0 53.5 38.1 34.3 27.0 10.4 8.4 6.1 Average Number of Citations by Journal (2000‐2011) Journal of Communication Political Communication Public Understanding of Science Communiction Theory Communication, Culture & Critique Journal of Computer‐Mediated Communication Public Opinion Quarterly Communication Monographs Human Communication Research 44 dyadic and higher-order relations and mechanisms. In order to conduct this type of analysis, the present work proposes an alternative network approach, which is outlined in Chapter 4. 45 CHAPTER 3: CONTEMPORARY CHALLENGES TO THE AGENDA-SETTING PERSPECTIVE While it is still the dominant academic paradigm dealing with topics in the news, in recent years the agenda-setting framework has become increasingly problematized. This chapter discusses two major types of challenges faced by agenda-setting research at present: Conceptual challenges: Transformations in the media system have prompted scholars to re-examine the relevance of traditional mass communication theories in a changing information environment. New digital media formats, user-generated content, proliferation of distribution channels and fragmentation of audiences have called into question the cohesion of media agendas and their effects on the priorities of the public. Opposing theories suggest fragmentation (Bennett & Iyengar, 2008) or, conversely, homogenization (Boczkowski & De Santos, 2007) of news coverage in a digital age. Policy Challenges: Studying agenda-setting is a task with increasingly important regulatory and policy implications. The need to understand the process is even more pronounced in light of recent social and technological shifts. Major regulatory concerns in this area revolve around the evaluation and promotion of media diversity (including the diversity of sources, viewpoints, and consumption patterns). While agenda-setting scholarship has significant policy ramifications, few works in the academic tradition address ongoing policy concerns directly (Napoli & Gillis, 2008). 46 In the following sections, the two types of challenges are examined in turn. First, the literature claiming substantial fragmentation in media production and consumption is summarized. Its major premises are outlined, identifying trends viewed as contributing to a presumed disintegration of the agenda-setting process. The chapter goes on to argue that virtually no empirical evidence supports the fragmentation hypotheses. In fact, research results quite consistently point in the opposite direction, leading to the conclusion that agenda setting is still taking place – and mainstream media sources continue to influence public opinion. The final section of the chapter provides an overview of regulatory objectives related to media diversity and discusses the contributions of agenda-setting studies to policy-oriented research. Conceptual Challenges: Agenda Fragmentation Back in the 1970s when the agenda-setting framework was conceived, researchers operated under the conditions of a relatively uniform, centralized media system. Three major television networks and a limited number of influential newspapers and magazines were credited with the shaping of public opinion (Bennett & Iyengar, 2008). A number of major technological and social transformations happening in the last two decades have radically changed the practices of media production, dissemination and consumption. Deregulation in the media sector along with the advent of digital technologies lowering the cost of content production facilitated the emergence of countless news outlets. Information overload, proliferation of distribution channels, and a 47 perceived shift of power from corporations to users characterize the media landscape of the 21 st century (Castells, 2005). Following those shifts, scholars began to question the relevance of current approaches to mass media (Chaffee & Metzger, 2001). Authors suggested that the agenda-setting paradigm needs to be reevaluated in view of the changing information environment (Bennett & Iyengar, 2008; Bennett & Manheim, 2006; Blumler & Kavanagh, 1999). The basic premise of McCombs and Shaw’s theory is that a limited number of outlets with converging agendas and shared journalistic culture are shaping the public perception of issue salience. A virtually unrestricted access to diverse content- delivery services can violate that assumption, compromising the media’s consensus- building function (Takeshita, 2006). As patterns of news consumption grow increasingly personalized (Tewksbury, 2005), a shared set of community priorities may become difficult to envision. In the extreme case outlined by Sunstein (2007, 2009), selective exposure would split mass audiences into many isolated like-minded groups, leading to “cyberbalkanization”. As journalists also have access to the same (or greater) abundance of information, the news agenda itself may also be threatened by fragmentation. The presumed disintegration of the media agenda is partly due to the expectation that the growing number of new outlets will disseminate content that reflects multiple, varied sets of priorities. This proliferation of available viewpoints would also allow for higher selectivity in one's exposure to media content. One related problem is what Bennett and Iyengar (2008) call the demise of the inadvertent audience and economists refer to as the unbundling of newspaper content (Kaye & Quinn, 2010). In the past, a limited set of media sources offered different types of content packaged together. As a 48 result, a large part of the news consumption back then was not purposeful. People were watching the evening newscast while waiting for the entertainment part of the network program – or browsing through the pages of a newspaper after reading the sports section. Today, this is no longer the case. The improved capacity of consumers to select preferred messages seems to reduce the likelihood that a coherent public agenda would form. This could decrease social cohesion and lead to a segmentation of audiences – a possibility articulated by Katz and others more than two decades ago (Blumler & Kavanagh, 1999). Figure 12. Conceptual challenges to the agenda-setting process. 49 In summary, the emerging trends viewed as potentially disruptive to the agenda- setting process can be narrowed down to the following three propositions (also presented on Figure 12): 1. Audience fragmentation: Selective exposure has made the mainstream media agenda irrelevant. The news audience is scattered across multiple information sources (including traditional media, but also user-generated content like blogs, tweets, YouTube videos, etc.). Those countless sources cover a very diverse set of issues, fragmenting the public agenda. 2. Bottom-up effects: The new information environment has effectively reversed the direction of influence in the agenda-setting process. Mainstream news coverage is now following the agenda of civic media and the networked public. 3. Media fragmentation: Technological advancement and increased information availability provide journalists with an unprecedented, diverse selection of issues that they may choose to cover. Relevant data, as well as work done by citizen journalists or non-profits, are readily available online. As a result, the news media agenda is fragmented – there is little consensus among outlets about the important issues of the day. While the media industry is undoubtedly changing in response to technological shifts, economic pressures and new regulation, there are reasons to believe that agenda- setting processes are still taking place. Changes in the nature of journalism and modes of news consumption are undoubtedly altering the way issues enter the media agenda and reach citizens. Yet newspapers, television programs, radio stations, and mainstream 50 sources of online news continue to play an important role in the formation of public opinion. The next three sections of this chapter provide support for that claim, addressing each of the three major potential disruptions in the agenda-setting process summarized above. One thing that should be noted here is that demonstrating the relevance of agenda- setting in a digital age is only a secondary task of this study. The main focus of the work is to propose and test a conceptual/analytical approach suitable for the examination of these questions in a new information environment. At present, the bulk of existing literature discusses normative consequences on the basis of speculation rather than evidence (Sayre, Bode, Shah, Wilcox, & Shah, 2010). There is a pressing need for more methodological and empirical work in the area (Webster & Ksiazek, 2012). The present project constitutes one attempt of that kind. To this end, it develops a framework applying network analytical tools to the study of media and public priorities, outlined in Chapter 4. Fragmented public agenda: selective exposure While the impact of the Web on media use is certainly unprecedented, its potential to cause audience fragmentation may not yet be fully realized. This notion is supported by a body literature discussing the attention economy (Goldhaber, 1997). According to works taking this perspective, the trend towards information proliferation is countered by an attention scarcity. Individuals are limited in the amount of time they can spend consuming media products. When audience members have to make a choice, they typically select bigger, easier to find, better-known news outlets (Nagler, 2007). 51 Furthermore, the information people are interested in seems to be fairly consistent, both within and across social groups (Neuman, 1991). Attention concentration patterns are also emerging online, where search engine ranking mechanisms often determine which sites will receive most of the attention (DiMaggio, Hargittai, Neuman, & Robinson, 2001). Academic investigations support the idea that online attention is highly concentrated rather than scattered across content providers. While the Internet offers an enormous number of information sources, people tend to cluster around a select few (McCombs, 2005). The high concentration seems to follow the 80-20 rule with 80% of the traffic going to less than 20% of the online news venues (Webster & Lin, 2002). Popular new sites are, furthermore, predominantly owned by large media companies. A study by the Berkman Center for Internet and Society (Miel & Faris, 2008) suggests that traditional media outlets are among the winners in the battle for consumer attention. Miel and Faris also claim that “the evidence to date has not confirmed any linkages between online news consumption and fragmenting audiences” (p. 33). Recent research reports confirm that the websites of traditional news organizations – particularly newspapers and cable TV stations – dominate the online information space (Pew Project For Excellence in Journalism, 2010c). Out of 4600 news and information sites examined by the Pew study, the top 7% attracted 80% of the visitor traffic. Legacy media accounted for two-thirds of the user visits. A 2011 report points to similar conclusions: out of the top 25 online news sources for 2010, 18 belonged to traditional media outlets (Pew Project For Excellence in 52 Journalism, 2011). Of the remaining 7 sites, 5 generated the majority of their traffic by aggregating mainstream media content. Similar trends have been recorded for online platforms like Twitter: a limited number of elite actors command the attention of social media users (S. Wu, Hofman, Watts, & Mason, 2010). The most prominent Twitter accounts belong to traditional media and celebrities (H. Kwak, Lee, Park, & Moon, 2010). Popular web sources of local coverage are likewise limited in number and mostly affiliated with traditional media. In an examination of the top 100 television markets in the U.S., Hindman (2011) identified 1074 local news sites and analyzed their audience reach and traffic volume. A large proportion of the markets had fewer than a dozen local news websites, most of which affiliated with TV stations or daily newspapers. TV and newspaper sites also attracted larger audiences, while independent online venues were less popular. Eighty-four of the 100 markets did not have any online-only sources of local news reaching over 1% of the audience. Assuming that public attention is concentrated on a limited number of outlets, it becomes important to evaluate the distribution of audience members across those venues. Individuals may seek a wider selection of news, or could be divided into isolated consumption communities as suggested by the fragmentation hypothesis (Sunstein, 2007, 2009). Using Nielsen data on television and Internet use, Webster and Ksiazek (2012) set out to examine audience fragmentation across 236 media outlets. The researchers point out that works envisioning a highly segregated world of niches and microcultures rarely 53 have direct empirical evidence. The study results paint a different picture of media realities, uncovering high levels of duplication of audiences across media outlets. The analysis reveals overlapping patterns of public attention rather than trends towards isolation in like-minded consumption groups. One of the most often discussed dividing lines in media consumption practices is grounded in political ideology. While some studies find evidence of ideologically motivated selective exposure (Stroud, 2011), others suggest that partisan preferences do not lead to selective avoidance (Holbert, Garrett, & Gleason, 2010). People may seek out outlets that support their views, but they do not actively avoid opposing opinions. In fact, individuals still seem to consume a relatively broad range of media (Webster, 2005, 2007), including sources that may carry contradictory viewpoints. Controlling for ideology, Holbert, Hmielowski, and Weeks (2012) found a strong positive association between the use of politically oriented cable networks like FOX News and MSNBC. Gentzkow and Shapiro (2010) investigated the role of the Internet in the ideological segregation of news consumption. They found that segregation in online news, while higher than that of most offline media use, was low in absolute terms. Online fragmentation of political news exposure was also significantly lower than that in face-to- face interactions with neighbors, co-workers, or family members. The study found no evidence that online media exposure is becoming more fragmented over time. When examining the potential of online news to fragment public opinion, it is crucial to also take into consideration the overall patterns of media use in the U.S. As of April 2012, 18% of American adults did not have Internet access and so relied solely on 54 traditional media and interpersonal contacts to stay informed (Pew Internet & American Life Project, 2012). Another 16% reported having some access, but did not have a broadband connection at home. The percentages were much higher for certain ethnicities and age groups. In 2010, 34% of the public got news online, compared to 58% who watched television news (Rosenstiel, 2010). Relatively few Americans (9%) relied exclusively on the Internet for news, while 39% used only traditional sources and 36% consumed a mix of the two. The average time spent with news on a given day was 57 minutes for traditional media (TV, radio, newspapers) and 13 minutes for the Web. Twitter, one of the most popular online platforms used for information dissemination and mobilization, was actively used by 8% of the U.S. Internet population (Smith & Brenner, 2012). Fox News alone had the regular attention of 23% of Americans (Rosenstiel, 2010) – not to mention a set of messages that are far more consistent and deliberate than a typical Twitter stream. As Althaus and Tewksbury (2002) point out, “to the extent that audiences continue to rely on traditional sources of news, the potential for online media to isolate people in highly personalized information environments will remain limited” (p.197). Theories of complementarity in online and offline news consumption suggest that individuals may continue to rely on a mix of both (Dutta-Bergman, 2004). If audience attention is concentrated on a limited number of news sources, the influence of news on public opinion can be expected to serve its traditional consensus- building function. That is, of course, provided that the media agenda itself still exhibits the homogeneity registered in early studies (Dearing & Rogers, 1996a) – a point discussed later in this chapter. 55 Reversed direction of influence: bottom-up effects The potential of digital audiences to influence the agenda of mainstream news sources is a major discussion topic in recent media studies. Much of the research in this area has focused specifically on political blogs, which are often deemed a new space for public deliberation. Several informative Pew PEJ reports (2010b, 2010c) provide some context to the relationship between user-generated content and “old-style” journalism. Monitoring and analysis performed on millions of blogs over a period of one year revealed that 99% of the stories they linked to came from traditional media. Furthermore, fully 80% of all blog links to news articles pointed to only four websites – those of BBC, CNN, the New York Times and the Washington Post. In another study of hyperlink use and citizen media, Reese et al (2007) found that the blogosphere relied heavily on professional journalists and news organizations. The authors concluded that “bloggers, for the most part, simply engage the facts and information carried in news accounts, accepting them at face value and using them to form their own arguments, reinforce views, and challenge opponents.” (p.257) Works relying on content analysis rather than linking patterns uncover similar trends. A study analyzing a sample of 120 blogs found that 73% of the sources they cited came from some form of news media (Messner & DiStaso, 2008). Comparisons between the agenda of political blogs and mainstream outlets also reveal a considerable similarity between the two, regardless of the bloggers’ political convictions. In a content analysis of 56 news stories and blog posts, Lee (2007) finds a +.79 correlation between conservative blogs and traditional media agendas and a correlation of +.73 for the liberal blogs. The fact that user-generated political content largely mirrors the agenda of traditional media is not surprising. As citizen journalists lack the resources available to news corporations, it can be expected that their work will often rely on news media. While they do reframe issues, supply new angles and add details, mainstream sources often dictate the topics that are discussed. Putting this in agenda-setting terms, researchers have found that professional journalism sets the issue agenda of user- generated content, while citizen posts may influence the second-level attribute agenda of news coverage (B. Lee, Lancendorfer, & Lee, 2005). Another explanation comes from the professionalization of citizen media and its assimilation into the mainstream. The popular blog format, once denoting independent opinions, has been co-opted and “normalized” by journalists (Singer, 2005). Completing the convergence from the other side, successful blogs have entered the mainstream, becoming increasingly professionalized. Blogs with large staff, established editorial processes, and advertisement revenue streams, have come to be almost indistinguishable from journalistic venues. Many blogs, now commercially run, have joined large media networks. The Huffington Post, one of the most popular left-wing representatives of the format, was sold to AOL in 2011, thus officially joining the mainstream. With the maturing of the blog format, the academic view of those outlets has also become normalized. The impact of blogs today can be conceptualized as another routine 57 – rather than unruly or disruptive – force in the intermedia agenda-setting process (V. Campbell, Gibson, Gunter, & Touri, 2009). All of this is not to say that members of the public cannot or do not put items on the media agenda – they can and they do. There is more than anecdotal evidence of emergent bottom-up effects (Wallsten, 2007). In several notable cases (e.g. the Trent Lott controversy, the Killian memo), individual bloggers have succeeded in generating considerable public attention to stories that might have otherwise gone unnoticed. Blogs, online social network services, and citizen journalists have brought to light important issues, provided relevant insights, and shared first-hands accounts of events. Recognizing this, however, we also need to acknowledge that it is not the norm. The primary direction of influence remains from mainstream sources to blogs and social network platforms. Furthermore, when alternative sources do play a role in the agenda- setting process, they are rarely a first-hand driver of opinion formation or social change. Due to their limited audience base, citizen-based media rarely reach the general public directly. Instead, stories have to be picked up by large news outlets, thus becoming part of the mainstream process of issue selection and prioritization. Farrell and Drezner (2008) make a related point – according to their research, there is strong evidence that media elites (editors, publishers, reporters, and columnists) pay attention to political blogs. It is through that editorial filter that issues brought up by bloggers may reach a larger audience. One revealing example of the indirect way in which bottom-up effects operate comes from the Arab Spring. In 2009, U.S. Internet users managed to put the elections in 58 Iran high on the media agenda. As Iranian authorities drove many foreign correspondents out of the country and put others under home arrest, television news networks were slow to provide election coverage. Tagging their messages with #CNNFail, Twitter users managed to convey their frustration with the lack of attention the elections were getting from the major cable network. Within a day, CNN had mobilized its resources to provide reporting from Tehran. In addition to being a new media success story, this case still demonstrates a degree of reliance on mainstream media. Even as bloggers were writing about the elections in Iran, the first instinct of online users was to seek real-time in-depth coverage on CNN. Furthermore, the way Twitter put Iran on the public agenda was indirect: it did so going through a traditional news outlet. Scholars have argued that the array of new information sources has weakened considerably the agenda-setting power of traditional media and eliminated its gatekeeping role (Williams & Delli Carpini, 2000, 2004, 2011). Yet there is an interpretation under which those diverse sources are just another one on the long list of exogenous factors influencing news coverage. There are still, as there have always been, many other media agenda-setters, including elite news outlets, public figures, corporate lobbies, and experts. Blogs, which these days are cited by journalists as an indication of the public opinion, constitute a very small percentage of all sources quoted in the news (Messner & Garrison, 2010). As long as both citizen media and the majority of Americans rely on news organizations for information, discarding the filtering function of journalism seems premature. 59 Fragmented media agenda: more voices, more diversity? Research on audience attention discussed above suggests that issues made prominent by traditional media are likely to end up on the public agenda. A key question for agenda-setting research remains: can the media agenda still be captured by examining a limited number of prominent news sources? Can we talk about convergent issue priorities in this new information environment? Do recent trends point to fragmentation, or are there stronger underlying factors that promote cohesion? Theoretical papers have outlined some challenges to studying agenda-setting in a decentralized media system. Still, little or no empirical research has so far supported the idea of news agenda disintegration which has remained mostly theoretical. In a recent book on news and public opinion, McCombs, Holbert, Kiousis, and Wanta (2011) make a strong statement going in the opposite direction: To take another, more recent example of media convergence, the media agendas of issues available to citizens now use both online and traditional media. These news agendas remain essentially as homogeneous today, however, as they were four decades ago during the golden age of television and newspapers. (p.11) While it seems difficult to defend the assertion that the homogeneity of news agendas has not declined since the 1970s (not least because the boundaries of acceptable news discourse have shifted), the study presented here adopts a weaker version of that claim. There seems to be enough consistency in the priorities of mainstream outlets in certain issue domains to justify the study of agenda-setting processes. Even in pre- 60 Internet times, the idea of a single media agenda was no more than a useful simplification allowing scholars to generalize findings made in a specific context. There were always multiple news agendas and issue publics that did, however, overlap in their major concerns within selected areas. It is likely that this is still the case in a new media environment. Studies employing traditional agenda-setting methods have uncovered convergence of online news outlet issue priorities, both in terms of attention and prominence (Aikat & Yu, 2005; Lim, 2010). Research has also found relatively strong agenda-setting effects on public opinion, even among younger people and heavy internet users (Coleman & McCombs, 2007). Many of the dynamics contributing to uniformity in news selection remain relevant today. Journalism education instills professional values that turn into accepted industry-wide standards. Job mobility in the sector promotes the spread of consistent norms and practices. Organizational routines – evolving as traditional sources enter the digital domain – are still key when it comes to issue priorities (Shoemaker & Vos, 2009). Reporters continue to track the work of their colleagues – even more carefully than they did in pre-Internet times – in order to validate their own decisions (Valenzuela & McCombs, 2009). The accelerated news cycle and the demands of fast-paced online journalism create additional pressures towards homogenization of content as little time is left for research and original reporting (Boczkowski & De Santos, 2007; Mitchelstein & Boczkowski, 2009). Elite publications have retained much of their influence – the New York Times can still confer legitimacy to an issue and trigger a public discussion around a topic that would have otherwise been ignored (McCombs et al., 2011). 61 The influence of elite publications, furthermore, seems to persist in online settings. In a study of hyperlinks patterns of news websites, Weber and Monge (2011) demonstrate the ability of key outlets to control the online flow of information. Their model identifies two important sources, The Associated Press and Reuters, feeding information through elite media like the New York Times and the Los Angeles Times to aggregators such as Google News and the Huffington Post. In a book-length study of news homogenization, Boczkowski (2010) examined the dynamics of imitation among media outlets and the facilitating role of technology. Analyzing online and print publications in Argentina, he found that the content overlap between news sources was not only pronounced, but also increased over time. Interviews with journalists suggested that the Internet had enhanced this trend in a number of ways. As traditional news media developed online presence, the monitoring of competitors’ content became much easier and less time-consuming. Real-time online updates made it possible to track rival news stories throughout the day instead of having to wait for the morning paper or the evening newscast. According to Boczkowski’s findings, the monitoring and reproducing of news from other media has turned into a newsroom requirement. The practice was institutionalized – an organizational routine rather than an individual shortcut a reporter might use to save time. This was particularly true of current affairs coverage. It was less pronounced in soft news journalism, perceived as more selective and not as time-sensitive. Another factor promoting cohesion in news agendas can be seen in the ongoing trend towards media concentration on a global scale. A relatively small group of multi- national media conglomerates own a large number of high-profile news production sites. 62 Those international corporations are well connected through a dense web of partnerships, cross-investments and personnel (Arsenault & Castells, 2008b). As the companies seek economies of scope, organizational knowledge, resources and staff are shared between the venues they own, across media formats. Smaller corporate-owned outlets borrow content from co-owned larger news sources. Journalistic and editorial practices are also shared. Similar processes affect the issue priorities of independent news providers – particularly the web services said to be most disruptive to the shared media agenda. While the Internet provides a space for numerous, varied news outlets, many of them still rely on partnerships with traditional media or news agencies to get access to content and enhance their consumer base (Duplessis & Li, 2004; Paterson, 2006). The financial problems that the media industry – and newspaper in particular – encountered in the last few years also contribute to the homogenization of news content. The collapsing business model of print has resulted in steep revenue declines for metropolitan newspapers (McChesney & Nichols, 2010). Under the circumstances, many organizations were forced to drastically cut down their news operations or go out of business (Curran, 2010). After a series of bankruptcies, closures, buyouts and layoffs, newspaper newsrooms are 30% smaller than they were in 2000 (Pew Project For Excellence in Journalism, 2011). Print publications, which have long been the main source of independent reporting in the U.S., are forced to reduce the scope and depth of the stories they cover (Downie & Schudson, 2009). Financial pressures also provide a strong incentive to consider multiple partnerships and content sharing. 63 The lack of funding for original reporting is preventing issues that would be considered high-priority in the past from entering the media agenda. This trend is especially noticeable in international reporting, as fewer and fewer U.S. news venues can afford to maintain foreign bureaus or send correspondents abroad. Original international coverage is particularly expensive to produce, which is why changes in media spending and resource allocations have a big impact on it. In the past couple of decades, U.S. foreign affairs news has decreased steadily in both scope and diversity (J. Carroll, 2007; Golan, 2006; Golan & Wanta, 2003). Wire services have become the most influential source of international stories (N. Kwak, Poor, & Skoric, 2006). As a result, nearly all media sources carry identical stories licensed from Reuters or AP (Zuckerman, 2008). In a study of news agency domination in foreign affairs reporting, Paterson (2006) claims that there are only four real sources of extensive international coverage: AP, Reuters, AFP and BBC. The crisis in journalism also affected local and ethnic news, which suffered from insufficient resources even before the economic downturn (Waldman, 2011). Community and ethnic sources are increasingly prone to reprinting or repurposing mainstream media stories in lieu of original reporting. As in the case with foreign affairs news, this is a long-term problem rather than a recent or transient trend. It was recorded, for instance, in a study of ethnic publications serving Asian and Latino neighborhoods in Los Angeles (Lin & Song, 2006). The analysis found that local coverage was much less prominent than international news and stories from the wire services. Research conducted years later in another diverse area reported similar findings from resident focus groups (N.-T. N. Chen, Dong, Ball-Rokeach, Parks, & Huang, 2012). 64 One striking example of the discrepancy between source abundance and information scarcity in local news comes from a Pew study conducted in Baltimore (Pew Project For Excellence in Journalism, 2010a). In a comprehensive analysis of the city’s news ecosystem, researchers identified 53 news outlets that regularly produced local content, ranging from radio, TV, and print, to online news sites and blogs. Yet content analysis found remarkably low levels of original reporting. Eight out of every ten stories repeated or repackaged previously published content. Of the original news items, 95% came from traditional news media, and particularly newspapers. Another study investigating the coverage of local government in 98 cities and 77 suburban communities similarly found that while many outlets were available, the majority of stories came from newspapers (Baldwin, Bergan, Fico, Lacy, & Wildman, 2010). Print sources published 53% of the government-related stories in cities and 75% in suburbs. Broadcast TV produced 36% of that content in the urban and 18% in suburban areas. Citizen news and blogs accounted for 3% in cities and less than 1% in suburban communities. Policy Challenges: Media Diversity Studying agenda-setting is becoming an increasingly important task with regulatory and policy implications. The need to understand the process is even more pronounced in light of recent social and technological shifts. One major regulatory concern in this area is the evaluation and promotion of media diversity. Explicating the reasoning behind diversity as a policy objective, U.S. scholars often invoke the marketplace of ideas metaphor (Entman & Wildman, 1992). European policy research 65 puts the issue in the context of the Habermasian public sphere ideal (Valcke, 2011). Both freedom of choice and upholding democratic values are major justifications behind the desire for maintaining a pluralism of voices in the media. One of the better-known analytical frameworks organizing knowledge around media diversity was proposed by Napoli (1999). His classification splits the diversity principle into three separate measurable components: source, content, and exposure diversity. The source dimension refers to the diversity of media organizations and program producers. Two key aspects assessed by policy-makers are (1) diversity of ownership, both in terms of media outlets and their programming, and (2) diversity in the workforce of individual organizations. The factors considered in evaluations of ownership diversity include market concentration and the demographic characteristics of owners. Workforce diversity measures have focused on the percentage of women and minority employees – a proportion expected to reflect the local labor market composition or population demographics (Butler, 2011; Karpel & Fleming, 2012; Papper, 2011). The content component of diversity examines the range of programing available to audience members. That includes diversity in (1) the type of programs, (2) the demographic targeting and characteristics (for instance ethnic and racial diversity of people featured in the programming), and (3) ideas or viewpoints. Viewpoint diversity, a key value in the marketplace of ideas paradigm, is linked to the representation of various ideological, cultural, and social perspectives in the media. As a highly subjective and 66 difficult to measure construct, it is often conceptualized as an expected outcome of other, easier to assess and regulate, diversity dimensions (Napoli, 2001). The idea of exposure diversity is based on McQuail’s (1992) diversity of content as sent and as received. This dimension shifts the focus from the content that is available to people to the content that they actually consume. To a large extent, the dynamics underlying this component overlap with the factors and processes discussed in the audience fragmentation section of this chapter (Napoli, 2011a). Exposure diversity metrics evaluate the number and type of outlets and programs selected by the public, as well as the range of viewpoints presented by those outlets. While it is certainly difficult to regulate directly, this is the dimension that has the most immediate impact on society, democracy and self-governance (Napoli, 2011b). Nonetheless, it has remained largely ignored by policy-makers (Webster, 2007). Multiple organizations and institutions have attempted to evaluate various aspects of media diversity. In a comprehensive report requested by the European Commission (2009), scholars from three universities and researchers from Ernst & Young proposed a set of indicators meant to assess media pluralism in EU member states. In addition to media distribution and use, the report looks at diversity of content in relation to (1) ownership and control, (2) media types and genres, (3) political viewpoints, (4) cultural expressions, and (5) local and regional interests. The document calls for an evaluation of the extent to which media provide fair and diverse representation of political and ideological viewpoints; local and regional interests; national, ethnic and linguistic groups. The diversity of available media formats (newspapers, TV and radio stations, websites, 67 etc.) and the impact of media ownership and control are also central to the suggested pluralism measures. In the United States, the Federal Communications Commission (FCC) is conducting preliminary studies expected to guide the design of a new analytical framework for the evaluation of media diversity. The framework should help the Commission evaluate the implications of cross-ownership and improve the regulation of media concentration in local markets. It is meant to replace the Diversity Index proposed under Chairman Powell, which was found unsuitable by the U.S. Third Circuit Court of Appeals (Lloyd & Napoli, 2007). The importance of content and exposure diversity and the ability to assess them were key points in the public discussions of the new index (Federal Communications Commission, 2009). In a report published by the Aspen Institute, the Knight Commission (2009) approaches the issue of media agenda diversity and homogeneity from a different perspective. The report stresses the necessity of evaluating and serving the information needs of local communities. Policy-makers in particular are tasked with ensuring that a flow of diverse and credible information reaches the residents. A recent analysis commissioned by the FCC demonstrates that the regulator is expanding its scope of inquiry to include concerns reflecting those of the Knight Foundation. The study (Friedland, Napoli, Ognyanova, Weil, & Wilson, 2012) links media issue priorities with the information needs of communities in the context of multi- level communication ecologies. It examines existing literature, identifying links between the diversity of views available to local communities and the diversity of media in local 68 markets. The work emphasizes a need to conduct further research on the forces shaping content flows in a changing information environment. One of the most daunting tasks the Federal Communication Commission faces is the substantiating of the assumed causal links between different diversity dimensions (Napoli, 1999). Media source diversity presumably leads to a broader range of content encompassing a wider variety of viewpoints. This content diversity, in turn, is expected to result in higher levels of exposure diversity as individuals take advantage of the multiple options available to them. Communication policy-making in the U.S. has largely been guided by the implicit assumption of underlying relationships between diversity components. Yet researchers, as well as a number of court rulings against the FCC, have underscored the fact that there is no sufficient empirical evidence to support the presumed causal links (Friedland et al., 2012). While the issue diversity of media coverage and public opinion are within the purview of agenda-setting research, works in the area have rarely explicitly addressed the question of policy relevance. In a policy-oriented theoretical work, Napoli and Gillis (2008) outline the potential contributions of social science to the FCC’s efforts to evaluate viewpoint diversity. The authors highlight the relevance of media effects scholarship, and agenda-setting studies in particular. Public agenda-setting studies could potentially supply information about the impact of different media outlets, providing valuable baseline data about their individual contributions to diversity. Intermedia agenda-setting work is even more promising, as it can capture sources of diversity and factors that promote homogeneity across local media markets. 69 The work of Napoli and Gillis (2008) is partly intended as a guide for policy- makers to important research within the social sciences. It is also meant, however, to encourage scholars in relevant fields to consider the significant policy ramifications of their work – something few have done in the past. The authors suggest that scholarship in established areas like agenda-setting should incorporate studies that more directly address ongoing policy concerns. Accordingly, this project proposes a conceptual framework sensitive to policy concerns, which allows for the testing of important relationships such as the one between ownership and content diversity. Outline and details are presented in Chapter 4. 70 CHAPTER 4: A NETWORK MODEL OF AGENDA-SETTING. ADDRESSING EXISTING CHALLENGES Pressures to rethink the simple model of influence proposed by McCombs and Shaw in the 1970s emerge from parallel developments in theory and society (Williams & Delli Carpini, 2011). Transformations in the media system, economic and technological shifts, have resulted in new production practices and consumption patterns (Castells, 2009). At the same time, media effects research – and agenda-setting scholarship in particular – is becoming increasingly nuanced in an attempt to capture multifaceted social dynamics. In addition to basic effects, new studies incorporate contingent conditions at the individual level, as well as intermedia interactions and external influences at the media level (Valenzuela & McCombs, 2009). Elaborating the initial agenda-setting framework, researchers began incorporating elements from other perspectives. Second-level studies focused on the attributes of objects in the news, spurring examinations of the theoretical links between agenda- setting, priming, and framing effects (Roessler, 2008; Scheufele, 2000; Shah, McLeod, Gotlieb, & Lee, 2009). The two-step flow of communication (Lazarsfeld et al., 1944) prompted investigations on the impact of interpersonal discussion on the agenda-setting process (Brosius & Weimann, 1996; Wanta & Wu, 1992; Yang & Stone, 2003). Media system dependency considerations (Ball-Rokeach, 1985; Ball-Rokeach & DeFleur, 1976) suggested examining the transfer of issue salience in the context of larger social systems and power relations (Matsaganis & Payne, 2005). News values and gatekeeping literature 71 pointed to a range of influences shaping media agendas (Shoemaker & Vos, 1996, 2009). The place of digital tools and platforms in the formation of media and public priorities came to the forefront of the field (Althaus & Tewksbury, 2002; Bennett & Iyengar, 2008). In order to address these added layers of complexity, a new conceptualization of the agenda-setting process needs to incorporate a variety of relevant features and relationships characterizing news outlets, audience members, and social issues. A framework of this kind would benefit from the instruments provided by network theory: a field that specializes in the examination of complex dynamics involving attributes and relations, as well as higher-order structures. This approach accommodates investigations of similarities, social ties, interactions, flows of information and resources among individuals, groups, and organizations (Borgatti, Mehra, Brass, & Labianca, 2009). Relational thinking allows for research examining key structural determinants of social, economic and political processes, patterns of power and influence (Castells, 2009), and the impact of interpersonal ties on individual preferences and public opinion (Wasserman & Faust, 1994). This theoretical orientation reflects the current state of the media system as it moves to networked forms of content production, delivery, and consumption. Persistent industry-wide trends increase the levels of consolidation, interorganizational collaborations, local and global partnerships (Arsenault & Castells, 2008b). Online formats and new technologies connect newsrooms and audience members (Cardoso, 2006), making content diffusion both faster and easier to track through digital traces 72 (Anderson, 2010). Professional and personal social ties affect individual news consumption and distribution habits (Boczkowski, 2010). In a recent theoretical work investigating those trends, Ognyanova and Monge (in press) propose a framework integrating media studies and network science. Their model of the media system is grounded in a multitheoretical, multilevel (MTML) approach outlined by Monge and Contractor (2003). MTML-based research incorporates a range of properties – from individual and dyadic, through more complex structural patterns, to network-level measures. The study of complex multilevel dynamics draws on multiple theoretical standpoints, some independent, others competing or complementary. In their application of MTML thinking to media studies, Ognyanova and Monge put forward an analytical strategy investigating the communication flows and relational ties within and across three domains: the media industry, audiences, and content. Detailing the need for a network perspective on media studies, they underscore the capacity of a relational approach to fill the gaps in existing literature as the field of mass communication undergoes a paradigm shift. Adopting the core principles outlined by Ognyanova and Monge (in press), the present work proposes one specific implementation of the broader multilevel model. The focus here is narrower, as theoretical explanations of network dynamics are drawn primarily from works within the agenda-setting perspective. The integrated model detailed below encompasses elements from the production, audience, and content domains – as well as a variety of links between them. 73 Figure 13. Agenda-setting: A network model example. A network approach to agenda-setting In order to capture major agenda-setting mechanisms, an integrated framework should reflect the dynamic flow of issues through public and media agendas. To this end, the model presented here (Figure 13) is structured as a dynamic multidimensional network of issues, individuals, and information sources. The following paragraphs provide a brief description of the relationship types incorporated in the framework, as well as the rationales for their inclusion. This section also sketches relevant actor and object characteristics. Finally, network mechanisms are mapped onto agenda-setting processes. 74 Relationship typology Issue adoption (issue – information source/ audience member) Key to the agenda-setting process, this type of tie indicates that an issue has become salient for media outlets or consumers. The connection between an issue and a news source is formed when the item is covered by the outlet. The link may be conceptualized in binary terms (present/absent connection), or weighted based on traditional salience dimensions – the placement prominence of a story or the time/space dedicated to it (Kiousis, 2004). Similarly, an issue adoption link to an audience member is recorded when it becomes clear that an issue has captured the person’s attention. This can be assessed through a typical agenda-setting survey instrument (McCombs, 2005). Alternatively, the link can be observed through digital traces (e.g. an individual posts a link to a story about the issue on a social networking platform – or mentions it in a blog post). Media use (audience member – information source) Researchers have examined a number of pertinent relationships between individuals and media outlets. Higher news consumption, as well as reliance on media, are known to enhance agenda-setting effects (McCombs & Reynolds, 2009; Wanta & Hu, 1994). In particular, exposure to a news source covering an issue is likely to increase the perceived importance of that issue (Stroud, 2011). This is, therefore, another key type of link in the model. As the literature has tested a number of related constructs (e.g. use, exposure, reliance, dependence), any of those can be substituted here. This allows for 75 conceptualizations ranging from a binary use/no use tie to a valued link weighted by exposure time or dependence strength. Interorganizational ties (information source – information source) A wide range of formal and informal relationships could constitute network ties between two media organizations. The list includes well-studied connections like partnership, ownership, and cross-investment (Arsenault & Castells, 2008a). Baker and Faulkner (2002) suggest a number of additional link types: market exchanges, strategic alliances, joint participation in syndicates, joint political action, interlocking directorates, family ties, even joint illegal activities such as collusion. Interorganizational relationships, both of cooperation and competition, are pertinent to the media agenda-setting process as they influence news selection (Dimmick, 2003). This may occur as a result of content sharing between outlets – or due to a transfer of organizational routines and news values. Social connections (audience member – audience member) Social ties include friendship, kinship, and other communication connections between audience members (e.g. friend/follow links in online social media platforms). These relationships are crucial as they provide a social infrastructure allowing for the spread of media preferences and the diffusion of news content. Interpersonal discussion is, furthermore, considered a major intervening variable in investigations of salience transfer between the media and public agendas (Dearing & Rogers, 1996a). When conversations deal with issues covered by news media, communication can enhance 76 agenda-setting effects (Wanta & Wu, 1992). This also means that direct exposure to specific news content may not always be a prerequisite for the effects to occur (Wanta & Ghanem, 2007). Combining agenda-setting research with two-step and diffusion models (Brosius & Weimann, 1996) has allowed researchers to study the interaction between interpersonal and media effects. The importance of examining social and media connections in parallel is recognized in a number of theoretical traditions. One example comes from the communication infrastructure theory (Kim & Ball-Rokeach, 2006), a framework incorporating interpersonal and mediated effects in a community context. Concept associations (issue – issue) Following McCombs’s (2004, 2010) definitions, issues on the agenda are broadly defined here to include any object which may draw attention, or about which one may hold an opinion. That allows research to explore general topics, public figures, organizations, or even countries in the news. Furthermore, under the network conceptualization proposed here, nodes denoted as “issues” may also be prominent object aspects or interpretations. In this way, the framework accommodates studies of second- level agenda-setting (McCombs, 2004, 2005; Weaver et al., 2004). Links between issues may connect items that have some association in meaning, a conceptual or semantic relationship. This is another broad definition allowing for multiple operationalizations, permitting the use of relationship ontologies such as the ones adopted in semantic web projects. 77 For instance, attitude objects such as “presidential elections”, “Barack Obama” and “Mitt Romney” could be considered conceptually associated. Such a conceptual tie between two issues may make them more likely to appear on the agenda together. Additionally, research has suggested that some issues may have a competitive relationship reducing the likelihood that they will be prominent at the same time (Djerf- Pierre, 2012). Link direction and agency One thing to note here is that the proposed model does not contain inherent assumptions about agency. Those could, however, be built in based on the theoretical grounding and research design of a particular study. While all relationships in the system are presented as symmetric (see Figure 13), it is possible to adopt an interpretation assuming a certain direction of influence. A directed link between individuals and media sources, for instance, would be grounded in an understanding of audiences as either active participants or passive consumers. The issue adoption links can also have a direction reflecting top-down processes or the view that individuals have the agency. An interesting alternative could build upon meme literature stemming from the work of Dawkins (2006), which implies that issues are the agents that propagate across hosts. Individual and dyadic attributes In addition to capturing the relationships between actors and objects, a network representation of agenda-setting allows for the inclusion of relevant node-level attributes. Information sources, for instance, may be characterized by revenue, geographic area, or format (e.g. radio, TV, print, online). Audience members have a range of demographic 78 characteristics potentially influencing the agenda-setting process (Wanta, 1997b; Wanta & Ghanem, 2007). Issues can also be evaluated or classified in a number of ways – e.g. by domain (politics, science, entertainment, etc.) or scope (local, regional, national, international). Some important agenda-setting constructs are dyadic in nature and need to be operationalized not as individual properties, but as link-level attributes. One such example is obtrusiveness, or the extent to which a particular issue is part of someone’s personal everyday experience (Coleman et al., 2008). Items like “unemployment” or “crime” may be obtrusive for some individuals and not others, making obtrusiveness a characteristic of the relationship between person and issue. Network mechanisms As discussed above, the network framework proposed here adopts some basic definitions of the agenda-setting perspective. The conceptualizations of issue, object, attribute, as well as measures of salience discussed in the Chapter 1 do also apply here. Other concepts and processes, however, require a network interpretation. The prominence of an issue on the agenda, for instance, is traditionally assessed based on a rank-ordered list of priorities (Valenzuela & McCombs, 2009). A simple network equivalent of that measure would be the issue's degree centrality: the number of individuals and/or media sources directly connected to an issue, potentially weighting for the strength of those relationships (Freeman, 1979). More advanced measures could take into account the extent to which an item is embedded in the overall network, or the 79 average number of steps to be traversed in order for the issue to reach every person/outlet included in the study (Borgatti & Everett, 2006). Figure 14. Network mechanisms underlying agenda-setting processes. The basic agenda-setting process is typically defined as a transfer of salience from media to the public agenda, with effect strength evaluated through correlation analysis (McCombs, 2004). There is evidence, furthermore, that audience members will perceive as salient an issue which features prominently in the news they consume (Stroud, 2011). In network terms, this process should result in a propensity for triadic closure within a particular source-individual-issue configuration (see Figure 14, panel 1). When an information source is connected to both an audience member and an issue, there should be an increased probability for tie formation between the issue and the individual. Though it has a different theoretical grounding, this mechanism operates somewhat 80 similarly to the balance principle known to predict transitivity in social relations (Granovetter, 1973). The capacity of individuals to place an issue on the media agenda could similarly be operationalized in network terms. Like its counterpart, bottom-up agenda setting can be expressed as a propensity towards the closure of triads in which an individual is linked to both an issue and a news source. However, while a single media outlet can influence a news consumer, the reverse effect is more likely to be a game of numbers. If a sufficiently large number of Americans have a shared concern, it may end up high on the news agenda, regardless of the media use patterns of those involved. Thus bottom-up agenda-setting effects may be produced by a preferential attachment mechanism (Easley & Kleinberg, 2010) similar to the one presented on Figure 14, panel 2. In a network context, this mechanism – also known as “cumulative advantage” or “the rich get richer” – describes a propensity to form links with nodes that are already well-connected. Preferential attachment to popular issues is more generally one plausible generative mechanism for an agenda network of the type described here. Both news sources and individuals are likely to form connections to issues already considered important by the media and the public. All of the processes described so far could potentially operate in conjunction to shape agenda-setting patterns. Combining those mechanisms in a single model provides a useful way to evaluate how well each one explains the observed structure. This is one advantage of taking a network approach, as it allows for the simultaneous testing of 81 multiple complementary and competing hypotheses operating at different levels of analysis (Contractor, Monge, & Leonardi, 2011; Monge & Contractor, 2003). Another network-centric analytical strategy aimed at predicting the adoption of issues comes from contagion and diffusion frameworks. Initially developed to track the spread of disease or technological innovations, those models have been used to study the propagation of topics through social networking platforms (Oh, Susarla, & Tan, 2008) and blogs (Leskovec, Backstrom, & Kleinberg, 2009; Leskovec, McGlohon, Faloutsos, Glance, & Hurst, 2007, April). Two types of models – threshold (Valente, 1996) and cascade (Cointet & Roth, 2009) – can be used to explore the diffusion of issues across outlets and individuals. In threshold models, adoption is based on the proportion of connections that have already adopted the issue. In a cascade model, each time an actor is "infected" with a new issue, there is a certain probability that the infection will spread to neighboring nodes. Reducing complexity The network model proposed above incorporates media effects, as well as intermedia and interpersonal influences. It provides a useful organizing framework encompassing different aspects and levels of agenda-setting. This comes at a cost, as data collection and analysis need to account for complex structures with multiple types of nodes and relationships. Not every research question requires that level of complexity, however. As Chapter 2 demonstrates, many studies in the field focus on a single dimension of the agenda-setting process and do not involve the full range of elements included here. 82 One simple way to reduce complexity while preserving the basic ideas behind the model is to focus on a limited subset of its elements. Studies could – and many do – only investigate issue adoption and media consumption links, discarding interorganizational, social, and conceptual associations (see Figure 15, panel 1). Traditional research such as the Chapel Hill study (McCombs & Shaw, 1972) includes even fewer factors, as it also ignores media use patterns. Figure 15. Reducing the complexity of networked agenda-setting models. Another way to simplify the analysis is to decrease the variety of node types present in the model. Reducing the number of modes (i.e. distinct sets of entities in a network) is a standard technique used with multimodal structures (Wasserman & Faust, 1994). The excluded elements are typically ones of less relevance or interest to the researcher (Borgatti, 2009). As discussed in Chapter 1, works within the agenda-setting 83 perspective are largely concerned with the impact of news on public opinion. The nature of particular issues is often less important than the degree of correspondence between media and audience priorities. This being the case, issues are one element type that can be removed from the model (see Figure 15, panel 2). In their place, a new relationship – an agenda convergence tie – is defined to represent the similarity in agendas between the remaining nodes (audience members and/or media outlets). The added links can be weighted based on one of several measures of similarity, the simplest of which is the number of shared issues (Borgatti & Halgin, 2010). Reducing complexity further, a study can focus on limited subsets of both nodes and links (Figure 15, panel 3). In the spirit of early agenda-setting research (Dearing & Rogers, 1996a), researchers may opt to examine the media use and issue overlap (agenda convergence) relationships between individuals and information sources (3a). Intermedia scholarship may similarly adopt models including interorganizational and shared issue relations between news outlets (3b). The network of overlap (or spread) of issues across individuals is also shown on Figure 15 (3c), although such research may fall outside the scope of traditional agenda-setting scholarship. Comparisons of issue associations across different agendas present another possibility (3d). This particular model was in fact adopted in a study by Guo and McCombs (2011) examining media and public agendas during Texas gubernatorial and U.S. senatorial elections. The analysis compared two issue networks. One of the networks represented conceptual associations between political figures and their attributes extracted from media content. The other contained similar associations reported by local residents. As the two concept maps exhibit high levels of similarity, Guo and McCombs concluded that media may be able to influence 84 relations between objects and attributes perceived by audience members. The article refers to this process as “third-level agenda-setting”. In several other cases, researchers have analyzed networks encompassing some subset of elements included in the model presented in Figure 13. Arsenault and Castells (2008a), for instance, investigated links of ownership and cross-investment among media corporations. Webster and Ksiazek (2012) studied a network of media outlets connected by ties of audience overlap. Research design and methodological challenges Testing empirically the approach outlined above, this project employs a reduced network model (Figure 15, panel 3b) to examine intermedia agenda-setting. The study is, furthermore, constructed with view to addressing a range of conceptual, policy, and methodological challenges faced by media effects scholarship. The following paragraphs outline major considerations about research design, many of which have been previously recognized in existing literature. A number of methodological issues emerge as a result of theoretical shifts in the agenda-setting literature. Chapter 3 of this work builds an argument against agenda disintegration claims. Yet this remains an open debate in the academic community and research within the tradition needs to be sensitive to its implications. One direct consequence of allowing for the possibility of increasing media fragmentation concerns sample size. Historically, agenda-setting studies have often focused on a small number of outlets, relying on the assumption of high redundancy 85 across media (McCombs, 2004). At present, the priorities of one or two sources can no longer be considered a viable substitute for the media agenda – at least not without a robust justification. As Meraz (2011) points out, this is the case even with elite publications like the New York Times, previously assumed to represent all news media. Agenda-setting scholarship published in the last twelve years, however, remains prone to examine a small number of influential sources (see Chapter 2). The average number of news outlets per article recorded here was 3.4 – and would be even lower if not for a handful of Internet studies that included several dozens to over a hundred blogs in their analyses. While research should also ideally include a wide range of formats – print, television, radio, and online – the focus is often limited to newspapers (Roessler, 2008). Among the agenda-setting research articles identified in Chapter 2, only 15% studied more than one type of media. Another important research design decision has to do with the selection and coding of news topics. As noted in the Acapulco typology (McCombs et al., 2011; Wanta & Ghanem, 2007, see also Figure 1), agenda-setting scholars can opt to focus on a single issue, or else examine multiple topics. In the first case, research would investigate the media and public salience of a specific problem – e.g. a health or environmental concern – often tracking shifts in attention over time (Ader, 1995; Barnes et al., 2008). The second case is more problematic. In studies examining the “full” media agenda, scholars often pick a small number of broad categories – for instance “the economy and jobs”, 86 “education”, “health”, “crime”, and “foreign affairs”. Those are regularly used to study agenda-setting in the context of elections or political campaigns (Coleman et al., 2008). Even before concerns of issue fragmentation were prominent, some found this setup conceptually and methodologically flawed. Early on, Rodgers and Dearing (1988) identified two major problems with it. One was the emphasis on political issues: “This primary emphasis on political issues is understandable, in the sense that a great many media news events are political in nature. But much other news content is not directly political in nature, and these news events should also be included in agenda-setting research, in order to determine the generalizability of public agenda-setting across various types of media content.” (p. 91) Even within the realm of politics, furthermore, recent works uncover strong influence from entertainment and other content (Holbert, 2005; Williams & Delli Carpini, 2011). The second problem identified by Rodgers and Dearing (1988) stems from the fact that agendas are operationalized as including a few very broad categories rather than a larger number of more specific issues. This crude operationalization also inflates correlation coefficients in agenda comparisons. Further demands are placed on studies interested specifically in examining agenda fragmentation. Such attempts, for instance, need to employ longitudinal design in order to capture trend dynamics. Fragmentation is seen as an ongoing process and as such can only be fully understood when recorded over time. Research of this kind, additionally, needs to examine the individual priorities of media sources, simply because an aggregated multi-source agenda would not capture 87 patterns of fragmentation. This brings up another methodological challenge. The issue priorities of individual outlets are shaped by multiple exogenous and endogenous factors, including the influence of some news sources on others. Traditional agenda-setting research relies largely on correlation and regression analyses (see Chapter 2) – two methods that assume independence across cases. Network analysis is one way to address interdependencies in the data. Network science can provide another key methodological advantage: a well- defined measure of fragmentation (Y. Chen et al., 2007). Much of the mass communication research discussing fragmentation relies on conjecture rather than empirical evidence (Sayre et al., 2010). As a result, there are very few studies that have actually operationalized the construct. Throughout the literature examined for this project – including the twelve years of articles presented in Chapter 2 – there were three studies that measured audience fragmentation (Tewksbury, 2005; Webster, 2005; Webster & Ksiazek, 2012). Not a single study proposed a measure of media agenda fragmentation. These and other benefits of network analysis have prompted scholars to suggest it as an instrument for policy-oriented research (Friedland et al., 2012). Many of the considerations outlined above, furthermore, are relevant for studies aiming to help guide media regulation. One additional set of concerns specific to such studies should be mentioned here. Much of the existing intermedia agenda-setting literature tests the level of similarity in issue coverage across media outlets. Regulatory bodies, however, would benefit most from research that reveals not only the overlap in agendas, but also the 88 factors that drive it. Obviously, this type of analysis would be particularly relevant if it includes drivers that are subject to regulation. The patterns of content diversity and homogeneity of different media formats, as well as the impact of ownership and cross- ownership are some of the key matters of interest in that regard (Napoli, 1999). The following chapters of this dissertation describe a study that tests the network approach to intermedia agenda-setting proposed here and address the concerns outlined above. In summary, those key conceptual and methodological points include: ‐ A large enough sample of media sources to capture potential fragmentation trends. ‐ A sample including outlets from a range of formats – print, radio, TV, and online. ‐ Media agendas that incorporate a wide range of specific issues rather than a handful of broad topics. ‐ Media agendas that incorporate issues extending beyond the sphere of politics. ‐ Longitudinal data allowing the tracking of fragmentation over time. ‐ Clear conceptualization and measurement of fragmentation. ‐ Methodology that allows for interdependencies in the data. ‐ Methodology that allows for the identification of policy-relevant factors. 89 CHAPTER 5: FORMULATING HYPOTHESES The following sections outline a study of media agendas designed to implement the network approach proposed in Chapter 4. The project is based on secondary data collected by the Pew Center’s Project for Excellence in Journalism (2008). The dataset contains U.S. news outlets of five different types: newspapers, radio, cable and network TV stations, and online services. The data includes a year’s worth of news stories divided into several hundred different topic categories. The sampling strategy and the issue coding are discussed in more detail in Chapter 6. Avoiding some of the methodological limitations of previous works, the analysis aims at testing the fragmentation hypothesis and identifying drivers of intermedia agenda-setting effects. As described in Chapter 4, the media agenda here is examined through an agenda convergence network of news sources (Figure 15, Panel 3b). A tie between two outlets in this network indicates their propensity to select and prioritize content in a similar fashion. The strength of the connection is assessed based on the overlap in key issues covered by media outlets over a period of time. Considerations presented in Chapter 3 suggest that even in a digital age, the agendas of key news sources have an impact on public opinion. Assuming that mainstream media still plays a central role in news selection and dissemination, this project proposes a look at agenda overlap across a number of diverse U.S. news sources. The first part of the proposed study addresses the following research question: 90 RQ1: Does the media agenda fragmentation for the sample of outlets included in the analysis increase or decrease during the period under observation? Demonstrating the relevance of a relational perspective, this work further employs network analytical tools to investigate a number of potential predictors of uniform story selection across sources. Grounded in literature on agenda-setting, news values, and gatekeeping, the following section lays out a series of factors expected to influence the convergence in news outlet agendas. News Values and Selection Mechanisms: Drivers of Agenda Convergence The notion of sustained practices that shape media agendas is a key tenet of academic works discussing news values (O'NeiIi & Harcup, 2008). Studying determinants of newsworthiness became particularly popular in the 1970s, when sociologists – most notably Herbert Gans (1979) – started taking an active interest in the selection of journalistic stories. In the following decades, numerous factors (see Chapter 1) affecting the chances of an event to become news were examined by researchers (F. Campbell, 1997). For the purposes of this study, the exact set of inherent properties making some stories more likely to get journalistic attention is not vitally important. In fact, as Schudson (1989) argues, it may even be the case that news values are a largely opaque structure. According to that hypothesis, even practitioners whose news sense has been instilled through years of journalism education and/or work in a newsroom can only give broad and general explanations of how they know news when they see it. 91 Regardless of the transparency of the selection process, news is a social construct organized and processed through a set of rules and routines. What gets covered is, therefore, not just a function of the intrinsic properties of a particular event. The choice of journalistic stories is governed by ongoing, stable practices put in place to reduce uncertainty and fulfill a socially imposed requirement to construct meaningful narratives (Kaplan, 2006). Institutionalized procedures structure the work of journalists and facilitate their attempts to differentiate between important and unimportant events, reliable and unreliable sources, accurate and inaccurate facts. The level of agenda convergence between outlets should therefore also reflect this underlying structure rather than simply be driven by a particular set of newsworthy events. As the news values and organizational norms guiding topic selection are relatively stable (even as technological and social processes transform the media system), it can be expected that common issue coverage patterns will also persist over time. Examining this, the present work seeks to answer: RQ2: What are the levels of similarity in the agenda convergence network of media sources across different points in time? Taking an institutional view of media routines, news values can be seen as operating on several levels – organizational, sector-specific and global (O'NeiIi & Harcup, 2008). The global level of news values reflects the existence of standards shared across the media industry. It is based on a set of practically universal conventions instilled through professional training. Those include generally accepted virtues like 92 objectivity, balance, veracity and serving the public interest – as well as more specific rules guiding the news production process (Paletz, 1999). At the level of media sectors, some format-specific requirements emerge (Kepplinger & Ehmig, 2006). Television newscasts, for instance, may give priority to materials that have good visual footage. Radio stations on the other hand will probably not choose stories that rely heavily on visuals. The specifics of the news cycle in each sector are also likely to affect the news selection process (Stevenson, 2003). Outlets releasing daily news reports will differ from monthly publications or fast online news constantly updated throughout the day. Those differences suggest that: H1: Two news outlets will be more likely to have converging agendas if they come from the same industry sector (printed press, network TV, cable TV, radio, Internet). At an institutional level, selection of stories depends on established internal routines. Media outlets vary in size, location, projected identity, social context, target audiences, political orientation, production technologies, available resources and ties to other organizations. All of those characteristics – and more – can affect content production (Allern, 2002). A number of scholars have suggested that news selection practices can be studied in the light of pure market forces, focusing on the effects of audience preferences, advertising demand and competition (Allern, 2002; E. L. Cohen, 2002; Dimmick, 2003, 2005; McManus, 1994, 1995). McManus (1994) asserts that there are two separate sets of factors influencing the selection of news stories. The first one, which used to be dominant 93 in the 20 th century, refers to journalistic norms and standards. The second, increasingly prevalent today, revolves around market logic and incorporates the pressures exerted on news outlets by consumers, advertisers and investors. Exploring the principles which govern the U.S. media sector, Dimmick (2003) focuses on competition and coexistence. He points out that competition takes place whenever organizations share the same resources – audiences and advertisers, but also story topics. Competition occurs both within industries (printed press, TV, etc. – firms using similar production technologies) and among populations of organizations within communities composed of members of many industries. Taking an ecological perspective on the media sector, Dimmick looks at resource niches. According to him, content in particular can be seen as a niche dimension – especially since the topics selected by media outlets determine the type of audience they will attract. Larger, generalist companies have broad content niches as they attempt to capture the widest variety of news appealing to large audiences. Smaller, specialized news enterprises have narrow niches and focus on a limited number and type of stories. The agenda-setting literature further affirms (Rogers et al., 1993) that large, popular media like the New York Times have the ability to influence the topic selection of other outlets. Meraz (2009) even suggests using network analysis to demonstrate the central position that big traditional media outlets hold in setting the issue agenda. Since large media should produce more diverse news, and elite outlets tend to influence other sources, it can be expected that: 94 H2: Larger outlets (identified through audience size) will be more central in the agenda convergence network. H3: There will be preferential attachment to elite outlets (identified through network characteristics) in the agenda convergence network. Both Dimmick (2003) and McManus (1994) devote particular attention to large international media conglomerates – and the implications that corporate ownership has for selection of content. Taking advantage of economies of scope, corporations will share resources – including staff, content and organizational knowledge – between the outlets they own. Thus venues positioned in different media industries may be more similar in terms of content if they are part of the same parent company. Co-owned outlets within an industry, however (especially if they are serving the same geographic area), are likely to diverge in their content selection. In an attempt to capture different audience segments, corporations will often diversify their properties within a sector, something that Dimmick refers to as part of a niche breadth strategy. This is particularly relevant in the case of multiple broadcast programs distributed over one channel (e.g. two TV shows running back to back on the same station). Based on those considerations, this study proposes that: H4: Outlets owned by the same parent company will be more likely to have converging agendas. H5: Programs that appear on the same channel will be less likely to have converging agendas. 95 Another important mechanism influencing the news agenda has to do with ideology and political power. While studies of media bias tend to focus on the framing of issues, selecting which items to cover can also be influenced by ideological considerations (Entman, 2007). Put in the context of competition and economic interests, the abandoning of political neutrality can be a profitable response to market forces (Hamilton, 2004). Research has identified political bias as one mechanism for product differentiation (Anand, Di Tella, & Galetovic, 2007). As media outlets attempt to outperform their competitors, catering to the audience's ideological preferences may be a useful strategy (Fengler & Russ-Mohl, 2008). A major reason for that is suggested by the media effects literature on selective exposure. It maintains that people have a preference for information which reinforces their existing opinions (Lazarsfeld et al., 1944). Politically-minded citizens have been found to favor those outlets which better match their viewpoints (Iyengar & Hahn, 2009), even if they do not specifically avoid conflicting viewpoints (Holbert et al., 2012). Selecting ideologically-preferred news sources is more pronounced on the Web (Bennett & Iyengar, 2008), but is also a known practice with other media formats – radio, television, newspapers (Garrett, 2009). This suggests that: H6: Media outlets serving audiences with similar political orientations will be more likely to have converging agendas. The considerations outlined above are most often discussed in the context of matching political leanings between media and audience members. We could, however, also look broadly at audience characteristics and their effect on media preference. People 96 will typically seek media that serve their information goals (Ball-Rokeach, 1985; Ball- Rokeach & DeFleur, 1976). Similarly, news outlets will offer materials expected to be of interest to their audiences. It is therefore likely that controlling for other important factors, two news sources serving similar demographic segments will be more likely to converge in their issue agendas. H7: Media outlets that serve population segments with similar demographic characteristics will be more likely to have converging agendas. Adopting the central premise of the news values literature, this work sets out to examine the forces that structure patterns of agenda convergence among media outlets. The seven hypotheses presented here span multiple gates, or levels, at which content is shaped (Shoemaker & Vos, 2009). Those include the level of media routines, encompassing format requirements, news cycle, and journalistic practices. The organizational level involves the effects of media size and ownership. Considerations related to media audiences, political institutions, and interest groups fall within the scope of the social institutions level. Taking a network approach to examine the multi-layered drivers of agenda convergence, this work employs dynamic actor-based modeling. A detailed description of the methods and data used in the analysis follows below. 97 CHAPTER 6: RESEARCH DESIGN Dataset Description The research questions and hypotheses presented above were tested through secondary analysis of two major datasets. The first one, collected by the Pew Project for Excellence in Journalism (2008), contained a list of topics covered on a daily basis by 64 U.S. news outlets over the span of one year. The sampling strategy designed by the Project for Excellence in Journalism (PEJ) aimed to capture the diversity of the U.S. mainstream media landscape. The multi-stage selection process included identifying media sectors, followed by selection of media outlets in each sector, choice of specific programs, and finally a protocol for story selection 1 . For a full list of outlets in the sample, see Appendix A. Throughout 2008, in-house trained research staff continuously recorded a variety of characteristics of key news stories produced by the selected outlets. A total of 69,942 stories were coded, including 7,350 newspaper stories, 6,539 online stories, 19,796 stories from network television, 21,892 stories on cable news, and 14,365 stories from radio programs. 1 Detailed description of the sampling, data collection, and coding protocol for the dataset is available online at http://www.people-press.org/category/datasets 98 Each individual story was classified according to its topic. The list of topics (or main issues mentioned in the story) was dynamically updated throughout the year to include emerging issues that did not previously exist as categories. Eventually each story in the 2008 PEJ dataset was coded as having one of 392 different main topics. The topics covered a wide range of areas: from immigration and same-sex marriage, through elections, conflicts in the Middle East and global warming, to sport scandals and celebrity news. Stories that covered more than one topic were coded into the topic that was more prominently mentioned. As the primary aim of the PEJ data collection was to capture high-priority content, the coding included key stories rather than the full content produced by outlets in the sample. The selection strategy employed by the researchers ensured the inclusion of materials scoring high on prominence – a dimension of story salience employed by agenda-setting research (Kiousis, 2004). Items were considered high-profile or prominent enough to include in the study if they appeared in the first 30 minutes of TV and radio shows and newscasts, began on the front pages of newspapers, or were in the top five items on the home page of online sources. The second dataset employed in this study comes from the National Annenberg Election Survey (NAES) 2008 Phone Edition from the Annenberg Public Policy Center, University of Pennsylvania. The nationally representative data were collected through telephone interviews with 57,967 randomly selected U.S. adults. The survey was conducted between December 17, 2007 and November 3, 2008. Participants were selected through random digit dialing and interviewed in English and Spanish. The response rate for the study was 23%. The NAES 2008 phone survey included only 99 respondents who had landlines in their households. Previous studies have indicated, however, that including a cell-only sample along with a landline sample produces general population estimates nearly identical to that of the landline sample alone (Keeter, Kennedy, Clark, Tompson, & Mokrzycki, 2007). The survey covered a wide range of topics related to the presidential election and politics in general. The analysis conducted here uses aggregated NAES data on media consumption, demographic characteristics, and political preferences of the respondents. Data cleaning procedures The R environment (2012) and RStudio software (2012) were used for data cleaning, formatting, and analysis. As a first step, the PEJ 2008 dataset was examined. Out of the 69,942 news stories it contained, 14,412 were coded as “999”, indicating that coders were unable to identify a single dominant topic. Those cases were removed from the data. Furthermore, out of the 64 PEJ sources, 27 were not coded throughout the year. They had between 13 and 44 weeks of missing data. Estimating models with such a large proportion of missing data was not desirable, as it was likely to skew the analysis and produce unreliable results (Ripley, Snijders, & Preciado, 2012). The problematic sources were thus excluded from the study. The remaining 37 media outlets from the PEJ sample were matched with news sources available in the NAES 2008 dataset. Upon examination, 32 of the 37 outlets were found present in the NAES dataset section on media consumption. The remaining 5 100 outlets were excluded from the study since the measures necessary to test the research hypotheses were not available for them. The final set of 32 outlets encompassed sources from all media sectors. Those included: ‐ Newspapers: New York Times, Washington Post, Wall Street Journal, USA Today, and LA Times. ‐ Network TV programs running on ABC, NBC, CBS, and PBS: World News Tonight, Today, NBC Nightly News, The Early Show, CBS Evening News, and NewsHour. ‐ Cable TV programs running on CNN, FOX News, and MSNBC: CNN Daytime, Situation Room, Anderson Cooper 360, Lou Dobbs, MSNBC Daytime, Hardball, Countdown with Keith Olbermann, Fox News Daytime, Special Report with Brit Hume, O'Reilly Factor, Hannity and Colmes, and Fox Report with Shepard Smith. ‐ Radio programs: Morning Edition, Rush Limbaugh, Sean Hannity, and Michael Savage. ‐ Online news sources: CNN.COM, Yahoo news, MSNBC.com, Google News, and AOL News. Measurement A table containing all outlet attributes used in the analysis is available in Appendix B. 101 Ownership For the purpose of this measure, the parent company owning each of the 32 outlets in the data (or the channels that carried them in the case of broadcast programs) was identified. This information was obtained from the Mint Global Business Intelligence database and confirmed through an examination of the news source websites. All parent companies were assigned a numeric ID. The owner ID vector was used as a categorical variable in the analysis. Channel A numeric ID was assigned to each “channel” in the data. Each non-broadcast outlet in the data was considered a separate channel. Television and radio programs were assigned IDs based on the station that carried them. For instance, O'Reilly Factor and Hannity received the same channel ID number since both shows appeared on Fox News. The channel ID vector was used as a categorical variable in the analysis. Sector Each outlet in the data was assigned a numeric ID corresponding to its media sector: print, network TV, cable TV, radio, or online. The sector ID vector was used as a categorical variable in the analysis. Audience size Audience size for each of the 32 outlets in the data was computed based on the 2008 National Annenberg Election Survey (NAES). The measure was calculated as the 102 number of respondents who identified each outlet as their main source of information in the week prior to the interview. The audience size (M = 1,542, SD = 2,439) ranged from 32 (for Special Report with Brit Hume) to 11,422 (for CNN Daytime). Since the NAES had a large, nationally-representative sample, the distribution of respondents across media sources was likely to correspond to the actual size of the outlet audiences. Audience demographics The audience demographics for each outlet were computed based on the 2008 NAES dataset. The average age, income, and education level of respondents who reported using each media source were recorded. The racial and ethnic compositions of program audiences were examined and the proportions of Hispanic and African- American respondents were included in the analysis. Other ethnic groups (Asian, Mixed Race, etc.) had too few cases or not enough variance to be added to the model. Age was measured in years, with the average age per news outlet ranging from 43.7 years (CNN.com) years to 60.2 years (the Situation Room). Education had 9 levels, (1 = Grade 8 or lower to 9 = Some graduate or professional degree). The lowest average education of a news outlet audience was 5.1 (5 = Some college, no degree) for The Early Show on CBS. The highest average education of a news source audience was 7.7 (8 = Graduate or professional school after college, no degree) for the New York Times. The household income variable had 9 levels (1 = Less than $10,000 to 9 = More than $150,000). The average income level per outlet audience ranged from 5.3 (Lou Dobbs) to 7.7 (Wall Street Journal). 103 Hispanic and African-American audiences were measured as percentages of the total number of respondents who reported using each media source. Hispanic audience proportions ranged from 0% (Special Report with Brit Hume) to 17% (LA Times). African-American respondents accounted for audiences ranging from 0% (Special Report with Brit Hume) to 18% (Washington Post). Party identification / Political ideology The party identification measure was computed based on the political leaning of respondents in the 2008 NAES dataset who reported using each media source. Survey participants were asked to identify as “Democrat”, “Republican”, or “Neither”. Party identification for news outlets was calculated as follows: Outlet Party ID = % Democrats among audience – % Republicans among audience Low negative values for an outlet score on this measure would thus point to an audience that leans Republican, while high positive values would suggest a more Democratic audience. The variable (M = 3.6, SD = 38) ranged from -68.8 (Special Report w Brit Hume on Fox News) to +64.7 (Countdown with Keith Olbermann on MSNBC). The political ideology measure was computed based on the ideological leaning of respondents in the 2008 NAES dataset who reported using each media source. Ideology was recorded on a scale ranging from 1 (Very conservative) to 5 (Very liberal). This variable was recoded into three levels: conservative (combining “Very Conservative” and “Somewhat conservative”), moderate, and liberal (combining “Very liberal” and “Somewhat liberal”). The political ideology for news outlets was calculated as follows: 104 Outlet Political Ideology = % liberal audience members – % conservative audience members Low negative values on the measure would thus indicate that an outlet has a more conservative audience, while high positive values would suggest a more liberal audience. A score close to zero would suggest ideologically balanced audience – or one that includes a large number of moderates. The ideology variable (M = -10.3 , SD = 42.4 ) ranged from -87.5 (Special Report with Brit Hume on Fox News) to +65.2 (Countdown with Keith Olbermann on MSNBC). As can be expected, party identification and political ideology were highly correlated (r = .983, p < .001). Agenda convergence network The network variables used in this study were constructed in R using several relevant packages (Butts, 2012; Butts, Handcock, & Hunter, 2012; Csardi & Nepusz, 2012). The media sources monitored by PEJ came from sectors with varying news cycles. A period of one week was selected to give every outlet under study sufficient time and opportunity to cover an issue. The dataset was accordingly divided into 52 one-week subsets. A binary affiliation network of topics (issues) and news outlets was constructed for each week. A link in this network indicated that a news source had covered the issue (cf. the issue adoption relationship discussed in Chapter 4) at least once within the selected timeframe. 105 This resulted in 52 matrices with dimensions 32 x 392 (number of outlets x number of topics). Since topic specifics were not of interest here, the outlet-by-outlet one-mode projection of the outlet-by-issue affiliation network was compiled and used in the analysis. Each of the 52 issue adoption outlet-by-topic matrices was transformed into an issue convergence outlet-by-outlet valued matrix of shared topics. Network literature discusses a large number of similarity measures used in unimodal approaches to affiliation data (Borgatti, 2009; Borgatti & Halgin, 2010; Wasserman & Faust, 1994). Many similarity indices – (simple matching, correlations, Jaccard coefficient, Hamming distance, etc.) – are commutative: the similarity of X to Y is equal to that of Y to X. Here, a non-commutative index was selected due to the nature of the data. In cases where outlets X and Y had a sizeable difference in the number of stories produced in a week, the shared topics between the two could account for a large percent of X’s stories, but a small proportion of Y’s content. In this case, it would be preferable for the X →Y link to carry more weight than the Y →X link. Accordingly, each agenda convergence link from outlet X to outlet Y was assigned a weight w calculated as the number of shared topics over the total number of topics covered by X: X→Y |X ∩ Y| |X| or in this case As some of the methods used in the analysis required binary data, the networks were further dichotomized using the global average similarity value (M = .495) as a cut- off point. Thus two outlets that had higher than the average similarity in agendas would 106 be connected by an agenda convergence link, while there would be no link between sources having average or lower similarity. In the testing of H1-H7, the 52 networks were further reduced to 13 binary matrices, each of which contained four weeks of data. In order to be recorded in one of those 13 networks, a link had to occur in at least two of the four weeks of that period 2 . This was done for two reasons: (1) to reduce noise and retain more stable structures that were less affected by the events of one particular week and (2) to obtain a reasonable number of time points that could be used in actor-based modeling. Analysis In order to answer RQ1, a fragmentation score was calculated for each of the 52 dichotomized weekly networks. As suggested by Borgatti (2006), fragmentation was measured as the proportion of node pairs in the network that could not reach one another (i.e. no path between them exists in the observed graph). For any two outlets i and j in the agenda convergence network, we can define r such that: 2 In an earlier version of the analysis, links were recorded in the aggregated four-week networks when they appeared in at least 3 of the four weekly networks. The analyses used to test H1-H7 on these less dense graphs produced results very similar to the ones reported here (identical direction and significance of the parameter estimates, as well as similar estimate sizes), but the model goodness of fit was worse. 1 if i can reach j 0 otherwise 107 The calculation of rij here relied on functions defined in the R package sna v.2.2 (Butts, 2012) to identify finite and infinite geodesic distances (shortest paths) between outlets. Fragmentation was then computed as: 1 ∑∑ To get additional insight into the network structure and its changes over time, the density, in-degree and out-degree centralization scores were also calculated for each of the 52 binary weekly matrices. Levels of centralization were relevant to RQ1, as agenda fragmentation has been linked to the idea of a decentralized media system (Bennett & Iyengar, 2008). Density was important as denser networks tend to be less fragmented. To evaluate the levels of similarity in the agenda convergence network across time points (RQ2), the product-moment correlations between the 52 weekly graphs were computed. Since correlations can be calculated on valued matrices, the weighted networks were used in this portion of the analysis. A quadratic assignment procedure test (Krackhardt, 1988) was performed to assess the significance of the similarities. Each observed measure was compared to statistics obtained from 1000 matrices randomly generated through row/column permutation of the original data. A result was considered significant if the observed correlation was greater than that for at least 95% of the simulated networks. Hypotheses 1 to 7 were tested through actor-based modeling using the RSiena package (Snijders, 2012a). Stochastic actor-based models are one analytical strategy that 108 captures network dynamics, allowing for the estimation of parameters for factors that influence network change over time (Burk, Steglich, & Snijders, 2007). One note to make here involves the assumptions of the method and the extent to which they apply to the data at hand. In general, this type of model assumes a network evolving as a stochastic process driven by actors. This makes it most appropriate for frameworks in which links are established and dissolved deliberately by rational agents (Snijders et al., 2010). Ties in this structure are supposed to be enduring rather than brief events (e.g. phone calls or e-mails between individuals). At a first glance, a relationship based on content similarity may seem to diverge from that assumption. However, as discussed in previous chapters (and in RQ2), the links of agenda convergence as conceptualized here represent an enduring underlying structure of intermedia relations. While specific events may introduce noise into the network, it is largely grounded in stable institutional practices and organizational routines. Furthermore, as Snijders et al (2010) suggest, more transient individual ties of the type considered here may be aggregated over a period of time. The researchers point out that the resulting structures can serve as indicators of state and be appropriately modeled through a stochastic actor- based network strategy. The model employed here contained parameters for each of the proposed hypotheses. The outlet attributes described in the Measurement section above were added as constant covariates. Several additional structural signatures were included. Density and reciprocity were estimated as both are basic effects that almost always need to be present when modeling a directed network (Ripley et al., 2012). The network variables used in the study, furthermore, were based on similarity scores obtained from affiliation data. 109 This type of network may have a higher clustering coefficient (Uzzi & Spiro, 2005), even after dichotomization. In order to account for that, a transitive triplets effect was estimated. This effect captures the propensity towards forming structures between three outlets (a, b, and c) of the type a → b, a → c, b → c. To test H1, a same sector parameter was added to the model. A significant positive estimate would indicate that news sources belonging to the same sector (print, network TV, cable TV, radio, online) would be more likely to share an agenda convergence link. For H2, an audience size alter effect was included. Here, a significant positive estimate would suggest that outlets with larger audiences receive more incoming ties. An in-degree popularity effect was added to the model to test H3: preferential attachment to elite outlets. The tendency of actors with high in-degree centrality to receive more incoming links (the rich get richer effect) would be confirmed by a significant positive estimate of this parameter. Note that elite status and popularity here refer to network characteristics rather than direct measures of an outlet’s agenda-setting power. While outlets capable of setting the agenda of other media were likely to be central in the agenda convergence network, the reverse would not necessarily be true. Two monadic covariate effects, same owner and same channel, were included to test H4 and H5. Those parameters would reflect the impact of sharing an owner or a channel on agenda convergence. 110 For H6, a political similarity 3 effect was included in the model. H7 was tested through several parameters related to outlet audience demographics: age similarity, income similarity, education similarity, Hispanic audience similarity, and African-American audience similarity. Significant positive similarity estimates would point to a higher likelihood for agenda convergence between outlets with similar values on the parameter in question. The deviation between observed and simulated statistics was examined to evaluate model convergence. Obtaining t-ratios smaller than 0.1 in absolute value for all included parameters would indicate very good convergence of the algorithm. A parameter estimate was considered significant if it was at least 1.96 times greater than its standard error (equivalent to p < .05). Goodness of fit for the model with regard to auxiliary statistics (Lospinoso & Snijders, 2011) was further assessed using the RSienaTest package (Snijders, 2012b). The in-degree and out-degree distributions generated by the model were compared to those of the observed data. A high p value would suggest a good fit of the model to the data. The initial assumption of longitudinal network analysis with RSiena is that parameters remain constant over time. If the importance of modeled effects changes, 3 Given the very high correlation between outlet party identification and political ideology (r = .98), only the first of those attributes (party ID) was used for the “political similarity” parameter included in the model. An alternative model estimated with the ideology parameter instead produced virtually identical results, with a slightly lower (but still excellent) convergence t-ratio for political similarity. 111 homogeneous specifications would result in estimates that are averaged across the time heterogeneity (Lospinoso, Schweinberger, Snijders, & Ripley, 2011). When there are theoretical reasons to expect shifts in a parameter across time periods, it is recommended that time heterogeneity is tested and accounted for in the model. This works examines media network dynamics throughout 2008: a year of presidential elections in the U.S. In the months leading up to the elections, news outlets in the sample were likely to focus their attention exclusively on election coverage, which would increase the level of similarity among their agendas. This – and other exogenous events during the year concentrating media attention on a single issue – could affect the density of the agenda network used to examine H1-H7. To address this possibility, score type tests of time heterogeneity were conducted in RsienaTest (Snijders, 2012b). As suggested by Lospinoso (2011, under review), time varying covariates (time dummy variables) were used for time periods with significant heterogeneity in network density. The resulting model would produce parameter estimates accounting for time heterogeneity in density. 112 CHAPTER 7: AGENDA CONVERGENCE: STUDY RESULTS RQ1: Fragmentation over time RQ1 asked whether the level of media fragmentation had increased or decreased during the period under observation. To address this question, the fragmentation of the 52 dichotomized weekly matrices representing the agenda convergence network over time was computed. To gain additional insight into the structure of the network, the density and centralization for all weekly snapshots were also calculated. The full table containing these network-level measures is available in Appendix C. Figure 16. Agenda convergence network: Fragmentation over time (2008) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0 5 10 15 20 25 30 35 40 45 50 55 Fragmentation Week Agenda convergence network: Fragmentation over time 113 The average fragmentation of the agenda convergence network was F =.287, SD = .11. This indicates that – on average – in less than a third of all possible outlet dyads dij, i was unreachable from j. Examining this as a linear trend over time (see Figure 16) suggested a decline in fragmentation throughout the year (F =-.002w +.35, where w is the week number). Controlling for autocorrelation at lag 1 (AC = .339, SE = 135; autocorrelations and partial autocorrelations for larger lags were non-significant) did not change the trend (F =-.002w +.27FLag1 + .26). While a linear model may not be the best fit for the data, the analysis conducted here did not attempt to capture more complex patterns over time. The data were examined only to confirm that there was no consistent increase in fragmentation during the period under observation. Out of the 52 weekly snapshots, a single matrix had a fragmentation coefficient of 0 with an agenda convergence network that formed a single component. The time period it represented was November 3 rd to 9 th (see Figure 17 for a snapshot of the agenda convergence network for that week). The singularly uniform media agenda that week emerged as a result of the U.S. presidential elections held on November 4 th . The next lowest-fragmentation time period (F = .06) was between March 10 th and 16 th when media attention was focused on the presidential primaries and a political scandal involving New York Governor Eliot Spitzer. The agenda convergence network had a relatively high centralization score (mean in-degree = .52, SD = .06; mean out-degree =.53, SD = .12). In comparison, the average centralization of 52,000 graphs obtained by generating a thousand row/column permutations of each weekly network was .21 for both in- and out-degree. 114 Figure 17. Agenda-convergence: a network snapshot for Nov. 3rd to 9th of 2008. Note: Node colors are based on media sectors, node sizes are proportional to degree centrality (calculated based on the number and weight of incoming links) for each outlet. Dense connections and low fragmentation characterize the agenda-convergence network during the week following the U.S. Presidential Elections in 2008. Outlet agendas converge (overlap) as media universally focus on the election coverage. 115 In social network literature, in-degree centrality is interpreted as a measure of prestige, while out-degree centrality is an indicator of actor activity (Wasserman & Faust, 1994). High levels of centralization point to a relatively small number of highly prestigious and very active nodes in the network. As a link from outlet X to outlet Y in the agenda convergence network only indicates similarity, it cannot be directly interpreted as evidence that X is taking cues about important stories from Y. It is likely, however, that elite sources capable of setting the agenda of other media would have a consistently high centrality over time. Thus while individual node centrality may not precisely map onto agenda-setting power, the high network centralization is a suggestive indicator. Network density (M = .41, SD = .08) – or the ratio of present ties to all possible ties in the network – is another key metric considered here. It is related to fragmentation in that denser networks tend to be less fragmented. The slight increase of density over time (D = 0.001w + 0.38, where w is the week number) could be one contributing factor explaining the decrease in fragmentation. Based on the agenda convergence network construction strategy, an increase of density over time suggested higher levels of similarity among outlets as the year progresses. One explanation for this comes from the 2008 presidential elections. In the months right before and after the elections, media attention was focused on the presidential race, narrowing the scope of outlet agendas and leading to higher convergence levels. 116 Figure 18. Agenda convergence network: Density over time (2008) RQ2: Levels of similarity across time points RQ2 sought to assess the consistency of the structure captured by snapshots of the agenda convergence network taken weekly during 2008. To examine the stability of the network over time, graph-level similarity indices among time points were computed. Correlations were calculated for all dyads Ni, Nj (i<j) where Nx was the state of the agenda convergence network for week x. In all graphs, each outlet had a full agenda convergence with itself, so that all elements along the diagonals of weekly networks had a value of 1. The diagonals were excluded from this part of the analysis to avoid artificially inflating the computed correlations. The correlations across all graph couples ranged between .125 and .790, with a mean of r = +.540, SD = .12. A quadratic assignment procedure test conducted to assess 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0 5 10 15 20 25 30 35 40 45 50 55 Density Week Agenda convergence network: Density over time 117 the significance of correlation coefficients resulted in p < .01 in each case. In a comparison with 1000 permutated networks, the observed correlation in each case was higher than that obtained from at least 990 of the generated graphs. Figure 18 show the average correlations between weekly networks based on their distance in time. As can be expected, time points that were closer (e.g. 0 or 1 week apart) were more similar than more distant ones (e.g. 50 or 51 weeks apart). Figure 19. Agenda convergence network: Average correlations (2008). H1-H7: Actor-based stochastic model estimates Thirteen aggregated four-week binary snapshots of the agenda convergence network (constructed as described in Chapter 6) were used to test hypotheses 1 to 7. Density, degree distributions and other descriptive characteristics of the thirteen graphs 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0 5 10 15 20 25 30 35 40 45 50 55 Correlation Weeks Apart Agenda convergence network: Average correlations across time periods 118 are available in Appendix D and Appendix E. The data included no missing network or attribute values. A model of the agenda convergence network dynamics, including rate parameters, as well as the structural and attribute covariate effects discussed in Chapter 6 was estimated in RSiena (Snijders, 2012a). After several re-estimations refining the values of parameter estimates, the resulting model had excellent convergence (see Table 2). To evaluate the potential time heterogeneity of the network density parameter, score type tests were conducted in RSienaTest (Snijders, 2012b). Based on the test results, time varying covariates (time dummy variables taking a value of 1 for the selected period and 0 for all other ones) were included for nine of the twelve periods in the model (Lospinoso, 2011, under review). This final model had excellent convergence with low deviations between observed and simulated statistics (see Table 3). The convergence t ratios for all parameters were lower than 0.1 in absolute value. The model fit to the observed data was further examined with regard to in-degree and out-degree distributions. Goodness of fit tests (Snijders, 2012b) suggested a very good match with observed data, with high p values for both in-degree (p = .52) and out- degree (p = .75) distributions. These results are also plotted on Figure 19 and Figure 20, which show the distribution of the simulated statistics and their observed values. 119 Table 2. Agenda Convergence Model 1: parameter estimates. Parameter Hypothesis Predicted Direction Parameter Estimate Standard Error Convergence t‐ratio Rate parameters Rate period 1 ‐ ‐ 12.5 *** 1.6 ‐ Rate period 2 ‐ ‐ 13.7 *** 1.9 ‐ Rate period 3 ‐ ‐ 11.7 *** 1.6 ‐ Rate period 4 ‐ ‐ 18.7 *** 2.9 ‐ Rate period 5 ‐ ‐ 10.5 *** 1.2 ‐ Rate period 6 ‐ ‐ 12.5 *** 1.5 ‐ Rate period 7 ‐ ‐ 17.2 *** 2.7 ‐ Rate period 8 ‐ ‐ 18.9 *** 3.0 ‐ Rate period 9 ‐ ‐ 14.3 *** 2.0 ‐ Rate period 10 ‐ ‐ 10.9 *** 1.4 ‐ Rate period 11 ‐ ‐ 14.7 *** 1.8 ‐ Rate period 12 ‐ ‐ 12.5 *** 1.7 ‐ Structural effects Density ‐ ‐ ‐2.224 *** 0.076 0.00 Reciprocity ‐ ‐ ‐0.760 *** 0.043 0.02 Transitive Triplets ‐ ‐ 0.072 *** 0.004 0.01 In‐degree popularity H2 positive 0.084 *** 0.002 ‐0.01 Monadic covariate effects Same sector H1 positive 0.541 *** 0.049 ‐0.01 Same owner H4 positive ‐0.034 0.105 ‐0.01 Same channel H5 negative ‐0.229 0.126 ‐0.02 Audience size alter H3 positive 0.000 31.61 ‐0.06 Political similarity H6 positive 0.164 * 0.084 ‐0.05 Age similarity H7 positive ‐0.061 0.087 0.04 Education similarity H7 positive 0.130 0.108 ‐0.02 Income similarity H7 positive 0.437 *** 0.127 0.01 Hispanic audience similarity H7 positive 0.340 *** 0.095 ‐0.01 Black audience similarity H7 positive 0.273 ** 0.091 ‐0.06 * p < .05, ** p < .01, *** p < .001 120 Table 3. Agenda Convergence Model 2: accounting for time heterogeneity in density. Parameter Hypothesis Predicted Direction Parameter Estimate Standard Error Convergence t‐ratio Rate parameters Rate period 1 ‐ ‐ 12.4 *** 1.7 ‐ Rate period 2 ‐ ‐ 11.5 *** 1.4 ‐ Rate period 3 ‐ ‐ 12.1 *** 2.0 ‐ Rate period 4 ‐ ‐ 14.1 *** 2.3 ‐ Rate period 5 ‐ ‐ 11.4 *** 1.3 ‐ Rate period 6 ‐ ‐ 11.5 *** 1.3 ‐ Rate period 7 ‐ ‐ 17.3 *** 4.9 ‐ Rate period 8 ‐ ‐ 18.8 *** 3.4 ‐ Rate period 9 ‐ ‐ 13.6 *** 1.8 ‐ Rate period 10 ‐ ‐ 11.6 *** 1.5 ‐ Rate period 11 ‐ ‐ 16.4 *** 2.1 ‐ Rate period 12 ‐ ‐ 13.4 *** 2.0 ‐ Structural effects Density ‐ ‐ ‐3.089 *** 0.130 ‐0.01 Reciprocity ‐ ‐ ‐0.659 *** 0.047 ‐0.07 Transitive Triplets ‐ ‐ 0.119 *** 0.007 ‐0.09 In‐degree popularity H2 positive 0.090 *** 0.003 ‐0.06 Monadic covariate effects Same sector H1 positive 0.556 *** 0.05 ‐0.01 Same owner H4 positive ‐0.135 0.107 ‐0.05 Same channel H5 negative ‐0.187 0.136 ‐0.05 Audience size alter H3 positive 0.000 31.61 ‐0.03 Political similarity H6 positive 0.221 ** 0.088 0.00 Age similarity H7 positive 0.063 0.094 0.02 Education similarity H7 positive ‐0.116 0.119 0.00 Income similarity H7 positive 0.452 *** 0.136 0.03 Hispanic audience similarity H7 positive 0.215 ** 0.106 0.01 Black audience similarity H7 positive 0.275 ** 0.100 ‐0.01 Time dummies: Density TD Period 2 ‐ ‐ 0.253 ** 0.115 0.05 TD Period 3 ‐ ‐ ‐0.549 *** 0.094 ‐0.06 TD Period 4 ‐ ‐ ‐0.521 *** 0.097 ‐0.10 TD Period 5 ‐ ‐ 0.349 *** 0.082 0.03 TD Period 6 ‐ ‐ 0.423 *** 0.101 0.03 TD Period 7 ‐ ‐ ‐0.377 *** 0.090 0.03 TD Period 8 ‐ ‐ 0.130 0.076 ‐0.03 TD Period 11 ‐ ‐ 0.348 *** 0.074 ‐0.01 TD Period 12 ‐ ‐ ‐0.348 *** 0.083 0.01 * p < .05, ** p < .01, *** p < .001 121 Figure 20. Goodness of fit: In-degree distribution. Figure 21. Goodness of fit: Out-degree distribution 122 The final model (see Table 3) provided support for four of the seven hypotheses presented in Chapter 5. The first hypothesis proposed that two news outlets will be more likely to have converging agendas if they come from the same industry sector (e.g. print, online, etc.). Consistent with the prediction of H1, the same sector parameter (Estimate = .556, SE = .05, p < .001) was positive and significant. As suggested by H3, the in-degree popularity (Estimate = .09, SE = .003, p < .001) indicated a tendency towards preferential attachment to outlet nodes with high in-degree centrality. The political similarity (Estimate = .221, SE = .088, p < .001) effect was also positive and significant, pointing to agenda convergence among outlets serving audiences with similar political orientation (H6). The estimates provided partial support for H7, which suggested that agenda convergence would be more likely among media serving audiences with similar demographic characteristics. The parameters estimates for income similarity (Estimate = .452, SE = .136, p < .001), Hispanic audience similarity (Estimate = .215, SE =.106, p < .01), and African-American audience similarity (Estimate = .275, SE = .1, p < .01) were positive and significant. The estimates for the age similarity (Estimate = .063, SE = .094, p > .05) and education similarity (Estimate = -.116, SE = .119, p > .05) effects, however, were not significant at the .05 level. Hypotheses 2, 4, and 5 were not supported by the results of the analysis. The same owner (Estimate = -.135, SE = .107, p > .05) and same channel (Estimate = -.187, SE = .136, p > .05) did not have a significant impact. Audience size (Estimate = 0, SE = 31.6, p > .05), expected to influence the in-degree centrality of outlets (H2), also resulted in a non-significant estimate. 123 The negative density (Estimate = -3.09, SE = .13, p > .001) and reciprocity (Estimate = -.659, SE = .047, p > .001) parameters suggested that agenda convergence links were not likely to be formed or reciprocated, except when tie formation was driven by other properties (e.g. same sector or audience similarities). As expected, the transitivity effect (Estimate = .119, SE = .007, p < .001) was positive and significant, which could partly be attributed to the strategy used to construct the agenda convergence network. Post-Hoc Analysis One plausible reason for the non-significant results for outlet ownership and channel was the redundancy of the two categorical variables in the secondary data used in the analysis. Outlet owner and channel overlapped to the extent that productions running on a channel were assigned its owner ID. As the data did not include a large number of cross-owned media, the two variables were fairly redundant. To examine this further, two separate variations of the model presented above (Model 2) were estimated. One retained the ownership but excluded the channel parameter (Model 3), the other estimated the channel effect, but excluded ownership (Model 4). Both models had excellent convergence with t ratios smaller than .01 for all included parameter estimates. The models also had good fit to the observed data with regard to both in-degree (Model 3 p = .39, Model 4 p = .41) and out-degree (Model 3 p =.75, Model 4 p = .66) distributions. 124 In both Model 3 and Model 4, the parameter estimates for effects other than ownership and channel were identical in direction and significance, as well as similar in size to those reported for Model 2. Both the same owner parameter in Model 3 (Estimate = -.241, SE = .094, p < .05) and the same channel parameter in Model 4 (Estimate = -.306, SE = .013, p < .01) were significant and negative. These results point to lower levels of agenda convergence between co-owned media – as well as between programs running on the same channel. 125 CHAPTER 8: DISCUSSION AND CONCLUSION This work proposed and empirically tested a conceptual framework and a network analytical strategy exploring the agenda-setting process in a new information environment. An examination of communication publications from the last twelve years (see Chapter 2) revealed that media-level or agenda building studies constituted a relatively small and less cited portion of the literature in the area. The few works that did explore that aspect of the theoretical perspective were prone to focus on a small number of outlets (an average of 3.4), often coming from a single sector. Such studies often analyzed newspaper (59%) or television (34%) content, followed by online media and blogs (15%). At the same time, the composition of media agendas became the focus of academic debates, as traditional theoretical frameworks were challenged by social, technological, and economic shifts. Early mass communication works would frequently – and sometimes implicitly – view “the media agenda” as a single construct that could be roughly approximated by measuring the salience of issues in specific outlets (Dearing & Rogers, 1996a). This approach seemed justified as a number of studies reported significant redundancy in source agendas – both among traditional (McCombs & Shaw, 1972) and online (Boczkowski & De Santos, 2007; Lim, 2010) sources. The homogeneity of issue priorities was attributed to social and institutional forces shaping news coverage (Reese, 1991). 126 Additional complexity was introduced as the relatively uniform, centralized media system of the 1970s was transformed through ongoing deregulation and a surge of new formats, channels, and information sources. Classic premises of mass communication and agenda-setting in particular were questioned in light of the seemingly unlimited number of new outlets with potentially divergent issue priorities (Bennett & Iyengar, 2008; Chaffee & Metzger, 2001). Media and audience fragmentation, as well as bottom-up effects stemming from user-generated content were identified as important and potentially disruptive trends. Even as media fragmentation was seen as a major challenge compelling scholars to rethink traditional theoretical perspectives, virtually no published articles proposed to measure and track its levels over time (Sayre et al., 2010). This lack of empirical work could partly be attributed to the analytical tools typically used in agenda-setting studies. Research in the area largely relied on rank-order correlations between issue priorities (Coleman et al., 2008, also see Chapter 2). This approach is not easily adapted to explore dyadic and higher order relationships and mechanisms. This is why, for instance, scholars have recently began analyzing audience balkanization through network measures (Webster & Ksiazek, 2012). Grounded in previous research, this study presented a range of arguments supporting the claim that mainstream traditional and online sources still play an important role in the formation of public opinion (see Chapter 3). The audience patterns of selective exposure and the bottom-up effects of civic media are relatively well-studied in communication and media effects literature. The present work took the road less traveled, examining the fragmentation of media priorities and identifying factors that contributed 127 to agenda convergence. To this end, a dynamic network of diverse media outlets over a one-year period was constructed and analyzed. The network framework proposed and tested here is perhaps the main contribution this work makes to contemporary mass communication research. Combined with a data collection scheme allowing for longitudinal analysis of diverse outlets and topical categories, this approach can facilitate the investigation of complex structures and processes driving news priorities and public opinion. Evaluating the agenda-setting framework and its relevance in the 21 st century is one important line of inquiry that falls within the scope of this analytical strategy. The use of a network design allowed this study to measure media fragmentation and test the major premises of an academic argument that did not previously have a solid empirical grounding. According to the results presented in Chapter 7, far from falling apart, the mainstream media agenda was rather homogeneous. The secondary data used in the analysis classified stories into hundreds of specific topics, in contrast to traditional research which often employs a small number of broad categories. In spite of the wide variety of story topics included in the study, the prominent issues covered by sources in the sample overlapped considerably. On average, during a week in 2008 two outlets would share close to 50% of the major stories each of them covered. The agenda convergence patterns were, furthermore, relatively consistent over time. The correlations between weekly network snapshots were significant and fairly strong. This pointed, as predicted by theory, to a structure that was shaped by industry values and institutional routines, rather than a random product of current events (Shoemaker & Vos, 2009). 128 In keeping with considerations outlined in Chapter 3, the fragmentation of the agenda convergence network did not increase over time. The level of media fragmentation also naturally declined during periods when a single issue occupied media attention: the 2008 primaries, the presidential election, or a political scandal such as the one that caused the resignation of New York Governor Spitzer. The findings of this study do not address the claim of McCombs and colleagues (2011, p. 11) that the media agenda is as homogeneous today as it was back in the 1970s. The results do, however, provide some empirical evidence going against the notion of an ever-increasing media fragmentation threatening to erode public consensus on issue importance. As suggested by previous research (Aikat & Yu, 2005; Lim, 2010), there appears to be enough consistency in the priorities of mainstream outlets to justify the study of agenda-setting processes. In keeping with Boczkowski’s (2010) findings, the average agenda convergence among outlets (and as a result, the density of the agenda convergence network constructed for this study) actually increased over time during the period under observation. The first part of the analysis conducted here found support for key premises of the agenda-setting and news values literatures. The findings discussed above pointed to the existence of an underlying structure driven by journalistic norms and routines, leading to stable patterns of agenda convergence across media over time. Modeling those patterns, the study set out to identify important organizational and industry-level factors that could enhance or suppress the convergence of outlet agendas. 129 Even as the boundaries between media formats are increasingly blurred and content is frequently repurposed and reused across news sources, the industry sector was still an important predictor of content similarity. As suggested by H1, agenda convergence was more prominent among outlets belonging to the same sector (print, radio, network TV, cable TV, and online news). This is explained by the distinctive demands of each media type, the format-specific production practices and news cycles, as well as the heightened levels of competition within sectors. Furthermore, to this day university education often provides specialized training for print, broadcast, and online journalism, leading to variations in industry standards across sectors. Two different ways of addressing outlet status were employed in the model. One examined the outlet’s popularity among audience members, while the other involved its capacity to set the agenda of other news sources. As suggested by Dimmick (2003), media with larger audiences were expected to cover a wider range of issues and be more central in the agenda convergence network (H2). This expectation was not supported by the results of the study. The non-significant effect was likely due to the specifics of the sample used in the analysis, which only included large news sources with national audience. The variance in outlet size may thus have been insufficient to adequately test the hypothesis. The second way of examining outlet popularity was through the in-degree centrality in the agenda convergence network. Consistent with H3, the study found that central outlets enhanced their status, becoming even more central over time. One key consideration that should be emphasized here is that in-degree centrality is not a direct 130 proxy for elite status or agenda-setting capacity. The measure only expresses a tendency for the agendas of many other sources to converge with that of the highly central outlet. Consequently, elite sources are likely to be central in the agenda-convergence network – but not all central outlets are necessarily elite. Popularity in network terms would thus not map precisely onto elite status. The analysis found some support for the expected higher convergence between outlets serving similar audience segments (H7). As individuals purposefully select sources that best match their information goals (Ball-Rokeach, 1988), media have a strong economic incentive to supply content relevant to their target populations. As a result, higher agenda-convergence can be expected between sources with a similar consumer base – e.g. ones catering to the interests of youth, Latinos, or affluent individuals. The results pointed to higher levels of agenda convergence between outlets with similar proportions of Hispanic and African-American audience members. Serving groups with similar income levels also led to converging agendas. The average age and education of audiences, however, had no significant effect. This result could be was at least partially due to the stronger media sector effect included in the model, as people of different age and education levels tend to prefer different media formats. For instance, the average age of respondents who used online media was 46.6, while that of television viewers was 54.3. TV viewers also had, on average, a two-year college degree, while the average education level of newspaper readers was a four-year college degree. 131 Another key aspect examined in this study was the role of political ideology in the shaping of issue priorities across news media. Research in this area tends to focus on content specifics (story framing) rather than the selection of issues that are covered (agenda-setting). The news priorities of outlets, however, are also subject to political influences (Entman, 2007). Ideologically tinted news values can, furthermore, emerge as a response to market forces (Hamilton, 2004). The results presented here confirm the impact of political factors on story selection. Outlets had higher levels of agenda convergence when they catered to audiences with similar political leanings (H6). While not addressing media bias directly, this finding indicates that the issue priorities of sources preferred by Republicans differed from those of outlets favored by Democrats. To the extent that media agendas match consumer preferences, dominant audience ideology will be associated with distinct politically-motivated issue selection patterns in news coverage. The present study thus highlights the importance of unpacking political slant not only in terms of story framing, but also at the agenda-building level. A relationship with major policy implications tested here was the link between media ownership and agenda convergence. Economies of scope driving the sharing of content and staff, as well as overlaps in organizational routines, were expected to produce higher agenda convergence between co-owned sources (H4). Product differentiation, on the other hand, could lead to divergence in the agendas of new sources sharing a single platform, such as TV programs running on the same channel (H5). The analysis conducted here found negative effects of both ownership and channel (examined separately in a post-hoc test). Controlling for other relevant factors, news sources owned 132 by the same parent company were less likely to have converging agendas. One explanation for this comes from the nature of the data used in the study. In almost all cases, the co-owned news sources in the sample belonged to the same media sector. According to previous research, media corporations tend to diversify their products within an industry sector in order to capture a wider audience (Dimmick, 2003). Examining the case of daily newspapers, for instance, George (2007) found that both product differentiation and the variety of topics in the news increased with greater ownership concentration. While the results indicate that co-owned outlets within a media sector have diverging agendas, further research is needed to fully assess policy implications. Findings going in the opposite direction – that of reduced issue diversity – would clearly indicate a negative impact of media concentration. Divergent issue priorities, however, are not a conclusive point in favor of a relaxed ownership regulation. It is possible that while it increases the diversity of topical coverage, ownership concentration still reduces the diversity of viewpoints in the news. Evaluating this possibility requires a careful and nuanced framing study employing an extensive content analysis. Another key policy implication of the results presented here is grounded in findings related to audience demographics. The analysis confirmed that media serving the same ethnic and socio-economic groups tend to cover similar issues. Ensuring content diversity across media would thus require the existence of sources targeting various demographic groups in each market. One road to enhancing content diversity goes through regulatory support for a variety of ethnic media outlets – a policy goal that 133 should go beyond concerns about minority ownership and workforce composition (V. S. Katz, Matsaganis, & Ball-Rokeach, 2012). Limitations and further research: data collection and methodological considerations A number of limitations of this work stem from the use of secondary data. The sample used in the analysis contained large media providing extensive political and current affairs coverage to their national audiences. While this was a useful way of capturing important trends in mainstream media, it could make generalizations to local or specialized publications seem problematic. Even though the empirical tests conducted in this study were based on a national media sample, many of the considerations outlined here apply at the local level. A homogenization of local and ethnic news is only to be expected as the economic troubles of journalism force outlets to decrease original reporting and reuse content produced by other sources (Waldman, 2011). A study conducted by Pew in Baltimore (2010a), for instance, found that 80% of local news content was based on reprinting or repackaging of already published stories. The financial difficulties faced by local journalism have led to a decline in the diversity of news issue agendas and, consequently, to lower levels of media fragmentation. The replication of the research presented here at the local level is certainly a promising avenue for future research. Works of this kind could feasibly examine a census of the news media operating in a local market – or identified as important by the members of a community. Research including the full range of information sources available to (or preferred by) local residents could capture a wide variety of outlets, from 134 large and general to smaller and more specialized, including public, commercial, mainstream and ethnic sources. This would also provide more robust results regarding the effect of media size, which was non-significant in the present study. The rationale behind most of the hypothesized drivers of agenda convergence – outlet size, popularity, sector, and ownership, audience demographics and socio- economic status – remains valid in the context of local/regional and ethnic news. Large organizations like the LA Times or Univision are likely to cover a broader range of topics relevant to the fairly diverse populations they serve. Smaller community outlets produce less content, covering a narrower selection of locally relevant issues. Elite sources – especially prominent newspapers – are also known to set the agendas of local print and television journalism (Protess & McCombs, 1991). Ownership and partnership links may have a stronger positive effect on agenda convergence for smaller venues that lack the resources to do extensive original reporting (Lin, Song, & Ball-Rokeach, 2010). Demographic characteristics of target audiences – ethnicity being an obvious example – are also likely to influence the news selection patterns of those outlets. The role of ideology, however, may be somewhat less pronounced in local and news, as coverage there is less prone to revolve around national politics. A local-level agenda convergence study will thus likely have to incorporate preliminary qualitative research and media monitoring in the selected area. This could help identify key forces, institutions, and affiliations in local politics that shape major debates in the news. Another important point about the secondary dataset used in this work concerns the time period under observation. Analyzing data from 2008, the study confirmed that industry-level and organizational factors had an important impact on news selection 135 patterns. The dynamics examined here, however, were also influenced by exogenous forces. The range of prominent topics discussed by news media shrank noticeably around the 2008 primaries and the presidential election. While these events can be seen as limiting the generalizability of research findings from that period, external disturbances of that kind are not that unusual. Elections and other widely covered issues (political scandals, war and conflict, natural disasters, the economic crisis, the Occupy movement, etc.) happen regularly and are bound to be present in longitudinal data. Nonetheless, a further examination of the agenda convergence network during a non-election year could provide a useful baseline for comparison. Chapter 4 of this work outlined a dynamic multidimensional network approach allowing researchers to address the complexities of agenda-setting processes in a new information environment. The proposed framework incorporated a variety of relevant characteristics and relationships among news sources, audience members, and social issues. The empirical investigation conducted here applied a reduced version of that model, focusing on the attributes and connections of media outlets. Further methodological work is needed to both refine this specific implementation and test in practice other parts of the proposed framework. As one example, the analytical strategy outlined in Chapter 4 could be employed in a network study of individual issue priorities, evaluating the extent to which they are shaped by social ties and connections to media sources. A limitation of the network model adopted here is that it registers agenda similarities, but does not directly capture the process of imitation. Based on the secondary data used in this work, it would be difficult to claim that an outlet has “borrowed” any 136 specific issue from another source. The temporal sequence of issue adoption, however, can be tracked and used to identify typical patterns that unfold over time. An examination of this kind could contribute a lot to our understanding of agenda building, as imitation is a key mechanism driving news content selection (Boczkowski, 2010). One additional challenge that still needs to be resolved involves the extensive data collection and processing needed in order to take full advantage of the analytical strategy proposed here. Media monitoring and content analysis can be prohibitively resource- intensive when the goal is to capture the full range of prominent stories produced by news organizations from all industry sectors. A common tactic in such cases is to reduce the scope of the examined content to several weeks of news stories. This option is certainly practical, but it limits the study of agenda convergence to one or two points in time. Finding a more robust way to address this problem is an important task for future works employing this methodological framework. Research efforts should be dedicated to the development of semantic analysis procedures that can replace or supplement the work of human coders. Semantic analysis relies on the idea that knowledge can be presented as networks of concepts and their relationships in a given domain (Carley, 1993). Though it has disadvantages, automated content analysis may also produce more consistent results across texts compared to human coders (Vlieger & Leydesdorff, 2011). While community specifics are a consideration in devising a set of categories and relationships to be used in the analysis, a general framework can be constructed and adapted to 137 different contexts. Frequencies of concepts in content covering different topics, as well as the patterns of concept relations (potentially assessed through term co-occurrence) can be used to measure the agenda convergence among media outlets. This is likely the most important extension of the model presented here, as it would significantly reduce the burden of data collection and processing, facilitating a variety of applications. A semantic approach could indeed render practical the use of research designs implementing in full the multidimensional network-analytical strategy proposed in this work. 138 REFERENCES Ader, C. R. (1995). A Longitudinal Study of Agenda Setting for the Issue of Environmental Pollution. Journalism & Mass Communication Quarterly, 72(2), 300-311. Aikat, D., & Yu, J. (2005). News on the Web: Agenda Setting of Online News in Web Sites of Major Newspaper, Television and Online News Services. Paper presented at the Annual Meeting of the International Communication Association (ICA), New York, NY. Allern, S. (2002). Journalistic and Commercial News Values. News Organizations as Patrons of an Institution and Market Actors. Nordicom Review, 23(1/2), 137-152. Althaus, S. L., & Tewksbury, D. (2002). Agenda Setting and the "New" News: Patterns of Issue Importance Among Readers of the Paper and Online Versions of the New York Times. Communication Research, 29(2), 180–207. doi: 10.1177/0093650202029002004 Anand, B., Di Tella, R., & Galetovic, A. (2007). Information or opinion? Media bias as product differentiation. Journal of Economics & Management Strategy, 16(3), 635-682. Anderson, C. W. (2010). Journalistic Networks and the Diffusion of Local News: The Brief, Happy News Life of the “Francisville Four”. Political Communication, 27(3), 289-309. doi: 10.1080/10584609.2010.496710 139 Arsenault, A., & Castells, M. (2008a). Switching Power: Rupert Murdoch and the Global Business of Media Politics: A Sociological Analysis. International Sociology, 23(4), 488. doi: 10.1177/0268580908090725 Arsenault, A., & Castells, M. (2008b). The Structure and Dynamics of Global Multi- Media Business Networks. International Journal of Communication, 2, 707-748. Baker, W. E., & Faulkner, R. R. (2002). Interorganizational networks. In J. A. C. Baum (Ed.), The Blackwell companion to organizations. (pp. 520-540): Oxford: Blackwell Publishers Ltd. Baldwin, T., Bergan, D., Fico, F., Lacy, S., & Wildman, S. S. (2010). News Media Coverage of City Governments in 2009. Quello Center for Telecommunication Management and Law. Retrieved from tc.msu.edu/news-media-coverage-city- governments-2009 Ball-Rokeach, S. J. (1985). The Origins of Individual Media-System Dependency: A sociological framework. Communication Research, 12(4), 485. doi: 10.1177/009365085012004003 Ball-Rokeach, S. J. (1988). Media System Dependency Theory. In M. DeFleur & S. J. Ball-Rokeach (Eds.), Theories of Mass Communication (pp. 297-327). White Plains, NY: Longman Group. Ball-Rokeach, S. J. (1998). A theory of media power and a theory of media use: Different stories, questions, and ways of thinking. Mass Communication & Society, 1(1), 5- 40. doi: 10.1080/15205436.1998.9676398 140 Ball-Rokeach, S. J., & DeFleur, M. L. (1976). A Dependency Model of Mass-Media Effects. Communication Research, 3(1), 3. doi: 10.1177/009365027600300101 Ball-Rokeach, S. J., Rokeach, M., & Grube, J. W. (1984). The Great American Values Test: Influencing Behavior and Belief Through Television. New York, NY: The Free Press. Barnes, M. D., Hanson, C. L., Novilla, L. M. B., Meacham, A. T., McIntyre, E., & Erickson, B. C. (2008). Analysis of media agenda setting during and after Hurricane Katrina: Implications for emergency preparedness, disaster response, and disaster policy. American Journal of Public Health, 98(4), 604-610. doi: 10.2105/AJPH.2007.112235 Belt, T. L., Just, M. R., & Crigler, A. N. (2012). The 2008 Media Primary Handicapping the Candidates in Newspapers, on TV, Cable, and the Internet. The International Journal of Press/Politics, 17(3), 341-369. doi: 10.1177/1940161212444124 Bennett, W. L. (2004). Gatekeeping and Press-Government Relations: A Multigated Model of News Construction. In L. L. Kaid (Ed.), Handbook of Political Communication Research (pp. 283-314). Mahwah, NJ: Lawrence Erlbaum Associates. Bennett, W. L., & Iyengar, S. (2008). A New Era of Minimal Effects? The Changing Foundations of Political Communication. Journal of Communication, 58(4), 707– 731. doi: 10.1111/j.1460-2466.2008.00410.x 141 Bennett, W. L., & Manheim, J. B. (2006). The One-Step Flow of Communication. The ANNALS of the American Academy of Political and Social Science, 608(1), 213. doi: 10.1177/0002716206292266 Blumler, J. G., & Kavanagh, D. (1999). The Third Age of Political Communication: Influences and Features. Political Communication, 16(3), 209-230. doi: 10.1080/105846099198596 Boczkowski, P. J. (2010). News at work: Imitation in an age of information abundance. Chicago, IL: University Of Chicago Press. Boczkowski, P. J., & De Santos, M. (2007). When more media equals less news: Patterns of content homogenization in Argentina's leading print and online newspapers. Political Communication, 24(2), 167-180. doi: 10.1080/10584600701313025 Borgatti, S. P. (2006). Identifying sets of key players in a social network. Computational & Mathematical Organization Theory, 12(1), 21-34. doi: 10.1007/s10588-006- 7084-x Borgatti, S. P. (2009). 2-Mode Concepts in Social Network Analysis. In R. A. Meyers (Ed.), Encyclopedia of Complexity and System Science. Borgatti, S. P., & Everett, M. G. (2006). A Graph-theoretic perspective on centrality. Social Networks, 28(4), 466-484. doi: 10.1016/j.socnet.2005.11.005 Borgatti, S. P., & Halgin, D. S. (2010). Analyzing Affiliation Networks. Lexington, KY: LINKS Center for Social Network Analysis, University of Kentucky. Lexington. 142 Borgatti, S. P., Mehra, A., Brass, D. J., & Labianca, G. (2009). Network analysis in the social sciences. Science, 323(5916), 892. doi: 10.1126/science.1165821 Brosius, H. B., & Weimann, G. (1996). Who Sets the Agenda: Agenda-Setting as a Two- Step Flow. Communication Research, 23(5), 561-580. doi: 10.1177/009365096023005002 Bryant, J., & Miron, D. (2004). Theory and Research in Mass Communication. Journal of Communication, 54(4), 662-704. doi: 10.1111/j.1460-2466.2004.tb02650.x Bulkow, K., Urban, J., & Schweiger, W. (2012). The Duality of Agenda-Setting: The Role of Information Processing. International Journal of Public Opinion Research. doi: 10.1093/ijpor/eds003 Burk, W. J., Steglich, C. E. G., & Snijders, T. A. B. (2007). Beyond dyadic interdependence: Actor-oriented models for co-evolving social networks and individual behaviors. International journal of behavioral development, 31(4), 397-404. doi: 10.1177/0165025407077762 Butler, B. (2011). NABJ Diversity Census: An Examination of Television Newsroom Management. College Park, MD: National Association of Black Journalists. Butts, C. T. (2012). sna: Tools for Social Network Analysis (Version 2.2) [R package]. Irvine, CA. Retrieved from CRAN.R-project.org/package=sna 143 Butts, C. T., Handcock, M., & Hunter, D. R. (2012). network: Classes for Relational Data (Version 1.7-1) [R package]. Irvine, CA. Retrieved from CRAN.R- project.org/package=network Campbell, F. (1997). Journalistic construction of news: information gathering. New library world, 98, 60-64. Campbell, V., Gibson, R., Gunter, B., & Touri, M. (2009). News Blogs, Mainstream News and News Agendas. In S. Tunney & G. Monaghan (Eds.), Web Journalism: A New Form of Citizenship? Eastbourne, UK: Sussex Academic Press. Cardoso, G. (2006). The Media in the Network Society: Browsing, News, Filters and Citizenship. Liboa, Portugal: CIES-ISCTE. Carley, K. M. (1993). Coding choices for textual analysis: A comparison of content analysis and map analysis. Sociological methodology, 75-126. doi: 10.2307/271007 Carroll, C. E., & McCombs, M. (2003). Agenda-Setting Effects of Business News on the Public’s Images and Opinions about Major Corporations. Corporate Reputation Review, 6(1), 36-46. doi: 10.1057/palgrave.crr.1540188 Carroll, J. (2007). Foreign News Coverage: The US Media’s Undervalued Asset: The Joan Shorenstein Center on the Press, Politics and Public Policy Working Paper Series. 144 Castells, M. (2005). Informationalism, Networks, And The Network Society: A Theoretical Blueprint. In M. Castells (Ed.), The network society: A cross-cultural perspective (pp. 3-45). London, UK: Edward Elgar Publishing. Castells, M. (2009). Communication Power. Oxford, UK: Oxford University Press. Chaffee, S. H., & Metzger, M. J. (2001). The End of Mass Communication? Mass Communication & Society, 4(4), 365-379. doi: 10.1207/S15327825MCS0404_3 Chen, N.-T. N., Dong, F., Ball-Rokeach, S. J., Parks, M., & Huang, J. (2012). Building a new media platform for local storytelling and civic engagement in ethnically diverse neighborhoods. New media & society. doi: 10.1177/1461444811435640 Chen, Y., Paul, G., Cohen, R., Havlin, S., Borgatti, S. P., Liljeros, F., & Stanley, H. E. (2007). Percolation theory and fragmentation measures in social networks. Physica A: Statistical Mechanics and its Applications, 378(1), 11-19. doi: 10.1016/j.physa.2006.11.074 Cohen, B. (1963). The Press and Foreign Policy. Princeton, NJ: Princeton University Press. Cohen, E. L. (2002). Online Journalism as Market-Driven Journalism. Journal of Broadcasting & Electronic Media, 46(4), 532-549. doi: 10.1207/s15506878jobem4604_3 145 Cointet, J. P., & Roth, C. (2009). Socio-semantic dynamics in a blog network. Paper presented at the IEEE SocialCom 09 International Conference Social Computing, Vancouver, Canada. Coleman, R., & McCombs, M. (2007). The Young and Agendaless? Exploring Age- related Differences in Agenda Setting on the Youngest Generation, Baby Boomers, and the Civic Generation. Journalism & Mass Communication Quarterly, 84(3), 495. doi: 10.1177/107769900708400306 Coleman, R., McCombs, M., Shaw, D., & Weaver, D. (2008). Agenda Setting. In K. Wahl-Jorgensen & T. Hanitzsch (Eds.), Handbook of Journalism Studies (pp. 147). New York, NY: Routledge. Contractor, N., Monge, P., & Leonardi, P. (2011). Multidimensional networks and the dynamics of sociomateriality: Bringing technology inside the network. International Journal of Communication, 5, 682–720. Csardi, G., & Nepusz, T. (2012). The igraph software package for complex network research (Version 0.6) [R package]: InterJournal, Complex Systems 1695. Retrieved from CRAN.R-project.org/package=igraph Curran, J. (2010). Future Of Journalism. Journalism Studies, 11(2). doi: 10.1080/14616701003722444 Dawkins, R. (2006). The Selfish Gene (3rd ed.). New York, NY: Oxford University Press. 146 Dearing, J. W., & Rogers, E. M. (1996a). Agenda-Setting. Thousand Oaks, CA: Sage Publications. Dearing, J. W., & Rogers, E. M. (1996b). What is agenda-setting? Communication concepts: Agenda-setting (pp. 1-23). Thousand Oaks, CA: Sage Publications. DiMaggio, P., Hargittai, E., Neuman, W. R., & Robinson, J. P. (2001). Social Implications of the Internet. Annual Reviews in Sociology, 27(1), 307-336. doi: 10.1146/annurev.soc.27.1.307 Dimitrova, D. V., Connolly-Ahern, C., Williams, A. P., Kaid, L. L., & Reid, A. (2003). Hyperlinking as Gatekeeping: online newspaper coverage of the execution of an American terrorist. Journalism Studies, 4(3), 401-414. doi: 10.1080/1461670032000099741 Dimmick, J. W. (2003). Media Competition and Coexistence: the theory of the Niche. Mahwah, NJ: Lawrence Erlbaum Associates. Dimmick, J. W. (2005). Media competition and levels of analysis. Handbook of media management and economics, 345–362. Djerf-Pierre, M. (2012). The Crowding-Out Effect: Issue dynamics and attention to environmental issues in television news reporting over 30 years. Journalism Studies, 13(4). doi: 10.1080/1461670X.2011.650924 Downie, L., & Schudson, M. (2009). The Reconstruction of American Journalism. Columbia Journalism Review. 147 Duplessis, R., & Li, X. (2004). Cross-Media Ownership and Its Effect on Technological Convergence of Online News Content - A Content Analysis of 100 Internet Newspapers. Paper presented at the Annual Meeting of the International Communication Association (ICA), New Orleans, LA. Dutta-Bergman, M. J. (2004). Complementarity in consumption of news types across traditional and new media. Journal of Broadcasting & Electronic Media, 48, 41. doi: 10.1207/s15506878jobem4801_3 Easley, D., & Kleinberg, J. (2010). Networks, Crowds, and Markets: Reasoning About a Highly Connected World New York, NY: Cambridge University Press. Entman, R. M. (2007). Framing Bias: Media in the Distribution of Power. Journal of Communication, 57(1), 163-173. doi: 10.1111/j.1460-2466.2006.00336.x Entman, R. M., & Wildman, S. S. (1992). Reconciling economic and non-economic perspectives on media policy: Transcending the “marketplace of ideas”. Journal of Communication, 42(1), 5-19. doi: 10.1111/j.1460-2466.1992.tb00765.x Fahmy, S. S., Wanta, W., Johnson, T. J., & Zhang, J. (2011). The Path to War: Exploring a Second-Level Agenda-Building Analysis Examining the Relationship among the Media, the Public and the President. International Communication Gazette, 73(4), 322-342. doi: 10.1177/1748048511398598 Farrell, H., & Drezner, D. W. (2008). The power and politics of blogs. Public Choice, 134(1), 15-30. 148 Federal Communications Commission (Producer). (2009). Media Ownership Workshop - Policy Scholars’ Panel. Retrieved from http://www.fcc.gov/ownership/workshop- 110209.html Fengler, S., & Russ-Mohl, S. (2008). The Crumbling Hidden Wall: towards an Economic Theory of Journalism. Kyklos, 61(4), 520-542. doi: 10.1111/j.1467- 6435.2008.00416.x Freeman, L. C. (1979). Centrality in social networks conceptual clarification. Social Networks, 1(3), 215-239. Friedland, L. A., Napoli, P. M., Ognyanova, K., Weil, C., & Wilson, E. J. (2012). Review of the Literature Regarding Critical Information Needs of the American Public. Washington, DC: Federal Communications Commission. Galtung, J., & Ruge, M. H. (1965). The structure of foreign news: The Presentation of the Congo, Cuba and Cyprus Crises in Four Norwegian Newspapers. Journal of peace research, 2(1), 64-90. doi: 10.1177/002234336500200104 Gandy, O. H. (1982). Beyond agenda setting: Information subsidies and public policy. Norwood, NJ: Ablex Publishing Company. Gans, H. J. (1972). The famine in American mass-communications research: Comments on Hirsch, Tuchman, and Gecas. The American Journal of Sociology, 77(4), 697- 705. doi: 10.1086/225195 149 Gans, H. J. (1979). Deciding what's news: A study of CBS Evening News, NBC Nightly News, Newsweek and Time. New York: Pantheon Books. Garrett, R. K. (2009). Echo chambers online? Politically motivated selective exposure among Internet news users1. Journal of Computer Mediated Communication, 14(2), 265-285. doi: 10.1111/j.1083-6101.2009.01440.x Gentzkow, M., & Shapiro, J. M. (2010). Ideological segregation online and offline NBER Working Papers: National Bureau of Economic Research. George, L. (2007). What’s fit to print: The effect of ownership concentration on product variety in daily newspaper markets. Information Economics and Policy, 19(3), 285-303. Gerbner, G., & Gross, L. (1976). Living with television: The violence profile. Journal of Communication, 26(2), 173-199. doi: 10.1111/j.1460-2466.1976.tb01397.x Golan, G. (2006). Inter-Media Agenda Setting And Global News Coverage: Assessing the influence of the New York Times on three network television evening news programs. Journalism Studies, 7(2), 323. doi: 10.1080/14616700500533643 Golan, G. (2008). Where in the World Is Africa?: Predicting Coverage of Africa by US Television Networks. International Communication Gazette, 70(1), 41. Golan, G., & Wanta, W. (2003). International elections on US network news: An examination of factors affecting newsworthiness. International Communication Gazette, 65(1), 25. 150 Goldhaber, M. H. (1997). The attention economy and the net. First Monday, 2(4-7). Google Inc. (2012). Google Scholar. Retrieved January 12th, 2012, from scholar.google.com Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360-1380. Hallin, D. C. (1989). The Uncensored War: The Media and Vietnam. Los Angeles, CA: Univ of California Press. Hamilton, J. T. (2004). All the News That's Fit to Sell: How the Market Transforms Information Into News. Princeton, NJ: Princeton University Press. Harcup, T., & O'Neill, D. (2001). What is news? Galtung and Ruge revisited. Journalism Studies, 2(2), 261-280. doi: 10.1080/14616700118449 Hindman, M. (2011). Less of the Same: The Lack of Local News on the Internet. Washington, DC: Federal Communications Commission. Holbert, R. L. (2005). A Typology for the Study of Entertainment Television and Politics. American Behavioral Scientist, 49(3), 436-453. doi: 10.1177/0002764205279419 Holbert, R. L., Garrett, R. K., & Gleason, L. S. (2010). A New Era of Minimal Effects? A Response to Bennett and Iyengar. Journal of Communication, 60(1), 15 - 34. doi: 10.1111/j.1460-2466.2009.01470.x Holbert, R. L., Hmielowski, J. D., & Weeks, B. E. (2012). Clarifying Relationships Between Ideology and Ideologically Oriented Cable TV News Use: A Case of 151 Suppression. Communication Research, 39(2), 194-216. doi: 10.1177/0093650211405650 Iyengar, S., & Hahn, K. S. (2009). Red media, blue media: Evidence of ideological selectivity in media use. Journal of Communication, 59(1), 19-39. doi: 10.1111/j.1460-2466.2008.01402.x Iyengar, S., & Kinder, D. R. (2010). News that matters: Television and American opinion (2nd ed.). Chicago, IL: University of Chicago Press. Kaplan, R. L. (2006). The News About New Institutionalism: Journalism's Ethic of Objectivity and Its Political Origins. Political Communication, 23(2), 173-185. doi: 10.1080/10584600600629737 Karpel, R., & Fleming, K. (2012). ASNE Newsroom Employment Census. Reston, VA: American Society of Newspaper Editors. Katz, E., Blumler, J. G., & Gurevitch, M. (1973). Uses and gratifications research. Public Opinion Quarterly, 37(4), 509-523. doi: 10.1086/268109 Katz, E., & Lazarsfeld, P. F. (1955). Personal influence: The part played by people in the flow of mass communication. Glencoe, IL: The Free Press. Katz, V. S., Matsaganis, M. D., & Ball-Rokeach, S. J. (2012). Ethnic Media as Partners for Increasing Broadband Adoption and Social Inclusion. Journal of Information Policy, 2(0). 152 Kaye, J., & Quinn, S. (2010). Funding Journalism in the Digital Age: Business Models, Strategies, Issues and Trends. New York, NY: Peter Lang Publishing. Keeter, S., Kennedy, C., Clark, A., Tompson, T., & Mokrzycki, M. (2007). What's Missing from National Landline RDD Surveys? The Impact of the Growing Cell- Only Population. Public Opinion Quarterly, 71(5), 772-792. doi: 10.1093/poq/nfm053 Kepplinger, H. M., & Ehmig, S. C. (2006). Predicting news decisions. An empirical test of the two-component theory of news selection. Communications, 31(1), 25-43. doi: 10.1515/COMMUN.2006.003 Kim, Y. C., & Ball-Rokeach, S. J. (2006). Community storytelling network, neighborhood context, and civic engagement: A multilevel approach. Human Communication Research, 32(4), 411-439. doi: 10.1111/j.1468- 2958.2006.00282.x Kinght Commission. (2009). Informing Communities: Sustaining Democracy in the Digital Age: The Aspen Institute. Kiousis, S. (2004). Explicating media salience: A factor analysis of New York Times issue coverage during the 2000 US presidential election. Journal of Communication, 54(1), 71-87. doi: 10.1093/joc/54.1.71 Kiousis, S. (2011). Agenda-Setting and Attitudes: Exploring the impact of media salience on perceived salience and public attitude strength of US presidential candidates 153 from 1984 to 2004 Journalism Studies, 12(3), 359-374. doi: 10.1080/1461670X.2010.501149 Kiousis, S., & McCombs, M. (2004). Agenda-Setting Effects and Attitude Strength: Political Figures During the 1996 Presidential Election. Communication Research, 31(1), 36-57. doi: 10.1177/0093650203260205 Kiousis, S., & McDevitt, M. (2008). Agenda Setting in Civic Development: Effects of Curricula and Issue Importance on Youth Voter Turnout. Communication Research, 35(4), 481. doi: 10.1177/0093650208315978 Kiousis, S., McDevitt, M., & Wu, X. (2005). The genesis of civic awareness: Agenda setting in political socialization. Journal of Communication, 55(4), 756-774. doi: 10.1111/j.1460-2466.2005.tb03021.x Kiousis, S., Mitrook, M., Wu, X., & Seltzer, T. (2006). First-and second-level agenda- building and agenda-setting effects: Exploring the linkages among candidate news releases, media coverage, and public opinion during the 2002 Florida gubernatorial election. Journal of Public Relations Research, 18(3), 265-285. doi: 10.1207/s1532754xjprr1803_4 Kosicki, G. M. (1993). Problems and Opportunities in Agenda-Setting Research. Journal of Communication, 43(2), 100-127. doi: 10.1111/j.1460-2466.1993.tb01265.x Krackhardt, D. (1988). Predicting with networks: Nonparametric multiple regression analysis of dyadic data. Social Networks, 10(4), 359-381. 154 Kwak, H., Lee, C., Park, H., & Moon, S. (2010). What is Twitter, a Social Network or a News Media? Paper presented at the World Wide Web Conference Kwak, N., Poor, N., & Skoric, M. (2006). Honey, I shrunk the world! The relation between Internet use and international engagement. Mass Communication & Society, 9(2), 189-213. doi: 10.1207/s15327825mcs0902_4 Lang, G. E., & Lang, K. (1981). Watergate: An exploration of the agenda-building process. Mass Communication Review, 2(447-468). Lazarsfeld, P. F., Berelson, B., & Gaudet, H. (1944). The people's choice: How the voter makes up his mind in a presidential campaign. New York, NY: Columbia University Press. Lee, B., Lancendorfer, K. M., & Lee, K. J. (2005). Agenda-setting and the internet: The intermedia influence of Internet bulletin boards on newspaper coverage of the 2000 general election in South Korea. Asian Journal of Communication, 15(1), 57-71. doi: 10.1080/0129298042000329793 Lee, J. K. (2007). The Effect of the Internet on Homogeneity of the Media Agenda: A Test of the Fragmentation Thesis. Journalism & Mass Communication Quarterly, 84(4), 745. doi: 10.1177/107769900708400406 Leskovec, J., Backstrom, L., & Kleinberg, J. (2009). Meme-tracking and the dynamics of the news cycle. Paper presented at the The 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Paris, France. 155 Leskovec, J., McGlohon, M., Faloutsos, C., Glance, N., & Hurst, M. (2007, April). Cascading behavior in large blog graphs: Patterns and a model. Paper presented at the Society of Applied and Industrial Mathematics Conference on Data Mining, Minneapolis, MN. Lim, J. (2010). Convergence of Attention and Prominence Dimensions of Salience among Major Online Newspapers. Journal of Computer-Mediated Communication, 15(2), 293-313. doi: 10.1111/j.1083-6101.2010.01521.x Lin, W. Y., & Song, H. (2006). Geo-ethnic storytelling: An examination of ethnic media content in contemporary immigrant communities. Journalism, 7(3), 362. doi: 10.1177/1464884906065518 Lin, W. Y., Song, H., & Ball-Rokeach, S. (2010). Localizing the global: Exploring the transnational ties that bind in new immigrant communities. Journal of Communication, 60(2), 205-229. doi: 10.1111/j.1460-2466.2010.01480.x Lippmann, W. (1922). Public Opinion. New York, NY: The Free Press. Lloyd, M., & Napoli, P. N. (2007). Local media diversity matters: Measure media diversity according to democratic values, not market values. Washington, DC: Center for American Progress. Lospinoso, J. A. (2011, under review). Testing and modeling time heterogeneity in longitudinal studies of social networks: A tutorial in RSiena. Connections. 156 Lospinoso, J. A., Schweinberger, M., Snijders, T. A. B., & Ripley, R. M. (2011). Assessing and accounting for time heterogeneity in stochastic actor oriented models. Advances in Data Analysis and Classification, 5(2), 147-176. doi: 10.1007/s11634-010-0076-1 Lospinoso, J. A., & Snijders, T. A. B. (2011). Goodness of fit for Social Network Dynamics. Paper presented at the Sunbelt XXXI, St. Petersburg, FL. Matsaganis, M. D., & Payne, J. G. (2005). Agenda Setting in a Culture of Fear: The Lasting Effects of September 11 on American Politics and Journalism American Behavioral Scientist, 49(3), 379-392. doi: 10.1177/0002764205282049 Matthes, J. (2006). The need for orientation towards news media: Revising and validating a classic concept. International Journal of Public Opinion Research, 18(4), 422- 444. doi: 10.1093/ijpor/edh118 Matthes, J. (2008). Need for orientation as a predictor of agenda-setting effects: Causal evidence from a two-wave panel study. International Journal of Public Opinion Research, 20(4), 440-453. doi: 10.1093/ijpor/edn042 McChesney, R. W., & Nichols, J. (2010). The Death and Life of American Journalism: The Media Revolution that Will Begin the World Again. Philadelphia, PA: Nation Books. McCombs, M. (1981). The agenda-setting approach. In D. D. Nimmo & K. R. Sanders (Eds.), Handbook of Political Communication (pp. 121-140). Beverly Hills, CA: Sage Publications. 157 McCombs, M. (1997). Building Consensus: The News Media's Agenda-Setting Roles. Political Communication, 14(4), 433-443. doi: 10.1080/105846097199236 McCombs, M. (2000). The Agenda-Setting Role of the Mass Media in the Shaping of Public Opinion. Paper presented at the LSE Mass Media Economics Conference, London, UK. McCombs, M. (2004). Setting the Agenda: The Mass Media and Public Opinion. Cambridge, UK: Polity Press. McCombs, M. (2005). A Look at Agenda-setting: past, present and future. Journalism Studies, 6(4), 543-557. doi: 10.1080/14616700500250438 McCombs, M. (2010). Extending our theoretical maps: Psychology of agenda-setting. Central European Journal of Communication, 3(2), 197-206. McCombs, M., & Bell, T. (1996). The Agenda-Setting Role of Mass Communication. In M. B. Salwen & D. W. Stacks (Eds.), An Integrated Approach to Communication Theory and Research (pp. 93-110). Mahwah, NJ: Lawrence Erlbaum Associates. McCombs, M., Danielian, L., & Wanta, W. (1995). Issues in the news and the public agenda: The agenda-setting tradition. In T. L. Glasser & C. T. Salmon (Eds.), Public Opinion and the Communication of Consent (pp. 281-300). New York, NY: Guilford Publications. McCombs, M., Holbert, R. L., Kiousis, S., & Wanta, W. (2011). The News and Public Opinion: Media Effects on Civic Life. Malden, MA: Polity. 158 McCombs, M., & Reynolds, A. (2009). How the News Shapes Our Civic Agenda. In J. Bryant & M. B. Oliver (Eds.), Media Effects: Advances in Theory and Research (3rd ed.). New York, NY: Routledge. McCombs, M., & Shaw, D. (1972). The Agenda-Setting Function of Mass Media. Public Opinion Quarterly, 36(2), 176. doi: 10.1086/267990 McCombs, M., & Valenzuela, S. (2007). The Agenda-Setting Theory. Cuadernos de informacion(20), 44-50. McLeod, J., Becker, L. B., & Byrnes, J. E. (1974). Another Look At the Agenda-Setting Function of the Press. Communication Research, 1(2), 131. doi: 10.1177/009365027400100201 McManus, J. H. (1994). Market-driven Journalism: Let the Citizen Beware? Thousand Oaks, CA: Sage Publications. McManus, J. H. (1995). A market-based model of news production. Communication Theory, 5(4), 301-338. doi: 10.1111/j.1468-2885.1995.tb00113.x McQuail, D. (1992). Media performance: Mass communication and the public interest. Thousand Oaks, CA: Sage Publications. Meijer, M. M., & Kleinnijenhuis, J. (2006). Issue news and corporate reputation: Applying the theories of agenda setting and issue ownership in the field of business communication. Journal of Communication, 56(3), 543-559. doi: 10.1111/j.1460-2466.2006.00300.x 159 Meraz, S. M. (2009). Is There an Elite Hold? Traditional Media to Social Media Agenda Setting Influence in Blog Networks. Journal of Computer-Mediated Communication, 14(3), 682-707. doi: 10.1111/j.1083-6101.2009.01458.x Meraz, S. M. (2011). The Fight for ‘How to Think’: Traditional Media, Social Networks, and Issue Interpretation. Journalism, 12(1), 107-127. doi: 10.1177/1464884910385193 Merritt, D., & McCombs, M. (2004). The Two W's of Journalism: The Why And What Of Public Affairs Reporting. Mahwah, NJ: Lawrence Erlbaum Associates. Messner, M., & DiStaso, M. (2008). The source cycle: How traditional media and weblogs use each other as sources. Journalism Studies, 9(3), 447-463. doi: 10.1080/14616700801999287 Messner, M., & Garrison, B. (2010). Sources without a name: An analysis of the source interaction between elite traditional news media and filter blogs. Paper presented at the Association for Education in Journalism and Mass Communication (AEJMC) Annual Conference, Denver, CO. Miel, P., & Faris, R. (2008). News and information as digital media come of age Media Re:public: Berkman Center for Internet & Society. Miller, J. M., & Krosnick, J. A. (2000). News media impact on the ingredients of presidential evaluations: Politically knowledgeable citizens are guided by a trusted source. American Journal of Political Science, 44(2), 301-315. 160 Mitchelstein, E., & Boczkowski, P. J. (2009). Between tradition and change: A review of recent research on online news production. Journalism, 10(5), 562. Monge, P., & Contractor, N. (2003). Theories of Communication Networks. New York, NY: Oxford University Press, USA. Monge, P., Heiss, B. M., & Margolin, D. B. (2008). Communication network evolution in organizational communities. Communication Theory, 18(4), 449-477. doi: 10.1111/j.1468-2885.2008.00330.x Moon, S. J. (2011). Attention, Attitude, and Behavior: Second-Level Agenda-Setting Effects as a Mediator of Media Use and Political Participation. Communication Research. doi: 10.1177/0093650211423021 Nagler, M. G. (2007). Understanding the Internet's Relevance to Media Ownership Policy a Model of Too Many Choices. The B.E. Journal of Economic Analysis & Policy, 7(1), 1-29. Napoli, P. M. (1999). Deconstructing the diversity principle. Journal of Communication, 49(4), 7-34. doi: 10.1111/j.1460-2466.1999.tb02815.x Napoli, P. M. (2001). Diversity and Localism: A policy analysis perspective. Roundtable on Media Ownership Policies, Federal Communications Commission. Retrieved from transition.fcc.gov/ownership/roundtable_docs/napoli-stmt.pdf 161 Napoli, P. M. (2011a). Diminished, Enduring, and Emergent Diversity Policy Concerns in an Evolving Media Environment. International Journal of Communication, 5, 1182–1196. Napoli, P. M. (2011b). Exposure Diversity Reconsidered. Journal of Information Policy, 1, 246-259. Napoli, P. M., & Gillis, N. (2008). Media Ownership and the Diversity Index: Outlining a Social Science Research Agenda. McGannon Center Working Paper Series. Retrieved from fordham.bepress.com/mcgannon_working_papers/5 Neuman, W. R. (1991). The Future of the Mass Audience. New York, NY: Cambridge University Press. Neuman, W. R., Just, M. R., & Crigler, A. N. (1992). Common knowledge: News and the construction of political meaning. Chicago, IL: University of Chicago Press. Neuman, W. R., Marcus, G. E., Crigler, A. N., & MacKuen, M. (Eds.). (2007). The Affect Effect: Dynamics of Emotion in Political Thinking and Behavior. Chicago, IL: University of Chicago Press. O'NeiIi, D., & Harcup, T. (2008). News Values and Selectivity. Handbook of Journalism Studies, 161. O'Neill, D., & Harcup, T. (2009). News Values and Selectivity. In K. Wahl-Jorgensen & T. Hanitzsch (Eds.), Handbook of Journalism Studies (pp. 161-174). New York, NY: Routledge. 162 Ognyanova, K., & Monge, P. (in press). A Multilevel, Multidimensional Network Model of the Media System: Production, Content, and Audiences. Communication Yearbook. Oh, J., Susarla, A., & Tan, Y. (2008). Examining the Diffusion of User-Generated Content in Online Social Networks. SSRN. Retrieved from ssrn.com/abstract=1182631 Paletz, D. L. (1999). The media in American politics: Contents and consequences. New York, NY: Longman Publishing Group. Papper, B. (2011). RTDNA / Hofstra University Annual Survey, Women and Minorities Data. Washington, DC: Radio-Television News Directors Association. Paterson, C. (2006). News Agency Dominance in International News on the Internet. In D. Skinner, J. Compton & M. Gasher (Eds.), Converging Media, Diverging Interests: A Political Economy of News in the United States and Canada (pp. 145- 164). Lanham, MD: Rowman & Littlefield Publishers. Pew Internet & American Life Project. (2012). Latest Statistics: Trend Data. Retrieved July 21st, 2012, 2012, from www.pewinternet.org/Static-Pages/Trend-Data- (Adults).aspx Pew Project for Excellence in Journalism. (2008). 2008 News Coverage Index Dataset. Washington, DC: Pew Research Center. 163 Pew Project For Excellence in Journalism. (2010a). How News Happens: A Study of the News Ecosystem of One American City Retrieved June 21st, 2012, from www.journalism.org/analysis_report/how_news_happens Pew Project For Excellence in Journalism. (2010b). New Media, Old Media: The Blogosphere Retrieved June 21st, 2012, from www.journalism.org/analysis_report/blogosphere Pew Project For Excellence in Journalism. (2010c). The State of The News Media 2010: An Annual Report on American Journalism The State of the News Media. Pew Project For Excellence in Journalism. (2011). The State of The News Media 2011: An Annual Report on American Journalism The State of the News Media. Protess, D. L., & McCombs, M. (Eds.). (1991). Agenda Setting: Readings on Media, Public Opinion, and Policymaking. Hillsdale, NJ: Lawrence Erlbaum Associates. R Development Core Team. (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from www.R-project.org Ragas, M. W., & Kiousis, S. (2010). Intermedia agenda-setting and political activism: MoveOn.org and the 2008 presidential elections. Mass Communication & Society, 13(5), 560-583. doi: 10.1080/15205436.2010.515372 Reese, S. D. (1991). Setting the media’s agenda: A power balance perspective. Communication yearbook, 14, 309-340. 164 Reese, S. D., & Danielian, L. H. (1989). Intermedia influence and the drug issue: converging on cocaine. In P. J. Shoemaker (Ed.), Communication campaigns about drugs: Government, media, and the public. Hillsdale, NJ: Lawrence Erlbaum Associates. Reese, S. D., Rutigliano, L., Hyun, K., & Jeong, J. (2007). Mapping the blogosphere: Citizen based media in the global news arena. Journalism: Theory, Practice, Criticism, 8(3), 235 262. Ripley, R. M., Snijders, T. A. B., & Preciado, P. (2012). Manual for SIENA. Oxford, UK. Roberts, M., & McCombs, M. (1994). Agenda Setting and Political Advertising: Origins of the News Agenda. Political Communication, 11(3), 249-262. doi: 10.1080/10584609.1994.9963030 Roessler, P. (2008). Agenda-Setting, Framing and Priming. In W. Donsbach & M. W. Traugott (Eds.), The Sage Handbook of Public Opinion Research (pp. 205). Thousand Oaks, CA: Sage Publications. Rogers, E. M., & Dearing, J. W. (1988). Agenda-Setting Research: Where Has It Been, Where Is It Going? Communication Yearbook, 11, 555-594. Rogers, E. M., Dearing, J. W., & Bregman, D. (1993). The Anatomy of Agenda-Setting Research. Journal of Communication, 43(2), 68-84. doi: 10.1111/j.1460- 2466.1993.tb01263.x 165 Rosenstiel, T. (2010). Americans Spending More Time Following the News. Washington, DC: Pew Research Center for the People and the Press. RStudio. (2012). RStudio: Integrated development environment for R (Version 0.96.330) [Statistical software]. Boston, MA. Retrieved from www.rstudio.org Sayre, B., Bode, L., Shah, D. V., Wilcox, D., & Shah, C. (2010). Agenda Setting in a Digital Age: Tracking Attention to California Proposition 8 in Social Media, Online News and Conventional News. Policy & Internet, 2(2), 2. doi: 10.2202/1944-2866.1040 Scheufele, D. (2000). Agenda-Setting, Priming, and Framing Revisited: Another Look at Cognitive Effects of Political Communication. Mass Communication & Society, 3(2-3), 297-316. doi: 10.1207/S15327825MCS0323_07 Scheufele, D., & Tewksbury, D. (2007). Framing, Agenda Setting, and Priming: The Evolution of Three Media Effects Models. Journal of Communication, 57(1), 9- 20. doi: 10.1111/j.1460-2466.2006.00326.x Schudson, M. (1989). The sociology of news production. Media, Culture & Society, 11(3), 263. doi: 10.1177/016344389011003002 Semetko, H. A. (2009). Media and Public Diplomacy in Times of War and Crisis. American Behavioral Scientist, 52(5), 639-642. 166 Shah, D. V., McLeod, D., Gotlieb, M., & Lee, N.-J. (2009). Framing and Agenda-Setting. In R. L. Nabi & M. B. Oliver (Eds.), The Sage Handbook of Media Processes and Effects (pp. 83-98). Thousand Oaks, CA: Sage Publications. Shaw, D., & Martin, S. (1992). The function of mass media agenda setting. Journalism & Mass Communication Quarterly, 69(4), 902-920. doi: 10.1177/107769909206900410 Shaw, D., & McCombs, M. (Eds.). (1977). The emergence of American political issues: The agenda-setting function of the press. St. Paul, MN: West. Shaw, D., McCombs, M., Weaver, D., & Hamm, B. (1999). Individuals, groups, and agenda melding: A theory of social dissonance. International Journal of Public Opinion Research, 11(1), 2-24. doi: 10.1093/ijpor/11.1.2 Sheafer, T., & Weimann, G. (2005). Agenda Building, Agenda Setting, Priming, Individual Voting Intentions, and the Aggregate Results: An Analysis of Four Israeli Elections. Journal of Communication, 55(2), 347-365. doi: 10.1111/j.1460- 2466.2005.tb02676.x Shoemaker, P. J. (1991). Gatekeeping. Newbury Park, CA: Sage Publications. Shoemaker, P. J. (1996). Hardwired for news: Using biological and cultural evolution to explain the surveillance function. Journal of Communication, 46(3), 32-47. doi: 10.1111/j.1460-2466.1996.tb01487.x 167 Shoemaker, P. J. (2006). News and Newsworthiness: A Commentary. Communications, 31(1), 105-111. doi: COMMUN.2006.007 Shoemaker, P. J., Chang, T. K., & Brendlinger, N. (1987). Deviance as a predictor of newsworthiness: Coverage of international events in the US media. Communication yearbook, 10, 348-365. Shoemaker, P. J., & Cohen, A. A. (2006). News around the world: Content, practitioners, and the public. New York, NY: Routledge. Shoemaker, P. J., Danielian, L. H., & Brendlinger, N. (1991). Deviant acts, risky business and US interests: The newsworthiness of world events. Journalism & Mass Communication Quarterly, 68(4), 781-795. doi: 10.1177/107769909106800419 Shoemaker, P. J., Johnson, P. R., Seo, H., & Wang, X. (2010). Readers as Gatekeepers of Online News: Brazil, China, and the United States. Brazilian Journalism Research, 6(1). Shoemaker, P. J., Kim, E., & Wrigley, B. (2001). Individual and routine forces in gatekeeping. Journalism & Mass Communication Quarterly, 78(2), 233-246. doi: 10.1177/107769900107800202 Shoemaker, P. J., & Reese, S. D. (1996). Mediating the Message. White Plains, NY: Longman USA. 168 Shoemaker, P. J., & Vos, T. P. (1996). Media Gatekeeping. In M. B. Salwen & D. W. Stacks (Eds.), An Integrated Approach to Communication Theory and Research (pp. 79-92). Mahwah, NJ: Lawrence Erlbaum Associates. Shoemaker, P. J., & Vos, T. P. (2009). Gatekeeping theory. New York, NY: Routledge. Singer, J. B. (2005). The political j-blogger ‘normalizing’a new media form to fit old norms and practices. Journalism, 6(2), 173-198. doi: 10.1177/1464884905051009 Smith, A., & Brenner, J. (2012). Twitter Use 2012. Washington, DC: Pew Internet & American Life Project. Snijders, T. A. B. (2012a). RSiena: Siena - Simulation Investigation for Empirical Network Analysis (Version 1.1-212) [R package]. Retrieved from CRAN.R- project.org/package=RSiena Snijders, T. A. B. (2012b). RSienaTest: Siena - Simulation Investigation for Empirical Network Analysis. (Version 1.1-216/r216) [R package]. Retrieved from R- Forge.R-project.org/projects/rsiena/ Snijders, T. A. B., Van de Bunt, G. G., & Steglich, C. E. G. (2010). Introduction to stochastic actor-based models for network dynamics. Social Networks, 32(1), 44- 60. doi: dx.doi.org/10.1016/j.socnet.2009.02.004 Stevenson, R. L. (2003). Wagging the Dog: Agenda-Setting and Inter-Media Competition. Paper presented at the Popular Culture Association Conference, New Orleans. 169 Stromback, J., & Kiousis, S. (2010). A New Look at Agenda-Setting Effects - Comparing the Predictive Power of Overall Political News Consumption and Specific News Media Consumption Across Different Media Channels and Media Types. Journal of Communication, 60(2), 271-292. doi: 10.1111/j.1460-2466.2010.01482.x Stroud, N. J. (2011). Niche news: The politics of news choice. New York, NY: Oxford Univ Press. Sunstein, C. R. (2007). Republic.com 2.0. Princeton, NJ: Princeton University Press. Sunstein, C. R. (2009). Going to Extremes: How Like Minds Unite and Divide. New York, NY: Oxford University Press. Tai, Z. (2009). The structure of knowledge and dynamics of scholarly communication in agenda setting research, 1996–2005. Journal of Communication, 59(3), 481-513. doi: 10.1111/j.1460-2466.2009.01425.x Takeshita, T. (2006). Current Critical Problems in Agenda-Setting Research. International Journal of Public Opinion Research, 18(3), 275. doi: 10.1093/ijpor/edh104 Tedesco, J. C. (2001). Issue and strategy agenda-setting in the 2000 presidential primaries. American Behavioral Scientist, 44(12), 2048-2067. doi: 10.1177/00027640121958483 170 Tedesco, J. C. (2005). Issue and strategy agenda setting in the 2004 presidential election: Exploring the candidate–journalist relationship. Journalism Studies, 6(2), 187- 201. doi: 10.1080/14616700500057270 Tewksbury, D. (2005). The Seeds of Audience Fragmentation: Specialization in the Use of Online News Sites. Journal of Broadcasting and Electronic Media, 49, 332- 348. doi: 10.1207/s15506878jobem4903_5 Tuchman, G. (1972). Objectivity as strategic ritual: An examination of newsmen's notions of objectivity. American Journal of Sociology, 660-679. doi: 10.1086/225193 Tuchman, G. (1978a). Making news: A study in the construction of reality. New York, NY: Free Press. Tuchman, G. (1978b). The News Net. Social Research, 45(2), 253-276. Uzzi, B., & Spiro, J. (2005). Collaboration and Creativity: The Small World Problem. American Journal of Sociology, 111(2), 447-504. doi: 10.1086/432782 Valcke, P. (2011). Looking For the User in Media Pluralism Regulation: Unraveling the Traditional Diversity Chain and Recent Trends of User Empowerment in European Media Regulation. Journal of Information Policy, 1(0), 287-320. Valcke, P., Lefever, K., Kerremans, R., Kuczerawy, A., Dunstan, M., Chanter, J., . . . Fug, O. (2009). Independent Study on Indicators for Media Pluralism in the 171 Member States - Towards a Risk-Based Approach. Brussels, Belguim: European Commission, Directorate General Information Society and Media. Valente, T. W. (1996). Social network thresholds in the diffusion of innovations. Social Networks, 18(1), 69-89. doi: 10.1016/0378-8733(95)00256-1 Valenzuela, S., & McCombs, M. (2009). The Agenda-Setting Role of the News Media. In M. B. Salwen & D. W. Stacks (Eds.), An Integrated Approach to Communication Theory and Research (2nd ed., pp. 90-106). New York, NY: Routledge. Vlieger, E., & Leydesdorff, L. (2011). Content Analysis and the Measurement of Meaning: The Visualization of Frames in Collections of Messages. The Public Journal of Semiotics, 1, 28-50. Waldman, S. (2011). The Information Needs of Communities: The Changing Media Landscape in a Broadband Age Working Group on Information Needs of Communities: Federal Communications Commission. Wallsten, K. (2007). Agenda Setting and the Blogosphere: An Analysis of the Relationship between Mainstream Media and Political Blogs. Review of Policy Research, 24(6), 567-587. doi: 10.1111/j.1541-1338.2007.00300.x Wanta, W. (1997a). The messenger and the message: differences across news media. In M. McCombs, D. Shaw & D. Weaver (Eds.), Communication and democracy: Exploring the intellectual frontiers of agenda-setting theory (pp. 137-151). Mahwah, NJ: Lawrence Elrbaum Associates. 172 Wanta, W. (1997b). The Public and the National Agenda: How People Learn about Important Issues. Mahwah, NJ: Lawrence Erlbaum Associates. Wanta, W., & Foote, J. (1994). The President - News Media Relationship: A Time Series Analysis of Agenda-Setting. Journal of Broadcasting & Electronic Media, 38(4), 437-448. doi: 10.1080/08838159409364277 Wanta, W., & Ghanem, S. (2007). Effects of Agenda-Setting. In R. W. Preiss, B. M. Gayle, N. Burrell, M. Allen & J. Bryant (Eds.), Mass media effects research: Advances through meta-analysis (pp. 37-51). New York, NY: Lawrence Erlbaum Associates. Wanta, W., Golan, G., & Lee, C. (2004). Agenda Setting and International News: Media Influence on Public Perceptions of Foreign Nations. Journalism & Mass Communication Quarterly, 81(2), 364-377. doi: 10.1177/107769900408100209 Wanta, W., & Hu, Y. W. (1994). The Effects of Credibility, Reliance, and Exposure on Media Agenda-Setting: A Path Analysis Model. Journalism & Mass Communication Quarterly, 71(1), 90-98. doi: 10.1177/107769909407100109 Wanta, W., & Wu, Y. C. (1992). Interpersonal communication and the agenda-setting process. Journalism & Mass Communication Quarterly, 69(4), 847-855. doi: 10.1177/107769909206900405 Wasserman, S., & Faust, K. (1994). Social Network Analysis: Methods and Applications. New York, NY: Cambridge University Press. 173 Weaver, D. (1984). Media agenda-setting and public opinion: Is there a link? Communication Yearbook, 8, 680-691. Weaver, D., McCombs, M., & Shaw, D. (2004). Agenda-Setting Research: Issues, Attributes, and Influences. In L. L. Kaid (Ed.), Handbook of Political Communication Research (pp. 257-283). Mahwah, NJ: Lawrence Erlbaum Associates. Weber, M., & Monge, P. (2011). The Flow of Digital News in a Network of Authorities, Hubs and Providers. Journal of Communication, 61(6), 1062–1081. doi: 10.1111/j.1460-2466.2011.01596.x Webster, J. G. (2005). Beneath the veneer of fragmentation: Television audience polarization in a multichannel world. Journal of Communication, 55(2), 366-382. doi: 10.1111/j.1460-2466.2005.tb02677.x Webster, J. G. (2007). Diversity of Exposure. In P. M. Napoli (Ed.), Media Diversity and Localism: Meaning and Metrics (pp. 309-325). Mahwah, NJ: Lawrence Erlbaum Associates. Webster, J. G., & Ksiazek, T. B. (2012). The Dynamics of Audience Fragmentation: Public Attention in an Age of Digital Media. Journal of Communication, 62(1), 39-56. doi: 10.1111/j.1460-2466.2011.01616.x Webster, J. G., & Lin, S. F. (2002). The Internet Audience: Web Use as Mass Behavior. Journal of Broadcasting & Electronic Media, 46(1), 1-12. doi: 10.1207/s15506878jobem4601_1 174 Williams, B. A., & Delli Carpini, M. X. (2000). Unchained reaction: The collapse of media gatekeeping and the Clinton–Lewinsky scandal. Journalism, 1(1), 61. doi: 10.1177/146488490000100113 Williams, B. A., & Delli Carpini, M. X. (2004). Monica and Bill all the time and everywhere: The Collapse of Gatekeeping and Agenda Setting in the New Media Environment. American Behavioral Scientist, 47(9), 1208-1230. doi: 10.1177/0002764203262344 Williams, B. A., & Delli Carpini, M. X. (2011). After Broadcast News: Media Regimes, Democracy, and the New Information Environment. New York, NY: Cambridge University Press. Wu, H. D. (2007). A Brave New World for International News? Exploring the Determinants of the Coverage of Foreign News on US Websites. International Communication Gazette, 69(6), 539. doi: 10.1177/1748048507082841 Wu, S., Hofman, J. M., Watts, D. J., & Mason, W. A. (2010). Who Says What to Whom on Twitter Categories and Subject Descriptors. Paper presented at the International World Wide Web Conference, Hyderabad, India. Yang, J., & Stone, G. (2003). The Powerful Role of Interpersonal Communication in Agenda Setting. Mass Communication & Society, 6(1), 57-74. doi: 10.1207/S15327825MCS0601_5 Yu, J., & Aikat, D. (2005). News on the Web: Agenda setting of online news in Web sites of major newspaper, television and online news services. Paper presented at the 175 Annual Meeting of the International Communication Association (ICA), New York, NY. Zuckerman, E. (2008). International News: Bringing about the Golden Age Media Re:public: Berkman Center for Internet and Society at Harvard University. 176 APPENDIX A: NEWS OUTLET LIST News outlets included in the PEJ News Coverage Index 2008 Media Outlet Media Type 1 Albuquerque Journal Newspaper 2 Arkansas Democrat-Gazette Newspaper 3 Austin American-Statesman Newspaper 4 Bakersfield Californian Newspaper 5 Boston Globe Newspaper 6 Chattanooga Times Free Press Newspaper 7 Chicago Tribune Newspaper 8 LA Times Newspaper 9 MetroWest Daily News Newspaper 10 Modesto Bee Newspaper 11 New Hampshire Union Leader Newspaper 12 NY Times Newspaper 13 Philadephia Inquirer Newspaper 14 San Francisco Chronicle Newspaper 15 Star Beacon (OH) Newspaper 16 Star Tribune (MN) Newspaper 17 Sun Chronicle (MA) Newspaper 18 The Gazette (Colorado Springs) Newspaper 19 USA Today Newspaper 20 Wall Street Journal Newspaper 21 Washington Post Newspaper 22 AOL News Website 23 CNN.COM Website 24 Google news Website 25 MSNBC.com Website 26 Yahoo news Website 27 CBS Evening News Network TV 28 Good Morning America Network TV 177 29 NBC Nightly News Network TV 30 Newshour with Jim Lehrer Network TV 31 The Early Show Network TV 32 Today Network TV 33 World News Tonight Network TV 34 1600 Pennsylvania Avenue Cable TV 35 Anderson Cooper 360 Cable TV 36 Campbell Brown: No Bias, No Bull Cable TV 37 CNN Daytime Cable TV 38 CNN Election Center Cable TV 39 CNN unspecified show Cable TV 40 Countdown with Keith Olbermann Cable TV 41 Fox News Daytime Cable TV 42 Fox News unspecified show Cable TV 43 Fox Report with Shepard Smith Cable TV 44 Hannity and Colmes Cable TV 45 Hardball Cable TV 46 Live with Dan Abrams Cable TV 47 Lou Dobbs Tonight Cable TV 48 MSNBC Daytime Cable TV 49 MSNBC unspecified show Cable TV 50 O'Reilly Factor Cable TV 51 Out in the Open Cable TV 52 Race for the White House Cable TV 53 Rachel Maddow Show Cable TV 54 Situation Room Cable TV 55 Special report with Brit Hume Cable TV 56 Tucker Cable TV 57 ABC News Headlines Radio 58 CBS News Headlines Radio 59 Ed Schultz Radio 60 Michael Savage Radio 61 Morning Edition Radio 62 Randi Rhodes Radio 63 Rush Limbaugh Radio 64 Sean Hannity Radio 178 APPENDIX B: NEWS OUTLET ATTRIBUTES Outlet Name Sector Owner NY Times Newspaper New York Times Company Washington Post Newspaper Graham Family Wall Street Journal Newspaper News Corporation USA Today Newspaper Gannett Company Inc. LA Times Newspaper Tribune Co. CNN.COM Online Time Warner, TBS Yahoo news Online Yahoo Inc. MSNBC.com Online NBC Universal, Comcast Google news Online Google AOL News Online Time Warner, TBS World News Tonight Network TV Walt Disney Company Today Network TV NBC Universal, Comcast NBC Nightly News Network TV NBC Universal, Comcast The Early Show Network TV National Amusements, Inc. CBS Evening News Network TV National Amusements, Inc. NewsHour Network TV Public Broadcasting Service CNN Daytime Cable TV Time Warner, TBS Situation Room Cable TV Time Warner, TBS Anderson Cooper 360 Cable TV Time Warner, TBS Lou Dobbs Cable TV Time Warner, TBS MSNBC Daytime Cable TV NBC Universal, Comcast Hardball Cable TV NBC Universal, Comcast Countdown with Keith Olbermann Cable TV NBC Universal, Comcast Fox News Daytime Cable TV News Corporation Special Report with Brit Hume Cable TV News Corporation O'Reilly Factor Cable TV News Corporation Hannity and Colmes Cable TV News Corporation Fox Report w Shepard Smith Cable TV News Corporation Morning Edition Radio National Public Radio, Inc. Rush Limbaugh Radio Clear Channel Communications Sean Hannity Radio Clear Channel Communications Michael Savage Radio Cumulus Media, KGO 179 Outlet Name Audience Size Average Age Average Education Average Income NY Times 2240 52.9 7.5 7.1 Washington Post 815 52.5 7.0 7.1 Wall Street Journal 788 55.0 7.5 7.7 USA Today 966 51.2 6.4 6.7 LA Times 689 53.3 6.6 6.6 CNN.COM 2369 43.7 6.9 6.9 Yahoo news 4548 44.8 6.1 6.3 MSNBC.com 1250 47.1 6.6 6.8 Google news 1003 47.8 6.6 6.7 AOL News 2203 49.8 6.1 6.5 World News Tonight 883 55.6 5.6 5.7 Today 515 48.1 6.0 6.2 NBC Nightly News 1134 53.8 6.0 6.2 The Early Show 1872 55.0 5.1 5.4 CBS Evening News 654 56.1 5.4 5.6 NewsHour 199 60.1 7.4 6.3 CNN Daytime 11422 52.3 5.8 6.0 Situation Room 55 60.2 5.4 5.9 Anderson Cooper 360 71 47.4 6.9 6.7 Lou Dobbs 79 57.3 5.7 5.3 MSNBC Daytime 2316 53.2 6.1 6.3 Hardball 147 52.7 6.7 6.4 Countdown with Keith Olbermann 187 52.5 6.8 6.5 Fox News Daytime 8262 55.7 5.6 6.2 Special Report with Brit Hume 32 47.9 6.4 7.0 O'Reilly Factor 213 59.2 5.7 6.1 Hannity and Colmes 140 53.2 6.1 6.6 Fox Report w Shepard Smith 252 57.3 5.7 5.9 Morning Edition 65 49.7 7.1 6.6 Rush Limbaugh 2702 57.0 5.9 6.4 Sean Hannity 1077 52.0 6.2 6.7 Michael Savage 201 51.8 5.9 6.1 180 Outlet Name Hispanic Audience African-American Audience Party ID Political Ideology NY Times 4.5% 6.4% 44 47 Washington Post 3.8% 18.3% 30 16 Wall Street Journal 3.4% 2.7% -31 -38 USA Today 4.7% 9.2% -3 -16 LA Times 16.8% 10.2% 29 19 CNN.COM 5.1% 10.2% 20 11 Yahoo news 6.8% 9.7% 5 -8 MSNBC.com 3.9% 6.5% 7 -3 Google news 6.4% 6.9% 10 -1 AOL News 5.9% 9.4% 5 -7 World News Tonight 4.4% 5.4% 10 -7 Today 3.1% 4.1% 9 -4 NBC Nightly News 3.2% 3.2% 15 -3 The Early Show 3.5% 9.7% 18 -13 CBS Evening News 2.9% 7.0% 20 -5 NewsHour 3.1% 0.5% 44 46 CNN Daytime 6.2% 14.3% 27 3 Situation Room 3.7% 14.8% 42 33 Anderson Cooper 360 8.5% 8.6% 37 21 Lou Dobbs 5.1% 3.8% 8 -10 MSNBC Daytime 4.2% 11.5% 39 23 Hardball 2.1% 5.5% 52 36 Countdown with Keith Olbermann 1.6% 7.1% 65 65 Fox News Daytime 3.4% 3.0% -47 -64 Special Report with Brit Hume 0.0% 0.0% -69 -88 O'Reilly Factor 1.0% 1.5% -55 -72 Hannity and Colmes 3.6% 0.0% -62 -77 Fox Report w Shepard Smith 4.1% 3.7% -52 -70 Morning Edition 6.3% 4.7% 37 45 Rush Limbaugh 3.0% 2.2% -57 -75 Sean Hannity 3.2% 2.6% -53 -71 Michael Savage 6.0% 1.0% -29 -61 181 APPENDIX C: AGENDA CONVERGENCE NETWORK METRICS I 2008 Agenda Convergence Network: Metrics over time (52 weekly binary graphs) Week # Density Fragmentation In-degree Centralization Out-degree Centralization 1 0.346 0.352 0.542 0.675 2 0.505 0.322 0.478 0.478 3 0.273 0.418 0.484 0.750 4 0.363 0.374 0.425 0.658 5 0.353 0.413 0.535 0.668 6 0.412 0.306 0.573 0.573 7 0.447 0.170 0.505 0.571 8 0.408 0.317 0.578 0.578 9 0.279 0.430 0.611 0.744 10 0.477 0.201 0.507 0.540 11 0.563 0.060 0.452 0.452 12 0.598 0.241 0.415 0.382 13 0.468 0.147 0.516 0.516 14 0.363 0.266 0.591 0.591 15 0.537 0.247 0.478 0.478 16 0.438 0.285 0.547 0.547 17 0.283 0.424 0.473 0.740 18 0.337 0.330 0.585 0.651 19 0.347 0.269 0.508 0.674 20 0.364 0.269 0.557 0.657 21 0.326 0.420 0.530 0.663 22 0.355 0.377 0.566 0.433 23 0.307 0.465 0.548 0.715 24 0.370 0.370 0.517 0.384 25 0.400 0.302 0.519 0.519 26 0.415 0.329 0.570 0.304 27 0.449 0.298 0.503 0.536 28 0.413 0.353 0.606 0.539 29 0.468 0.348 0.516 0.416 30 0.337 0.454 0.552 0.685 31 0.367 0.327 0.620 0.287 32 0.352 0.360 0.602 0.336 33 0.340 0.151 0.482 0.448 34 0.425 0.262 0.493 0.560 35 0.409 0.462 0.477 0.576 36 0.454 0.452 0.464 0.497 182 37 0.471 0.321 0.513 0.546 38 0.469 0.148 0.515 0.448 39 0.434 0.331 0.584 0.550 40 0.390 0.092 0.563 0.596 41 0.449 0.293 0.403 0.569 42 0.373 0.285 0.348 0.647 43 0.369 0.259 0.452 0.651 44 0.425 0.399 0.593 0.560 45 0.595 0.000 0.418 0.418 46 0.468 0.123 0.483 0.416 47 0.383 0.215 0.604 0.370 48 0.411 0.268 0.541 0.441 49 0.482 0.093 0.502 0.402 50 0.511 0.121 0.471 0.471 51 0.517 0.094 0.498 0.332 52 0.408 0.296 0.578 0.344 Mean 0.413 0.287 0.519 0.531 Min 0.273 0.000 0.348 0.287 Max 0.598 0.465 0.620 0.750 183 APPENDIX D: AGENDA CONVERGENCE NETWORK METRICS II 2008 Agenda Convergence Network: Metrics over time (13 four-week binary graphs) Period Density Fragmentation In-degree Centralization Out-degree Centralization 1 0.447 0.352 0.538 0.571 2 0.474 0.322 0.510 0.543 3 0.587 0.418 0.427 0.427 4 0.542 0.374 0.472 0.439 5 0.418 0.413 0.567 0.600 6 0.404 0.306 0.548 0.582 7 0.506 0.170 0.510 0.343 8 0.456 0.317 0.562 0.529 9 0.509 0.430 0.507 0.507 10 0.522 0.201 0.493 0.460 11 0.482 0.060 0.535 0.535 12 0.567 0.241 0.447 0.348 13 0.570 0.147 0.444 0.278 184 APPENDIX E: NEWS OUTLET DEGREE DISTRIBUTIONS Out-Degree Distribution: 2008 Agenda Convergence Network (13 four-week binary graphs) Media Outlet \ Time Period: 1234 5 6 7 8 9 10 11 12 13 NY Times 11 14 15 6 4 7 7 8 6 4 5 1112 Washington Post 17 19 16 19 11 8 13 14 12 10 15 20 18 Wall Street Journal 20 21 21 20 24 8 11 20 17 22 19 26 24 USA Today 21 7 22 21 20 20 20 18 23 22 16 16 23 LA Times 6 11 15 16 8 10 16 12 19 15 11 21 10 CNN.COM 17 15 20 16 15 17 19 16 19 16 13 22 20 Yahoo news 15 13 17 18 18 13 18 15 16 13 13 18 19 MSNBC.com 13 16 15 17 13 11 16 18 18 14 11 21 16 Google news 16 13 15 17 8 12 17 16 14 14 5 14 20 AOL News 16 16 16 19 10 16 18 17 18 12 7 15 20 World News Tonight 9 10 17 16 7 9 13 10 10 6 5 12 16 Today 121316141013159 16 8 5 9 14 NBC Nightly News 9 12 12 15 8 12 13 11 12 9 6 12 13 The Early Show 12 13 19 16 13 14 15 16 17 16 13 15 17 CBS Evening News 12 11 13 14 7 10 15 12 16 11 7 11 17 NewsHour 8 10 11 9 7 11 18 15 13 10 6 15 16 CNN Daytime 13 16 21 18 14 17 19 13 13 21 22 18 21 Situation Room 16 13 19 20 19 18 14 9 15 9 12 18 19 Anderson Cooper 360 17 15 19 22 14 17 16 24 26 29 26 24 22 Lou Dobbs 8 15 22 20 12 4 17 15 7 14 14 19 16 MSNBC Daytime 11 13 22 15 14 19 17 9 21 15 28 22 20 Hardball 32 32 31 20 31 31 23 19 13 31 32 28 24 Countdown, Keith Olbermann 11 2 31 19 6 13 20 23 18 31 32 29 24 Fox News Daytime 9 17 18 22 15 12 13 14 13 15 15 20 21 Special Report, Brit Hume 14 13 9 12 6 5 14 14 13 10 8 13 12 O'Reilly Factor 18 24 24 16 11 6 17 9 30 31 27 25 23 Hannity and Colmes 32 32 32 31 32 20 23 31 32 31 32 28 24 Fox Report w Shepard Smith 13 18 19 17 15 14 15 14 15 16 14 9 19 Morning Edition 8 7 7 6 2 5 8 7 2 5 3 4 7 Rush Limbaugh 27 21 22 24 16 19 23 22 17 31 28 27 20 Sean Hannity 9 24 32 31 32 28 27 22 29 30 32 27 23 Michael Savage 23 26 26 24 25 14 24 12 27 29 28 25 27 185 In-Degree Distribution: 2008 Agenda Convergence Network (13 four-week binary graphs) Media Outlet \ Time Period: 1 2 3 4 5 6 7 8 9 10 11 12 13 NY Times 15 24 25 19 15 17 24 24 28 17 23 21 25 Washington Post 7 11 11 19 10 5 5 8 11 15 14 7 6 Wall Street Journal 4 3 4 1 6 1 1 2 7 2 10 1 1 USA Today 3 4 6 5 3 3 1 3 9 9 10 1 2 LA Times 8 9 13 7 10 3 11 9 10 12 16 15 2 CNN.COM 20 24 26 27 11 16 28 16 21 17 16 18 24 Yahoo news 21 30 30 27 16 24 26 20 27 27 20 18 29 MSNBC.com 24 26 31 24 17 22 25 14 23 22 16 27 29 Google news 9 25 26 26 15 19 24 12 19 21 17 23 30 AOL News 26 25 25 23 17 19 21 11 22 19 14 24 24 World News Tonight 28 30 32 32 31 28 32 32 30 32 28 30 32 Today 25 31 32 31 27 28 31 29 30 31 32 32 32 NBC Nightly News 30 30 32 31 28 30 32 31 29 30 28 32 32 The Early Show 23 28 28 27 23 24 29 29 28 29 20 32 28 CBS Evening News 31 30 31 32 29 28 31 30 28 27 24 31 32 NewsHour 28 31 32 30 28 30 32 32 32 32 29 30 30 CNN Daytime 18 13 21 27 17 15 21 24 27 17 14 22 27 Situation Room 26 10 16 14 10 10 18 22 14 17 13 20 25 Anderson Cooper 360 6 8 18 8 5 4 2 6 8 10 9 13 21 Lou Dobbs 8 7 18 12 9 4 8 13 3 10 17 22 11 MSNBC Daytime 13 9 18 22 14 15 13 15 5 15 11 27 20 Hardball 2 2 6 5 3 2 3 3 5 8 4 8 1 Countdown, Keith Olbermann 3 4 4 3 4 3 1 2 2 6 4 5 2 Fox News Daytime 10 9 17 13 16 10 18 8 21 18 15 15 14 Special Report, Brit Hume 23 15 32 28 13 11 21 29 25 23 20 30 29 O'Reilly Factor 6 4 5 4 4 5 6 6 4 8 9 9 12 Hannity and Colmes 2 2 4 2 3 3 1 2 4 6 4 6 6 Fox Report w Shepard Smith 19 16 25 29 20 18 29 13 18 17 24 30 30 Morning Edition 26 31 32 32 30 29 32 32 31 31 30 32 32 Rush Limbaugh 4 5 6 4 4 2 6 4 6 6 7 6 5 Sean Hannity 3 3 3 2 4 2 1 2 6 8 4 6 2 Michael Savage 4 3 5 4 5 3 1 1 4 8 8 1 2

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Contagious: social norms about health in work group networks

PDF

Civic games with 'local fit': embedding with real‐world neighborhoods and place‐based networks

PDF

Voice, participation & technology in India: communication technology for development in the mobile era

PDF

Network hermeneutics: interpretation of texts as social practice and performance in the age of digital media

PDF

The evolution of knowledge creation online: what is driving the dynamics of Wikipedia networks

PDF

The evolution of scientific collaboration networks

PDF

The evolution of multilevel organizational networks in an online gaming community

PDF

Nosopolitics of postwar trauma: a study in rhetorical networks (1967–2015)

PDF

Food is the medium: food movements, social justice and the communication ecology approach

PDF

Pathways to immigrant civic engagement: the storytelling network and church participation

PDF

Immigrant health in context: a communication ecology approach to understanding health behavior

PDF

Effects of economic incentives on creative project-based networks: communication, collaboration and change in the American film industry, 1998-2010

PDF

Political coming-of-age in the era of WeChat: understanding the ethnic media ecosystem and group politics of first-generation Chinese Americans

PDF

The impact of engaged communication scholarship on the re-imagination of space and place in South L.A.

PDF

Communicating organizational knowledge in a sociomaterial network: the influences of communication load, legitimacy, and credibility on health care best-practice communication

PDF

The 2010 Haiti network relief movement: rhetoric in natural disaster, institutional intervention, and biosecurity

PDF

Media reinvented: the transformation of news in a networked society

PDF

Social networking in the modern zeitgeist: an examination of why we're obsessed and how social media will define our future

PDF

Understanding normative influence of neighborhoods: a multilevel approach to promoting Latinas’ cervical cancer prevention behaviors in urban ethnic communities

PDF

Pundits, panels, but no policy: the normative role of cable news in communicating health care policy to the public

Asset Metadata

Creator Ognyanova, Katherine (author)

Core Title Intermedia agenda setting in an era of fragmentation: applications of network science in the study of mass communication

School Annenberg School for Communication

Degree Doctor of Philosophy

Degree Program Communication

Publication Date 03/21/2013

Defense Date 10/25/2012

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag actor-based modeling,agenda,agenda setting,fragmentation,issue salience,journalism,mass communication,media studies,network analysis,network science,new media,OAI-PMH Harvest,political ideology

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ball-Rokeach, Sandra J. (committee chair), Crigler, Ann N. (committee member), Monge, Peter R. (committee member), Overholser, Geneva (committee member)

Creator Email ognyanov@usc.edu,thesis@ognyanova.net

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-225755

Unique identifier UC11294638

Identifier usctheses-c3-225755 (legacy record id)

Legacy Identifier etd-OgnyanovaK-1478.pdf

Dmrecord 225755

Document Type Dissertation

Rights Ognyanova, Katherine

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA