Student
Fynn Gerken
11110805
Supervisor
Dr. Toni G.L.A. van der Meer
Date of completion
24.06.2016
The dynamic framing process:
Investigating the interplay between news
media and governmental organisations in a
long-lasting crisis
Master’s Thesis
Graduate School of Communication, University of Amsterdam
Master in Communication Science: Corporate Communication
1
Abstract
In times of crisis, the communicative interplay between actors is highly complex and has important implications for the crisis magnitude, evolution, and consequences. This study aims to understand the communicative interaction over time between news media and
governmental organisations, by focusing on the concepts of news frame diversity and frame alignment. The recent Ebola crisis (2014 – 2015), with its international significance and large impact, offered a suitable test case. To empirically analyse the actors’ interplay, a method innovation of semantic-network analysis was conducted to automatically identify frames and their presence in newspaper articles (N = 1,079) and organisational press releases (N = 324). The findings of a time series analysis illustrate how the rise of certain frames (e.g., Contagion frame) pushed others from the media agenda, while the presence of other frames (e.g.,
Support frame) afforded greater plurality in framing. In addition, the results of a reciprocal time series analysis show that a rise in frame alignment caused an increase in news frame diversity. This implies that a level of common interpretation and understanding between actors can foster the openness of news media for a variety of narratives and viewpoints, which might be instrumental for the resolution of a crisis situation.
Keywords: crisis communication, agenda setting, news frame diversity, frame
2
The dynamic framing process: Investigating the interplay between news media and governmental organisations in a long-lasting crisis
In today’s mediated society, actors’ communication is of fundamental importance for the magnitude, evolution, and consequences of a crisis. Indeed, while crises are often caused by external circumstances, actors socially co-constructed the meaning of these successive events in their communicative interplay (Kleinnijenhuis, Schultz, & Oegema, 2015; Schultz & Raupp, 2010). Faced with a continuous stream of information about corporate scandals, terrorist attacks, or disease outbreaks, actors try to make sense of these rapidly unfolding and interlocking events (Schultz & Raupp, 2010; Van der Meer, Verhoeven, Beentjes, &
Vliegenthart, 2014; Weick, 1988). The stakes for involved actors are often high, as their interpretations might disrupt social order (Patriotta, Gond, & Schultz, 2011), inflict panic among the public (Klemm, Das, & Hartmann, 2014), and cause turmoil on the financial markets (Kleinnijenhuis, Schultz, Oegema, & van Atteveldt, 2013). As a result, the
communicative interplay becomes highly complex and dynamic (Kleinnijenhuis et al., 2015). Research in the field of crisis communication has started only recently to focus on the interplay of actors (e.g., Kleinnijenhuis, Schultz, Utz, & Oegema, 2013; Schultz,
Kleinnijenhuis, Oegema, Utz, & van Atteveldt, 2012; Van der Meer et al., 2014). Yet, a research deficit remains regarding the complex understanding of communicative interaction between actors in crisis situations, especially concerning the aspect of time (Schultz et al., 2012). Indeed, while crises are intrinsically related to the notion of time, time-centred research, has largely been absent from the field (Fleischer, 2013). Hence, to further
understand the communicative interplay between actors, a more dynamic approach is required that examines the overtime process of communication that constitute the crisis. This paper pays close attention to the element of time and examines the communicative interaction of news media and governmental organisations (GOs) in a long-lasting crisis.
3
Both actors play a key role for the construction and evolution of the crisis. News media make crucial information available (Sorribes & Rovira, 2011) and offer interpretations of the events (Van Gorp, 2007). This can significantly influence other actors’ crisis
understanding and guide their actions. Thus, news media can cause or prevent crisis escalation (Van der Meer et al., 2014) and hence have a pivotal role for the crisis
development. At the same time, GOs are key holder of information, which makes them an important source for journalist (Liu & Horsley, 2007). As public organisation, GOs have the social function to notify other actors about crisis developments (Lee, 2001; Liu & Horsley, 2007) and thus are important for the reduction of confusion (Liu & Kim, 2011). Despite its relevance, the communicative interplay between these actors has been widely overlooked (Schultz & Raupp, 2010). Especially, since the transferability of results between private and public sector is limited (Horsley & Barker, 2002), further research is urgently needed to understand how these actors interact and co-construct the crisis.
To advance understanding of crisis situations and the actors’ interaction, it is essential to untangle the underlying communicative processes. Framing theory offers a powerful body of literature to study the crisis communication patterns and their evolution over time
(Kleinnijenhuis et al., 2015; Schultz et al., 2012; Snow, Vliegenthart, & Corrigall-Brown, 2007; Van der Meer et al., 2014). Especially, analysing how frames vary in communication and across actors promises critical insight into the communicative interplay and its
consequences. The variety of frames in communication can be studied through the concept of frame diversity (Hung, 2009). Focusing on news frame diversity can reveal how open news media are for actors’ varying interpretations and viewpoints over time. This may have substantial consequences for the trajectory of the crisis, since the predominance of a single frame might signal a breakdown of communication, crisis intensification, and hinder crisis resolution (Kleinnijenhuis et al., 2015). In addition, the concept of frame alignment is useful
4
to compare frames across actors (Snow, Rochford, Worden, & Benford, 1986; Van der Meer et al., 2014). A temporary similarity in framing can indicate that actors have come to common interpretations about the events (Hellsten, Dawson, & Leydesdorff, 2010; Snow et al., 1986), which fosters dialogue and might signal a communicative end of crisis (Van der Meer et al., 2014). Although both concepts are instrumental for the communicative interaction between actors and the resolution of the crisis, it remains unclear whether a degree of openness for alternative views results in common interpretations, or if a level of common interpretations motivates other actors to open up for varying views. Thus, more analysis is needed to understand the causal link between frame alignment and news frame diversity.
In sum, this study aims to shed light on the communicative interaction between GOs and media in a long-standing crisis and to further investigate the concepts of frame alignment, and frame diversity, by addressing the research question: How does the presence of certain
frames and the alignment of frames between actors relate to news frame diversity in a long-lasting crisis? To answer the question, this study applies a method innovation of
semantic-network analysis (Hellsten et al., 2010) on the recent Ebola crises, studying the framing dynamics between U.S. news media and GOs throughout the 54 weeks of crisis.
Theoretical Framework Framing literature
Framing theory offers a powerful approach to investigate the dynamic process of meaning construction and negotiation (Gamson & Modigliani, 1989) and has its roots in the field of psychological as well as sociological (Borah, 2011; Van Gorp, 2007). Research into framing has drastically increase in recent years (Vliegenthart & van Zoonen, 2011) and has become the most employed approach in the field of mass communication (Bryant & Miron, 2004). This, however, has led to a lack of empirical and theoretical consistency, with studies employing diverging operationalisations and conceptualisations (Borah, 2011; Entman, 1993;
5
Scheufele, 1999; Van Gorp, 2007; Vliegenthart & van Zoonen, 2011). As a consequence, the term frame has been ascribed a variety of meanings (Van Gorp, 2007).
The underlying theoretical idea is that a single issue can be regarded from multiple angles (Chong & Druckman, 2007c). One of the first definitions of framing can be traced to classic works of Goffman (1974), who asserted that to categorise seemingly random events and construct meaning, individuals draw on "schemata of interpretation” (p. 21), which he referred to as frames. Closely resembling Goffman’s conceptualisation, Gamson and
Modigliani (1989) referred to frames as “interpretative package” that give meaning to issues enable individuals to make “sense of relevant events” (p. 3). Also drawing on Goffman’s work, Gitlin (1980) defined frames as ‘‘persistent patterns of cognition, interpretation, and presentation, of selection, emphasis, and exclusion’’ (p. 7). Consequently, frames are tools that help individuals organise series of events, ascribe meaning to them, and guide action (Benford & Snow, 2000; Entman, 2007), by bringing certain aspects to the attention of the individuals (De Vreese, 2005; Druckman, 2001). In other words, frames can be understood as constructed realities that help to understand the meaning of issues and events by defining them in ways that allows individuals to understand them (Benford & Snow, 2000; Entman, 2007). In present research, Entman's (1993) definition of framing has established itself as standard reference:
To frame is to select some aspects of a perceived reality and make them more salient in the communicating text, in such a way as to promote a particular problem
definition, causal interpretation, moral evaluation and/or treatment recommendation for the item described. (p. 52)
Crisis framing
Framing also plays a fundamental role in crisis situations. Crises are characterised by a high level of uncertainty and ambiguity for involved actors (e.g., Thelwall & Stuart, 2007;
6
Vasterman, Yzermans, & Dirkzwager, 2005). Since frames are central to the sensmaking process (Weick, 1988), helping individuals to construct meaning of complex events (Benford & Snow, 2000; Entman, 2007), they allow actors to come to an understanding of the crisis (Schultz & Raupp, 2010; Weick, 1988). Thus, frames assist actors in establishing coherence to an issue (Hellsten et al., 2010; Snow et al., 1986) and decrease uncertainty (Leydesdorff & Ivanova, 2014). This makes frames crucial for the prevention of confusion, panic (Liu & Kim, 2011; Van der Meer & Verhoeven, 2013), and crisis escalation (e.g., Seeger, 2002).
News media’s crisis framing. During crises, news media serve as a central platform
for communication, especially when the public health is at risk (Glik, 2007). News media can shape public opinion by calling attention to certain issues (McCombs & Shaw, 1972;
McCombs, 2005) and offering interpretations of the events (Van Gorp, 2007). Often, the public’s meaning construction and actions are based on information provided by news media (Sorribes & Rovira, 2011). Particularly in crisis situations, when the need for information is high, journalists are important gatekeepers, who can influence the crisis through their selection of items and sources (Van der Meer, Verhoeven, Beentjes, & Vliegenthart, 2016). Moreover, they make sense of incidences and organise them into a meaningful succession, thereby influencing the construction of reality of the public (Hallahan, 1999). Thus, news media can spark or prevent crisis escalation (Van der Meer et al., 2014) and ultimately impact the evolution of the crisis (Kleinnijenhuis, Schultz, Utz, et al., 2013).
GOs’ crisis framing. Another key actor in times of crisis are GOs. As public
organisations, they are a key holder of information and have the social function to communicate quickly to protect the public (Lee, 2001; Liu & Horsley, 2007) and reduce confusion (Liu & Kim, 2011). Their corporate communication aims to notified the public about the action they should take to avoid physical harm (Coombs, 2007). For instance, the public might be advised to refrain from using certain products or travel routes. Furthermore,
7
to reduce psychological stress among the public, GOs are expected to disclose actions that are implemented to solve the problem and prevent similar occurrences in the future (Coombs, 2007). The information and decision that are released can have far reaching consequences crisis developments, as other actors are prone to act upon it such as consumer and investors (Kleinnijenhuis et al., 2015).
Crisis framing across actors. Since the meaning of a certain issues can be
constructed, or framed, differently (Van Gorp, 2007), crisis interpretation across actors can vary in their level of similarity and presence in communication over time. Indeed, crisis frames often differ between actors (Van der Meer et al., 2014). For instance, differences in framing over time have been observed during the financial crisis (Schultz & Raupp, 2010), the H1N1 flue pandemic (Liu & Kim, 2011), and the BP oil spill crisis (Schultz et al., 2012). Given the relevance of crisis frames, identifying and analysing frames constructed by news media and GOs is essential to gain a deeper understanding of the meaning that actors assigned to a series of successive events. Moreover, comparing whether and how this meaning
construction differs between actors can offers insight into the interactive negotiation of reality among actors that is central to crisis communication. Therefore, the first research question reads as follows:
Research Question 1 (RQ1): What frames are constructed by news media and
governmental organisations during crisis?
The dynamic framing process
Frames are not static, but develop over time with the evolvement of an issue (Gamson & Modigliani, 1989; Hellsten et al., 2010; Jacobi, van Atteveldt, & Welbers, 2015), as actors engage in the collective construction and negotiation of meaning as part of the framing process (Benford & Snow, 2000; Vliegenthart & van Zoonen, 2011). Since framing is essential during crisis, its evolution is likely to depend upon the dynamic development of
8
frames (e.g., Kleinnijenhuis et al., 2015; Schultz et al., 2012; Van der Meer et al., 2014). Ergo, analysing news frame diversity and frame alignment might provide further insight into the evolution of frames and hence the negation between actors over time in crisis situations.
News frame diversity. Research into the communicative diversity in terms of actors,
issues, and frames in communication has its origins in agenda setting research (Allen & Izcaray, 1998; Jennings, Bevan, & John, 2011; Jennings, Bevan, Timmermans, et al., 2011; John & Jennings, 2010; Peter & Vreese, 2003; Zhu, 1992). Framing research recently has taken up the idea. Hung (2009) studied media frame diversity and its correlation with audience frame diversity, conceptualised as the semantic variety of frames present in text. Hitherto, however, literature on news frame diversity does not provide conclusive evidence how the variety of news media frames influence the evolution of a crisis and what underlying process cause changes in frame diversity. This study focus on news frame diversity
conceptualise as the variety of frames present in news media and their prominence in relation to one another.
In order to gain further insight into diversity in news media framing, it is essential to understand how the individual frames affect frame diversity. The framing environment is contested, with differently strong frames competing in text for salience (Chong & Druckman, 2007a, 2007b). As Entman (2003) pointed out, “the framing of a given issue, or event during a defined time period can be arrayed along a continuum from total dominance by one frame to a completely even-handed standoff between competing frames” (p. 418). This relates to the idea of a zero-sum dynamic in media where a rise in salience to one subject, comes at the cost of the salience of another (Jennings, Bevan, Timmermans, et al., 2011; Zhu, 1992; Zuh & McCombs, 1995). Essentially, it highlights that the rise of certain frames might push other frames from the media agenda, leaving limited room for different narratives, interpretation, and viewpoints.
9
In crisis situations, this dynamic can be expected to be especially pronounced. An outbreak or intensification of a crisis is likely to cause a singularity in perspectives, reducing frame diversity in news media. Indeed, research by Kleinnijenhuis et al. (2015) provides evidence for this process. Their study focused on frame complexity, examining the number of different associations between issues and actors and the connection between the latter present in news media. Thus, framing was considered to be more complex, when more issues and actors co-occurred together in text. Importantly, the study provides evidence that frame complexity decreased at the begin of the financial crisis and increased again with the resolution of the crisis situation.
During times of crisis, problems are likely to push to the forefront of the media agenda, which results in a drop in complexity (Kleinnijenhuis et al., 2015), a phenomenon that has also been observed in seminal complexity research (e.g., Jennings, Bevan, & John, 2011; Suedfeld & Tetlock, 1977). This implies that when the crisis intensifies, frames focusing on, for instance, problems might become predominant in media, leaving limited space for other narratives. As a result, news frame diversity might decrease. Yet, over time the media becomes again more open for a divers set of narratives, thereby producing solutions (Kleinnijenhuis et al., 2015). This increase in narratives would by definition lead to greater news frame diversity. In other words, news frame diversity might increase when frames emphasise a wider set of issues and solutions, which can be collectively discussed.
This dynamic is of fundamental importance for the crisis evolution and its
consequences. A reduction in frame diversity in news media can produce an atmosphere of crisis, alarming actors about an outbreak or identification of a crisis (Kleinnijenhuis et al., 2015). Actors are likely to react to this communicative signal, which can have dramatic consequences such as panic on the financial markets (Kleinnijenhuis, Schultz, Oegema, et al., 2013). In addition, the level of news frame diversity is also likely to have important
10
implications for the resolution of the crisis. As long as a narrow focus on few problems persist “a trade-off with the issues of other stakeholders is more unlikely” (Kleinnijenhuis et al., 2015, p. 4). Thus, until the number of frames on the media agenda has not increased and frame diversity recovered, a resolution might remain distant. Once more narratives become present, solutions can be discussed and the crisis can resolve (Kleinnijenhuis et al., 2015). To provide a more thorough understanding of the concept news frame diversity and how it develops in the competitive framing environment the second research question reads as follows:
Research Question 2 (RQ2): How does the presence of frames affect news frame
diversity in times of crisis?
Frame alignment. Besides the variety of frames in news media, the similarity of
frames between actors can offer further insight into the communicative crisis process. Crisis situation are, by definition, ambiguous and confusing (Coombs, 2007). During crisis, actors desire to resolve differences and come to a collective understanding of the crisis (Snow et al., 1986; Van der Meer et al., 2014). After the individual meaning production, actors engage in collective sensemaking, which is likely to result in a temporal construction of similar frames (Van der Meer et al., 2014). This similarity in frame construction can be understood as frame alignment. Therefore, frame alignment in this paper will be conceptualised as the similarity in presence of comparable frames in actors’ communication. Previous studies have already recognised signs of frame alignment (Schultz et al., 2012; Schultz & Raupp, 2010; Snow et al., 2007) and provided qualitative (Snow et al., 1986) as well as quantitative (Van der Meer et al., 2014) evidence for different degrees of frame alignment throughout crisis.
The level of alignment can also have substantial consequences for the crisis
development. As long as confusion and incoherence is the prevailing state of crisis, a solution for the crisis is improbable (Seeger, 2002; Van der Meer et al., 2014; Weick, 1988). Before a
11
crisis can be solved, actors need to reach a degree of consensus about what happened and what the complex events mean (Van der Meer et al., 2014). As interpretation of events become more similar, actors can reach a common understanding (Snow et al., 1986; Van der Meer, 2014). This makes further escalation is unlikely (Weick, 1988) and creates an important precondition for the resolution of the crisis (Van der Meer et al., 2014).
Frame alignment and news frame diversity. The similarity of frames between actors
and the diversity of frames in news media, are likely to stand in close relationship with each other. The literature indicates that both common interpretations (Seeger, 2002; Snow et al., 1986; Van der Meer et al., 2014; Weick, 1988), and a degree of openness for alternative viewpoints (Kleinnijenhuis et al., 2015; Wong, Ormiston, & Tetlock, 2011) are instrumental for the communicative interaction of actors and the crisis resolution. However, the question arises what, initially, forms the basis for an inter-actor dialogue? Does a degree of openness for alternative views result in common interpretations, or is first a degree of common interpretations across actors required to motivate actors to open up for other opinions?
On the one hand, it might be that first an intra-actor shift from a close focus on few problems to a divers set of narratives is required (Kleinnijenhuis et al., 2015) to reach a common understanding what the events signify. Often, an increasing complexity promotes an increase in perspectives on the situation and openness for new information (Wong et al., 2011). This openness, in turn, might motivate actors to interact (Kleinnijenhuis et al., 2015), which results in a rapprochement of actors’ frames and common understanding (Van der Meer et al., 2014), and ultimately to the resolution of the crisis.
On the other hand, common interpretations might be the precondition for actors to be open for multiple interpretations and viewpoints. It might be, that with an increasing
similarity in actors’ interpretations, a level of mutual understanding can be reached (Van der Meer et al., 2014). This, in turn, might increase actors’ approachability for different
12
viewpoints, thereby providing an important basis for further communicative interaction. The plurality in narratives might, then, further accelerate discussion, as multiple actors are provided with room to satisfy their desire to share speculations about the events (Gilpin & Murphy, 2008). With the increasing complexity, an increase in perspectives is further stimulated and openness for new information motivated (Wong et al., 2011). Thus, with the widening of opinions, away from the narrow set of problems, possible solutions can be discussed and the crisis can come to a communicative end (Kleinnijenhuis et al., 2015).
Since news media provide the central platform for crisis communicate (Glik, 2007), and represents the social space where actors’ multifaceted narratives interact (Van der Meer et al., 2014), the complex communicative process can be disentangled by focusing on frame alignment and its relationship with news frame diversity. Therefore, the third research question is formulated as follows:
Research Question 3 (RQ3): How do frame alignment between news media and GOs
and news frame diversity influence each other over time during crisis?
Method Data collection
To answer the research questions, an automated content analysis of U.S. newspaper coverage and GOs press releases was conducted. The recent Ebola crisis served as a test case, which was selected for several reasons, most notably, because the epidemic provides a unique opportunity to study frame dynamics between GOs and news media over a long period of time. The crisis began in March, 2014 when the first cases were reported in Guinea (WHO, 2015). While the large majority of cases were reported in West Africa, attention in the U.S. was especially high, when four people were diagnosed with the virus in the U.S. in September and October, 2014. Given the significance and the scale of the outbreak, the crisis offers an interesting test case to answer this study’s research question.
13
Newspaper articles and press releases were collected between June 29, 2014 because that was when the attention started to increase in U.S. news media, and July 5, 2015, as attention faded away. For news media, the five U.S. newspapers with highest circulation rates, available with complete articles on LexisNexis, were selected. Relevant articles that appeared in print were selected by applying a search string, containing the words Ebola, EVD (for Ebola virus disease), or EHD (for Ebola hemorrhagic disease). Only headlines were searched to limit the selection to relevant articles, avoiding articles that predominantly focused on other content (e.g., weekly summaries). This resulted in a total sample of 1079 newspaper articles. The selected newspapers were: The New York Times (n = 477), The Washington Post (n = 242), USA Today (n = 139), Daily News New York (n = 125), and The New York Post (n = 96). In total, four GOs were selected and their websites searched for press releases about the outbreak, resulting in a total sample of 324 press releases. The selected GOs were: United Nations (UN) (n = 200), World Health Organisation (WHO) (n = 51), World Bank (n = 40), and The Center for Disease Control and Prevention (CDC) (n = 32), since they are among the most important GOs responding to large scale epidemics. To validate that materials were relevant to the case, a random sampled of press releases and newspaper articles was reviewed.
Operationalisation
Frames. This study applies an inductive method to automatically identify frame in
actors’ crisis communication. More specifically, a semantic-network analysis was conducted, identifying latent patterns in text based on word (co-)occurrences (Hellsten et al., 2010). This automated approach draws on the idea that a text can be seen as a network of words that conveys their meaning, with each network serving as an indicator for the frame their represent (Hellsten et al., 2010). By clustering groups of correlating words, a higher-order structure in text and between text can be identified and frames detected (Hellsten et al., 2010). In other
14
words, analysing the (co-)occurrence of words allows one to quantify meaning in measureable units of analysis, thereby avoiding subjective bias (Hellsten et al., 2010). Research has
already successfully applied this method, for instance, to examine debates over time
(Leydesdorff & Hellsten, 2006), to compare discourses (Leydesdorff, 2005), to explore crises frame dynamics (Van der Meer et al., 2014), and to explore the effect of crisis strategies (Van der Meer, 2014). Thus, frames in this research will be operationalised as (co-)occurrences in communication, which generate latent semantic-networks, that conveys their meaning (adapted from Hellsten et al., 2010).
The main premise of such automated approaches is that documents can be considered as a bag of words, where set of words are sufficient to understand the meaning of text
(Grimmer & Stewart, 2013; Hopkins & King, 2010). This idea is in line with the underlying theoretical assumptions of framing theory that “text contains frames, which are manifested by the presence or absence of certain keywords, stock phrases, stereotyped images, sources of information, and sentences” (Entman, 1993, p. 52). For a more detailed description of why frames can be analysed in this way, see, for example (Grimmer & Stewart, 2013; Hellsten et al., 2010).
In practice, the automated approach requires several steps that are based on a
scientifically published manual by Vlieger & Leydesdorff (2011). In this study, frames were identified in all actors’ documents combined. First, a frequency list, containing the 255 most frequently used words, was created with the help of the programme FrequencyList. Common words and organisational names were removed with a stop-word list as well as the letter ‘s’ at the end of words deleted to avoid plural forms. Second, after a manual revision, the list served as input for the programme FullText, which created a document-term occurrence matrix. Third, the matrix was used to conduct a principle component factor analysis, which was limited to 11 components. To maximise the variable loadings on each factor, Varimax
15
rotation was selected (Field, 2013). The retrieved components represent the frames. Fourth, to ensure the validity of the identified frames, a random sample of documents were reviewed and frames manually identified. Finally, labels were assigned to the frames according to the words that are part of the word clusters that make up the frame.
Fame diversity. Research into communicative diversity (e.g., agenda diversity and
frame diversity) relates to the idea that a difference in presence can be observed for a defined number of categories (e.g., Kleinnijenhuis et al., 2015). In this study, frame presence was measured in frequency of occurrence of each frame, in terms of articles that contained a particular frame.
To analyse the frequency of occurrence in U.S. news media, AmCat (The Amsterdam Content Analysis Toolkit, cf. van Atteveldt, 2008) was utilised1. First, for each frame, a search string was created, combining the words that are part of the word clusters (or frame) in Apache Lucene query language with Bolean operators (AND, OR) and wildcards (*, ?). Words with factor loadings above 0.50 were combined with the operator AND, words relating to the same concept (e.g., traveller, passenger) combined with OR, and wildcards were used (e.g., the word country was reduced to countr*, which enabled searching for both countries and country). The search queries were developed, tested, and improved with the help of samples of the population to ensure the validity of the search strings. Second, all news documents were searched for the presence of these frames in AmCat. The toolkit construed a data matrix with rows representing the time (weekly level), and columns representing frame presence, indicating the variety of frames present in a given week. Third, to calculate frame presence and their prominence, in relation to one another, the (Shannon & Weaver, 1949) entropy measure was used, represented inEquation 1. This is a widely used measure for agenda and frame diversity (Chaffee & Wilson, 1977; Green-Pedersen & Wilkerson, 2006; Hung, 2009; Jennings, Bevan, & John, 2011; Jennings, Bevan, Timmermans, et al., 2011;
16
John & Jennings, 2010; Kleinnijenhuis et al., 2015; Peter & Vreese, 2003; Zuh & McCombs, 1995). The measure is probabilistic and calculates the spread of observations across a defined number of (discrete) categories, which is weighted by the occurrence these categories.
𝐻 = (−1) ∑ 𝑝(𝑥𝑖) ln(𝑝(𝑥𝑖)) 𝑛
𝑖=1
(1)
In the formula, H represent the entropy score which is calculated as the negative sum for all frames of the likelihood that a document x, falls within a particular frame i, multiplied by the natural log of that likelihood. The value of H increases (high degree of entropy) if frames are more equally present; thus indicating greater news frame diversity. The highest possible entropy score would be achieved if all 11 frames were equally present, which would result in taking the natural log of 11 (H = 2.398). Conversely, the value of H decreases (low degree of entropy) if fewer frames are present; thus, indicating lower news frame diversity. Hence, the lowest possible entropy score would be achieved if only one frame were present (H = 0). When no frames were present it was assumed that that 0*ln(0)=0, because logs of zero cannot be calculated.
Frame alignment. The concept of frame alignment relates to the idea that different
actors construct and use frames that differ in their level of similarity at certain points in time (Van der Meer et al., 2014). This study sets out to measure this similarity in framing by comparing frame presence, across actors over time. In other words, by measuring occurrence of a discrete number of frames, for each actor, in each defined time period, the presence of frames can be compared across actors.
In practice, this analysis is divided into several steps. First, the search string for the frames of pervious steps were used to construe a data matrix in AmCat for each actor, with rows representing the time, and columns indicating frame presence. Second, a new dataset was created from these matrices, where the cases represent the frames and a variable each
17
actor and every week (see Appendix A for example). Third, frame presence of both actors were tested by means of Pearson's r Correlations test, week by week. The results provided insight into the level of alignment over time. These correlation scores were then used to create a new variable (frame alignment)2.
Analysis
Partial adjustment autoregressive distribution lag (ADL) model. To test the
influence of frames on frame diversity in news media (RQ2) a partial adjusted ADL (Koyck) model was estimated.Since the data was aggregated in regular time intervals (weekly),
representing a systematic suggestion of values of the dependent variables, time-series analysis is especially suitable. The model is appropriate, given the time series characteristic of the data and the expected unidirectional causal relationship between independent and the dependent variable. In this analysis, frame diversity is the dependent variable, whereas news frames and attention serve as independent variables. The model can be written as in Equation 2:
𝑌𝑡 = 𝛼0+ 𝛼1𝑌𝑡−1+ 𝛽1𝑋𝑡+ 𝛽2𝐶𝑡+ 𝜀𝑡 (2) In this model (t) represents time, where news frame diversity (Yt) is a function of a constant term (𝛼0), plus a fraction of its past value of itself (𝑌𝑡−1), the degree of presence of the frame (𝑋𝑡), the number of documents in that week (𝐶𝑡) plus a random shock (𝜀𝑡). Augmented Dickey-Fuller test for unit root (stationarity) was conducted to decide whether changes in variables (first-order difference) rather than variables should be used for analysis (see Appendix B for test results). After analysis, the series was tested for absence of
autocorrelation in residuals (i.e., white noise).
Vector Autoregression (VAR). Following previous research (Kleinnijenhuis et al.,
2015; Kleinnijenhuis, Schultz, Utz, et al., 2013; Wu, Stevenson, Chen, & Guner, 2002), a VAR model was estimated. To test the causal hypothesis RQ3, VAR is an appropriate model. The analysis tests for bidirectional causality in time series data, and consist of a series of OLS
18
regressions in which each variable is treated as both dependent and independent variable (Vliegenthart, 2014). The model is based on Granger causality, where external causes should only be assumed when the own autoregressive past is not sufficient to explain a current value (Brandt & Williams, 2007). In this model estimation, each value is regress on its former value and the former value of the endogenous variable as well as cross-lagged influence of other variables. This is represented in Equation 3, with two variables x and y, constants 𝛼1 and 𝛼2, and only one time lag:
𝑦𝑡 = 𝛼1+ 𝛽1𝑦𝑡−1+ 𝛽2𝑥𝑡−1+ 𝜀𝑦𝑡; 𝑥𝑡 = 𝛼2+ 𝛽3𝑥𝑡−1+ 𝛽4𝑦𝑡−1+ 𝜀𝑥𝑡 (3) In this analysis, news frame diversity and frame alignment served as endogenous variables, with news attention as exogenous variable to control for. Note that the logarithm of the variable news frame diversity was entered into the model, following seminal literature, to account for non-normal distributions (e.g., Hollanders & Vliegenthart, 2011). Again, the data was analysed on a weekly level and Dickey-Fuller test were conducted to test for stationarity (see Appendix B for test results). The maximum number of lags was limited to three, with the assumption that a direct impact had only occurred within three weeks or less. This restriction is in line with prior literature on media effects (Vliegenthart, 2014) and ensures robust findings due to the limited number of estimated coefficient (Hollanders & Vliegenthart, 2011). Following the common approach in the literature, the final lag length, included in the model, is determined by the Akaike Information Criterion (AIC) (e.g., Hollanders &
Vliegenthart, 2011; Vliegenthart, 2014). After analysis, tested for absence of autocorrelation (in residuals) were conducted (i.e., white noise).
19
Results Frames in communication (RQ1)
A semantic network analysis was conducted, which investigates the frames that were constructed by both actors. In total, 11 frames were identified, labelled, and their presence in communication analysed to answer RQ1. Appendix C shows the developed search strings and
Support 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Victim 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Protection 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Intensification 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Outbreak 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Contagion 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Local infections 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Politics 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Prevention 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Research 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Consequences 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014Sept. 2014Oct. 2014 Nov. 2014Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
20
Figure 1 as well as Table 1 frame presence in communication of news media and GOs.
Table 1. Identified frames in news media and GOs communication.
Frames 1.Support 2.Victim 3.Protection 4.Intesificati
on
5.Outbreak 6.Contagi on
21
In terms of frame presence, the results reveal that news media dominantly focused on the spreading virus and the related risks (Contagion frame: 39.85%). In addition, frames that focused on pandemic outbreak in Africa (Outbreak frame: 15.2%), on the first U.S. case that intensified the crisis (Intensification frame: 16.22%), and on the measures that need to be taken to protect patients and health personnel (Protection frame: 11.87%), were high on the news media agenda. In contrast, GOs heavily emphasised the need for international support (Support frame: 50.92%). Moreover, they focused strongly on the development of the
Indicators example Support, countries, community Family, died, home, Protect, training, worker Texas, Duncan, hospital Leone, border, capital Disease, spread, infected Occurrence News 71 (6.58%) 58 (5.37%) 136 (12.6%) 175 (16.22%) 164 (15.2%) 430 (39.85%) GOs 165 (50.92%) 2 (0.61%) 12 (3.7%) 0 (0%) 35 (10.8%) 101 (31.17%) Total 236 (16.82%) 60 (4.28%) 148 (10.55%) 175 (12.47%) 199 (37.84%) 531 (37.85%) Frames 7.Local Infections
8.Politics 9.Prevention 10.Research 11.Consequences
R2 = 1.62 R2 = 1.55 R2 = 1.4 R2 = 1.32 R2 = 1.31 Indicators example Spencer, Bellevue, quarantine Obama, president, house Screening, airport, temperature Vaccine, trial, research Economic, million, impact Occurrence News 38 (3.52%) 89 (8.24%) 21 (1.95%) 39 (3.61%) 15 (1.39%) GOs 0 (0%) 2 (0.62%) 1 (0.31%) 4 (1.23%) 57 (17.59%) Total 38 (2.71%) 91 (6.49%) 22 (1.57%) 43 (3.14%) 72 (5.13%)
Note. Cells contain number of documents with identified frames. Percentages in parentheses are
calculated based on frame presence relative to overall number of identified documents per actor to enable comparison.
22
virus and its risk (Contagion frame: 31.17%), the economic consequences of the crisis (Consequence frame: 17.59%), and the pandemic outbreak centre in Africa (Outbreak frame: 10.8%). In sum, it becomes evident that while some frames are only (or more strongly) emphasised by the individual actors, others were high on both agendas (e.g., Outbreak frame, Contagion frame).
Frames and news frame diversity: Partial adjusted ADL model (RQ2)
To provide insight into the framing dynamics during a crisis, the association between the specific frames and frame diversity was analysed, answering RQ2. The partial adjusted ADL model is presented Table 2 with news frame diversity serving as dependent variable, and a fraction of its past value, the news frames, and news attention as independent variables.
0 1 2 3 4 5 6 News f ra m e d iv e rs ity 0 5 10 15 20 25 30 35 40 45 50 55 60
June 2014 Jul. 2014 Aug. 2014 Sept. 2014 Oct. 2014 Nov. 2014 Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015 June 2014 Jul. 2014 Aug. 2014 Sept. 2014 Oct. 2014 Nov. 2014 Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
Support Victim Protection Intesification Outbreak
Contagion Local infections Politics Prevention Research
Consequences News frame diversity
23
The analysis demonstrates a positive significant effect of news attention on news diversity, (B* = 1.70, p < .10), meaning for each additional document, significantly increase frame diversity in news media. Turning next to the effect of frames on news frame diversity, the results provide evidence that the rise of certain frames can decrease the variety of frames present in news media. More specifically, the increase in presence of the ‘Contagion frame’ (B* =-1.63, p < .01) and the ‘Intensification frame’ (B* =-1.26, p < .10) in news media significantly decreased news frame diversity. This indicates that a rise in presence of frames in news media that emphasised the spreading the infectious virus (Contagion frame) and the intensification of the crisis situation (Intensification frame) had the potential to push other
Table 2. Partial adjustment autoregressive distribution lag model predicting news frame
diversity based on identified frames.
News frame diversity L.News frame diversity .05 (.141)
News frames Support .51 (.07)* Victim .60 (.08)* Protection .37 (.06) Intensification -1.26 (.40)† Outbreak -.28 (.04) Contagion -1.63 (.04)* Local infections -.29 (.05) Politics .59 (.07) Prevention .10 (.06) Research .16 (.12) Consequences .11 (.17) News attention 1.70 (.02)† R2 0.58
Note: Cells contain standardized (B*) regression coefficients with standard errors (SE).
24
frames from the media agenda. Thus, their growing frame presence seemed to have left limited room for the occurrence of other frames. This is also reflected in Figure 2, which displays the time series of news frame presence and news frame diversity. The graph shows that the present of both frames increased considerably at the same time around September, which led to a visible drop in news frame diversity. In addition, it is also observable that when the ‘Contagion frame’ increased in presence (e.g., around November), decreases in news frame diversity are also visible. An opposite effect, however, is evident for other frames. The results reveal that the increasing presence of the ‘Support frame’ (B* = .51, p < .01) and the ‘Victim frame’ (B* = .60, p < .01), significantly increased news frame diversity. This suggest that the with the increasing presence of these frames, the variety of narratives and
interpretation also increased. In sum, the findings of this analysis seem to indicate a zero-sum dynamic in the framing process, where the rise of certain frames pushes other frames from the agenda, thereby decreasing news frame diversity, whereas other frames seem to foster a plurality of frames in news media and co-exist alongside each other.
Antecedents and effects for frame alignment and news frame diversity: VAR model (R3)
To explore the relationship between frame alignment and news frame diversity a VAR model was estimated. Table 3 displays the estimated reciprocal effects of the two dependent variables. In addition, news attention was included in the model as exogenous variable. The model is absent of autocorrelation and heteroscedastic as in indicate by the Ljung-Box Q test and the Lagrange Multiplier test (Vliegenthart, 2014), revealing that the model is well-specified. Moreover, the model explains a considerable amount of variance of both series, with a R-square value of 0.19 for new fame diversity and 0.41 for frame alignment.
The Granger causality test reveals that frame alignment granger causes news frame diversity, indicated by the significant effect. In other words, the prediction of news frame diversity is increased by taking into account the level of frame alignment between the actors
THE DYNAMIC FRAMING PROCESS
25
in the previous week.
This is also reflected in Figure 3, which displays the times series for news frame diversity and frame alignment with transformed z-scores. The graph shows that news alignment often in- and decreased before news frame diversity, which then developed
Table 3. Granger-causality tests for frame alignment and news frame diversity
Frame alignment News frame diversity Frame alignment
Granger 6.56*
News frame diversity
Granger .19 Ljung-Box Q(20) 16.56 17.86 Lagrange M test (20) 21.38 14.90 AIC 25.55 BIC 38.65 R2 0.19 0.41
Note. *p<.05. The model includes the variable news attention as exogenous variable.
-2 -1 .5 -1 -.5 0 .5 1 1 .5 2
June 2014 Jul. 2014 Aug. 2014 Sept. 2014 Oct. 2014 Nov. 2014 Dec. 2014 Jan. 2015 Feb. 2015 Mar. 2015 Apr. 2015 May 2015 June 2015
News frame diversity Frame alignment
26
accordingly. The impulse response function provides further insight (see Appendix D for graph). The function indicates that a one-point increase frame alignment triggered an increase in 0.28 of news frame diversity in the following week. While the effect is positive and
immediate, it seems the impact of the effect declines slowly, but gradually during the
following weeks. In sum, the results indicate that when frames between actors become more similar, the diversity of frames in media increases, thereby answering RQ3.
Discussion and Conclusion
The aim of this study was to shed light on the overtime process of communication between actors during a long-lasting crisis, by answering the overall RQ: How does the
presence of certain frames and the alignment of frames between actors relate to news frame diversity in a long lasting crisis? To answer this question an innovation of semantic-network
analysis was conducted that automatically identified frames and their presence in text. The recent Ebola crisis served as a test case to study the interplay between governmental and news media crisis communications over 54 weeks.
First, this study automatically identified crisis frames that both actors had constructed (RQ2). The results reveal that frames such as the ‘Outbreak frame’ and ‘Contagion frame’ were salient in both actors’ communication. At the same time, other frames were actor specific (e.g., GOs: Support frame; U.S. news media: Intensification frame), revealing a degree of differences in framing. Surprisingly, the results suggest that frames that emphasised the intensification of the crisis due to the first Ebola case in the U.S. (Intensification frame) and the later Ebola infection on U.S. soil (Local infection frame) were largely absent from the GOs agenda. On the one hand, this might suggest that GOs payed little attention to these incidents, possibly because the millions of new cases in West Africa outweighed the few U.S. Ebola infections. On the other, GOs might have adopted a more strategic approach. The organisations might have used the high level of attention to voice the need for more support, a
27
frame that was high on the GOs agenda. In fact, such a tactic was recently recommended in crisis health literature (e.g., Reintjes et al., 2016).
Second, a partial adjusted ADL model was estimated, investigating the relationship between the presence of identified frames and news frame diversity (RQ2). The analysis provides evidence that a rise in presence of certain frames, decreases the variety in frames present in news media. More specifically, with an increasing presence of the ‘Contagion frame’ and the ‘Intensification frame’, limited space seemed to have been left for varying interpretations and hence news frame diversity decreased. Strikingly, both fames seemed to have focused on the problems of the crisis such as the continuous spreading of the virus (Contagion frame). The drop in news frame diversity can, therefore, possibly be attributed to the intensification of the crisis. A perceived crisis escalation might have resulted in a narrow focus (Suedfeld & Tetlock, 1977) and fostered a concentration on problems (e.g., Jennings, Bevan, & John, 2011; Kleinnijenhuis et al., 2015). Due to the limited capacity of news media (Jennings, Bevan, Timmermans, et al., 2011; Zhu, 1992; Zuh & McCombs, 1995), alternative viewpoints and interpretations remained absent, crowded out by these predominant negative frames (Entman, 1991, 2003). As a consequence, news frame diversity decreases.
Such decline in diversity might have substantially affected the crisis evolution and its consequences. The predominance of a limited number of frames on the news agenda can send strong signals of crisis intensification which, for instance, can impact finical markets and consumer confidence (Kleinnijenhuis et al., 2015). Even more fundamentally, low news frame diversity might foster a continuous focus on problems (Wong et al., 2011) and a
reduces the chance of a “trade-off” with other actors (Kleinnijenhuis et al., 2015, p. 4). Hence, as long as the variety of frames remains low, solutions are unlikely to be discussed,
compromises improbable, and hence a crisis resolution remains distant.
28
and the ‘Victim frame’, increased news frame diversity. The finding suggests that a rising emphasis on these frames resulted in a plurality in framing of news media. This effect might be related to a progressing dialogue and exchange of information. Previous research already observed that with the advancement of the crisis, news media became more open for a diverse set of narratives (Kleinnijenhuis et al., 2015), which increases diversity and fosters a variety of perspectives on the situation (Wong et al., 2011). Especially the ‘Support frame’ is in line with this idea, since a greater involvement of the international community might be debated as solution to the crisis. Thus, it might be that, as actors’ communication turned to the discussion of solutions, communicative intensity increased.
Finally, VAR model was estimated to analyse the relationship between the level frame alignment between actors and news frame diversity. The underlying question was whether openness to alternative views creates the basis for common understanding, or if a degree of common interpretations of the crisis is at the core of being open for a variety of
interpretations. The model provides evidence that an increase in frame alignment produced greater frame diversity in news media, thereby answering RQ3. The direction of the
relationship seems to suggest that with an increasing rapprochement of narratives and opinions, a degree of mutual understanding between actors about the meaning of the events was reached (Hellsten et al., 2010; Snow et al., 2007; Van der Meer et al., 2014). This appears to have been instrumental for the opening of the media for diverse narratives, which likely fostered communication among actors (Van der Meer et al., 2014). As actors were provided the space to give their interpretation and debate the events (Gilpin & Murphy, 2008), communicative interaction throve and frame diversity further increased.
Thus, the increase in alignment might have been the starting point for the resolution of the situation. With increasing alignment of frames, actors seemed to have been more open to contradicting viewpoints (Wong et al., 2011) and nurtured discussion of solution and
29
compromises (Kleinnijenhuis et al., 2015). In addition, the degree of alignment might have served as an indicator for involved actors that a degree of consensus had been reached (Van der Meer et al., 2014), the situation de-escalated (Weick, 1988), and hence the crisis soon would reach a communicative end.
In addition, it is noteworthy that frame alignment between actors was characterised by changing phases of alignment and de-alignment, with news frame diversity often successively changing, over the course of the crisis. This not only confirms previous literature that
observed specific patterns of frame alignment in short-lasting crisis, (Van der Meer et al., 2014), but also suggest that a similar dynamic might occur repeatedly in a long-lasting crisis. Each crisis intensification might trigger a temporary alignment of frames, cause by the need to understand the crisis event, which subsidises as the situation resolves (Van der Meer et al., 2014).
The results of this study contribute to the body of literature on crisis communication as well as framing theory and have certain practical implications. First, the findings enrich the crisis literature by providing further insight into the dynamic framing process over time, thereby answering the call for a more complex and dynamic investigation of the
communicative interplay between actors in times of crisis (Fleischer, 2013; Schultz et al., 2012). Second, by focusing on organisations from the public sector, this study started to fill an important gap and enhances understanding of GOs interaction with news media in times of crisis (Horsley & Barker, 2002; Liu & Horsley, 2007; Schultz & Raupp, 2010). Third, this study contributes to the literature on an empirical level, by proposing an innovative way to quantify the presence of automatically identified frames in lager amount of text. Moreover, a new way of analysing frame alignment was put forward based on document co-occurrence. Fourth, on a practical level, crisis managers need to realise that capacity of news with regard to frame presence is limited. When a limited number of frames predominate the news during
30
crisis, other frames are likely to be pushed aside. Yet, since common interpretations seems to have fostered the openness of news media for varying narratives, timely communicative interaction between the actors seemed to be key to resolving the crisis.
While this study makes some valuable contributions, a number of shortcomings must be considered when interpreting the results. First, the findings are based on a case study, which limits generalisability (Petersen, 2008). Second, the automated analysis started from the assumption that bags of words are sufficient to retrieve the general meaning of the text. Although this neglecting of syntax enables analysing large amount of text (Grimmer &
Stewart, 2013), it bears the risk to obscure relevant information. Finally, despite the automatic identification of frames, the analysis faced a degree of subjective interference, because frame presence was identified based on a manually created search strings. The chosen method to determine the strings resulted, for some frames, in a limited number and selective
combination of words. Although the resulting search strings were carefully validated, this might have influenced the findings. Therefore, future research should further validate the results and see whether the findings of the Ebola crisis also hold in other crisis situations – especially in the private sector.
This study, despite these limitations, reached its objective to understand the
underlying communicative dynamics that influence the construction and evolution of crisis over time, by analysing the concepts of news frame diversity, and frame alignment in the communicative interplay between GOs and news media during a long-lasting crisis.
31
Notes
1. AmCat (van Atteveldt, 2008) is an online platform that offers a navigator and database that can be used for document management, keyword-based analysis, and other
content analysis task. See http://amcat.nl.
2. Since this automated approach enables analysis of large amount of text processing, all data could be included in the analysis. Therefore, (random) sampling could be avoided and correlation scores were used in the subsequent analysis regardless of the
32
References
Allen, R. L., & Izcaray, F. (1998). Nominal agenda diversity in a media-rich, less-developed society. Communication Research, 15(1), 29–50.
http://doi.org/10.1177/009365088015001002
Benford, R. D., & Snow, D. A. (2000). Framing processes and social movements: An overview and assessment. Annual Review of Sociology, 26(2000), 611–639. http://doi.org/http://dx.doi.org/10.1146/annurev.soc.26.1.611
Borah, P. (2011). Conceptual issues in framing theory: A systematic examination of a decade’s literature. Journal of Communication, 61(2), 246–263.
http://doi.org/10.1111/j.1460-2466.2011.01539.x
Brandt, P. T., & Williams, J. T. (2007). Multiple time series models. Thousand Oaks: SAGE Publications.
Bryant, J., & Miron, D. (2004). Theory and research in mass communication. Journal of
Communication, 54(4), 662–704. http://doi.org/10.1093/joc/54.4.662
Chaffee, S. H., & Wilson, D. G. (1977). Media rich, media poor: Two studies of diversity in agenda-holding. Journalism & Mass Communication Quarterly, 54(3), 466–476. http://doi.org/10.1177/107769907705400304
Chong, D., & Druckman, J. N. (2007a). A theory of framing and opinion formation in competitive elite environments. Journal of Communication, 57(1), 99–118. http://doi.org/10.1111/j.1460-2466.2006.00331.x
Chong, D., & Druckman, J. N. (2007b). Framing public opinion in competitive democracies.
American Political Science Review, 101(4), 637–655.
http://doi.org/10.1017/S0003055407070554
Chong, D., & Druckman, J. N. (2007c). Framing theory. Annual Review of Political Science,
33
Coombs, W. T. (2007). Ongoing crisis communication: Planning, managing, and responding. Los Angeles: SAGE.
De Vreese, C. H. (2005). News framing: Theory and typology. Information Design Journal,
13(1), 51–62. http://doi.org/0.1075/idjdd.13.1.06vre
Druckman, J. N. (2001). The of framing effects for citizen competence. Political Behavior,
23(3), 225–256. http://doi.org/doi.org/10.1023/A:1015006907312
Entman, R. M. (1991). Framing U.S. coverage of international news: Contrasts in narratives of the KAL and Iran Air incidents. Journal of Communication, 41(4), 6–27.
http://doi.org/10.1111/j.1460-2466.1991.tb02328.x
Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. Journal of
Communication, 43(4), 51–58. http://doi.org/10.1111/j.1460-2466.1993.tb01304.x
Entman, R. M. (2003). Cascading activation: Contesting the White House’s frame after 9/11.
Political Communication, 20(4), 415–432. http://doi.org/10.1080/10584600390244176
Entman, R. M. (2007). Framing bias: Media in the distribution of power. Journal of
Communication, 57(1), 163–173. http://doi.org/10.1111/j.1460-2466.2006.00336.x
Field, A. (2013). Exploratory Factor Analysis. In M. Verboord (Ed.), Quantitative methods in
media and communication (pp. 207–261). London: Sage Publications.
Fleischer, J. (2013). Time and crisis. Public Management Review, 15(3), 313–329. http://doi.org/10.1080/14719037.2013.769852
Gamson, W. A., & Modigliani, A. (1989). Media discourse and public opinion on nuclear power: A constructionist approach. American Journal of Sociology, 95(1), 1–37. http://doi.org/http://dx.doi.org/10.1086/229213
Gilpin, D. R., & Murphy, P. J. (2008). Crisis management in a complex world. New York: Oxford University Press. http://doi.org/10.1017/CBO9781107415324.004
34
Gitlin, T. (1980). The whole world is watching: Mass media in the making & unmaking of the
new left. Berkeley: University of California Press.
Glik, D. C. (2007). Risk communication for public health emergencies. Annual Review of
Public Health, 28, 33–54.
http://doi.org/http://dx.doi.org/10.1146/annurev.publhealth.28.021406.144123
Goffman, E. (1974). Frame analysis: An essay of the organization of experience. New York: Harper and Row.
Green-Pedersen, C., & Wilkerson, J. (2006). How agenda-setting attributes shape politics: Basic dilemmas, problem attention and health politics developments in Denmark and the US. Journal of European Public Policy, 13(7), 1039–1052.
http://doi.org/10.1080/13501760600924092
Grimmer, J., & Stewart, B. M. (2013). Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political Analysis, 21(3), 267–297. http://doi.org/10.1093/pan/mps028
Hallahan, K. (1999). Seven models of framing: Implications for public relations. Journal of
Public Relations Research, 11(3), 205–242.
http://doi.org/10.1207/s1532754xjprr1103_02
Hellsten, I., Dawson, J., & Leydesdorff, L. (2010). Implicit media frames: Automated analysis of public debate on artificial sweeteners. Public Understanding of Science,
19(5), 590–608. http://doi.org/10.1177/0963662509343136
Hollanders, D., & Vliegenthart, R. (2011). The influence of negative newspaper coverage on consumer confidence: The Dutch case. Journal of Economic Psychology, 32(3), 367– 373. http://doi.org/10.1016/j.joep.2011.01.003
35
Horsley, J. S., & Barker, R. T. (2002). Toward a synthesis model for crisis communication in the public sector: An initial investigation. Journal of Business and Technical
Communication, 16(4), 406–440. http://doi.org/10.1177/105065102236525
Hung. (2009). Frame-rich, frame-poor: An investigation of the contingent effects of media frame diversity and idividual difference on audience frame diversity. International
Journal of Public Opinion Research, 22(1). http://doi.org/10.1093/ijpor/edp024
Jacobi, C., van Atteveldt, W., & Welbers, K. (2015). Quantitative analysis of large amounts of journalistic texts using topic modelling. Digital Journalism, 1–18.
http://doi.org/10.1080/21670811.2015.1093271
Jennings, W., Bevan, S., & John, P. (2011). The agenda of British government: The speech from the throne, 1911-2008. Political Studies, 59(1), 74–98.
http://doi.org/10.1111/j.1467-9248.2010.00859.x
Jennings, W., Bevan, S., Timmermans, A., Breeman, G., Brouard, S., Chaques-Bonafont, L., … Palau, A. M. (2011). Effects of the core functions of government on the diversity of executive agendas. Comparative Political Studies, 44(8), 1–30.
http://doi.org/10.1177/0010414011405165
John, P., & Jennings, W. (2010). Punctuations and turning points in British politics: The policy agenda of the Queen’s speech, 1940–2005. British Journal of Political Science,
40(03), 561–586. http://doi.org/10.1017/S0007123409990068
Kleinnijenhuis, J., Schultz, F., & Oegema, D. (2015). Frame complexity and the financial crisis: A comparison of the United States, the United Kingdom, and Germany in the period 2007-2012. Journal of Communication, 65(1), 1–23.
36
Kleinnijenhuis, J., Schultz, F., Oegema, D., & van Atteveldt, W. (2013). Financial news and market panics in the age of high-frequency sentiment trading algorithms. Journalism,
14(2), 271–291. http://doi.org/10.1177/1464884912468375
Kleinnijenhuis, J., Schultz, F., Utz, S., & Oegema, D. (2013). The mediating role of the news in the BP oil spill crisis 2010: How U.S. news is influenced by public relations and in turn influences public awareness, foreign news, and the share price. Communication
Research, 42(3), 408–428. http://doi.org/10.1177/0093650213510940
Klemm, C., Das, E., & Hartmann, T. (2014). Swine flu and hype: A systematic review of media dramatization of the H1N1 influenza pandemic. Journal of Risk Research,
9877(September), 37–41. http://doi.org/10.1080/13669877.2014.923029
Lee, M. (2001). Intersectoral differences in public affairs: The duty of public reporting in public administration. Journal of Public Affairs, 2(2), 33–43.
http://doi.org/10.1002/pa.92
Leydesdorff, L. (2005). Metaphors and diaphors in science communication: Mapping the case of stem cell research. Science Communication, 27(1), 64–99.
http://doi.org/10.1177/1075547005278346
Leydesdorff, L., & Hellsten, I. (2006). Measuring the meanings of words in contexts: Automated analysis of “Monarch Butterflies”, “Frankenfoods” and “Stem Cells”.
Scientometrics, 67(2), 231–258.
Leydesdorff, L., & Ivanova, I. A. (2014). Mutual redundancies in interhuman communication systems: Steps toward a calculus of processing meaning. Journal of the American
Society for Information Science and Technology, 65(2), 386–399.
37
Liu, B. F., & Horsley, J. S. (2007). The government communication decision wheel: Toward a public relations model for the public sector. Journal of Public Relations Research,
19(4), 377–393. http://doi.org/10.1080/10627260701402473
Liu, B. F., & Kim, S. (2011). How organizations framed the 2009 H1N1 pandemic via social and traditional media: Implications for U.S. health communicators. Public Relations
Review, 37(3), 233–244. http://doi.org/10.1016/j.pubrev.2011.03.005
McCombs, M. (2005). A look at agenda-setting: Past, present and future. Journalism Studies,
6(4), 543–557. http://doi.org/10.1080/14616700500250438
McCombs, M., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public
Opinion Quarterly, 36(2), 176–187. http://doi.org/10.1086/267990
Patriotta, G., Gond, J.-P., & Schultz, F. (2011). Maintaining legitimacy: Controversies, orders of worth, and public justifications. Journal of Management Studies, 48(8), 1804–1836. http://doi.org/10.1111/j.1467-6486.2010.00990.x
Peter, J., & Vreese, C. H. de. (2003). Agenda-rich, agenda-poor: A cross-national
comparative investigation of nominal and thematic public agenda diversity. International
Journal of Public Opinion Research, 15(1), 55–64.
http://doi.org/http://dx.doi.org/10.1093/ijpor/15.1.44
Petersen, N. J. (2008). Designing a rigorous small sample study. In Best practices in
quantitative methods (pp. 137–147). Thousand Oaks, California: SAGE Publications,
Inc.
Reintjes, R., Das, E., Klemm, C., Richardus, J. H., Keßler, V., & Ahmad, A. (2016). “Pandemic public health paradox”: Time series analysis of the 2009/10 influenza A / H1N1 epidemiology, media attention, risk perception and public reactions in 5 european countries. Plos One, 11(3), e0151258. http://doi.org/10.1371/journal.pone.0151258