Master’s Thesis
Revisiting Intermedia Agenda-Setting and News Values in a
Hybrid Media System: How Twitter Dominates the News Agenda
on ‘Fridays for Future’
Lara Heinz (12306630)
MSc Political Communication
Graduate School of Communication
University of Amsterdam
Date: January 31
st, 2020
Supervisor: Dr. Damian Trilling
Abstract
With the rise of social media, journalists have lost their monopoly on the production of news. The special affordances of digital media, such as speed and unlimited time or content, make them potentially important agenda-setters alongside traditional media. The purpose of this study is to explore the intermedia attention dynamics and differences in news values between tweets and newspaper articles on the topic of ‘Fridays for Future’. Results of time-series analyses suggest a unidirectional relationship from Twitter’s news agenda on newspapers’ agenda. In terms of news treatment, a quantitative content analysis revealed that news values are used much less in tweets compared to newspaper articles, which was, especially regarding the news values ‘controversy’ and ‘emotionality’, astonishing. The findings imply that in the context of protest coverage, Twitter is the decisive agenda-setter but deviates significantly from traditional media logic on the use of news values. The results urge scholars to move towards a more nuanced understanding of the dynamics between different media and their respective characteristics in the
construction of news in an increasingly complex media environment.
Keywords: news values, intermedia agenda-setting, hybrid media system, Twitter,
Revisiting Intermedia Agenda-Setting and News Values in a Hybrid Media System: How Twitter Dominates the News Agenda on ‘Fridays for Future’
Social media have revolutionized the news environment by altering traditional relationships between content producers and consumers. The platforms are no longer only used for interpersonal communication, but also for the ‘(quasi-journalistic) gathering and sharing of news information’ (Araujo & van der Meer, 2018, p. 2). Citizens can communicate their view of political reality to a wider audience through user-generated-content, competing with professional journalism (Eilders, Geißler, Hallermayer, Noghero, & Schnurr, 2010). However, only few studies examine social and traditional media at the same time regarding their differences and intermedia dynamics in news production. There is a great need for further research since, with the expanded opportunities for participation in the political discourse, the conditions of publicly mediated communication have decisively changed. First and foremost, digital media challenge the mediating role of journalists as gatekeepers and agenda-setter. Through social media, citizens gain influence on the political agenda and on the discussion alongside the political and media elites. Despite this development, traditional mainstream media are still assessed as the most powerful agenda-setter for public debate (Djerf-Pierre & Shehata, 2017).
The media coverage is influenced by many factors (see hierarchy influences model, Shoemaker & Reese, 1991). Although, real-world events are the foundations of news
production, the media’s inner logic decides whether and how stories are covered. This article uses news value theory as a theoretical framework to understand how journalists and Twitter users spin a story by emphasizing certain characteristics. Thereby, one news outlet can also give another one an impulse to follow a story. The theory of intermedia agenda-setting deals with the ability of one content producer to influence another in terms of issue salience. In times where the ‘public arena appears to be more crowded, diverse, competitive, and unpredictable than ever’ (Waldherr, 2018, p. 293), the decisive agenda-setter is hard to
identify. Apart from real-world events and the influence from other media, also
inner-dynamics and the own agenda’s history determine today’s news. Statistically speaking, ‘news coverage may then be seen as a partly autoregressive process: news being news, because it was news yesterday’ (Hollanders & Vliegenthart, 2008, p. 48). This study uses time-series analysis to look at the attention dynamics for both media and the influence they have on each other’s agenda levels.
Much of the research on social media has been focused on Twitter due to the
platform’s specific affordances (Bucher & Helmond, 2017). Given the wide range of groups, who communicate via tweets, not only citizens but also professionals like politicians and journalists, the microblog platform has become a powerful tool in political communication (Kwak, Lee, Park, & Moon, 2010; Boukes, 2019). Twitter is not only an instrument for receiving and disseminating information but can be used for mobilizing people to participate in protests (Earl, McKee-Hurwitz, Messinas, Tolan, & Arlotti, 2013; Jungherr & Jürgens, 2014). Due to Twitter’s digital architecture including the hashtag feature and the retweet function, political messages and calls for action can spread rapidly and widely (Segesten & Bossetta, 2016). The environmental movement 'Fridays for Future' has also taken advantage of the platform’s special features in order to mobilize people.
At the ‘Fridays for Future’ school strikes thousands of students worldwide are taking their frustrations to the streets, demanding imminent action against man-made climate change. The movement, started by 15-year-old Swedish activist Greta Thunberg in 2018, brought new attention to the issue of climate change. While the movement quickly spread in other
countries, the ‘Fridays for Future’ demonstrations in Germany took on the largest, long-lasting proportions within Europe. The climate movement also influenced the political agenda pressuring the German government to pass a climate change policy package. It also might have contributed to the great success of the Greens in the European Election in May 2019, where they emerged as the second-strongest force in Germany, reaching an all-time high with
20.5 percent of the vote (European Parliament, 2019). The dimensions and impacts of the movement make it an important effort to examine the media coverage about it and justify the societal relevance of this research. This study narrows the topic of climate change down to the media coverage of the ‘Fridays for Future’ movement in order to investigate the differences in news values rather than issue differences (Boukes & Vliegenthart, 2017).
The theories of news values and agenda-setting date back to a time when the relationship between journalists and the audience was a one-way, asymmetrical
communication model. The current level of knowledge about the applicability of traditional media theories to the contemporary ‘hybrid media system’ (Chadwick, 2011) is rather limited and empirical studies have been inconsistent so far. Since both traditional and social media data have been publicly accessible for a significant amount of time, new possibilities for (time-series) analysis and further research methodologies have emerged (Neuman,
Guggenheim, Jang, & Bae, 2014). This study advances current research through a quantitative analysis between social and traditional media that combines intermedia agenda-setting and news value theory to acquire knowledge on how media report about a civic movement. The overall research question that will be investigated on the issue of ‘Fridays for Future’ in Germany is:
What are the attention dynamics between newspapers’ and Twitter’s agenda levels; and does the protest coverage differ between both media in terms of news values?
Theoretical Background
Intermedia Agenda-Setting
First, the origin and further development of the (intermedia) agenda-setting theory is examined in more detail, before the concept is applied to the contemporary modern media system. The idea of agenda-setting can be traced back to Walter Lippmann’s work ‘Public Opinion’ (1922), in which he argued that the images in our heads are determined by a
pseudo-environment constructed by the news media. Accordingly, the issues closely covered in news reports are becoming the focus of public attention, while laying the foundation of public opinion. Half a century later, McCombs and Shaw (1972) coined the term ‘agenda-setting’ with their well-known ‘Chapel Hill study’, marking the starting point for many future
empirical studies on agenda-setting. According to that theory, news media tell people ‘what to think about’ by transmitting issue salience from the media agenda to the public agenda
(McCombs & Shaw, 1972, p. 22). Since then, many researches have provided evidence of the congruency between priorities of media agenda and public agenda in many different
geographical and historical settings on a wide variety of public issues (McCombs, 2014). With the emergence of more communication channels and the resulting complex dynamics of issue agenda, a new era of agenda-setting research, focused on reciprocal influences, has begun. The effects of institutional media agendas on each other are called ‘intermedia agenda-setting’ and at first have been mainly investigated between newspapers and TV (Golan, 2006). While in the past the ‘New York Times’ or ‘Washington Post’ have proven to influence the content of other media, the directionality of the agenda-setting influence has become blurred since social media have become more popular. To describe the ever-evolving relationship between different types of media that are characterized by adaptation, interdependences and transition, Chadwick (2013) developed the concept of the ‘hybrid media system’.
Hybrid Media System.
The hybrid media system presents an ontological concept beyond dichotomies, like simple distinctions between old and new media or professional writers and amateur bloggers, that have dominated debates about political communication in the past. The contemporary political information cycle diffuses power to a more diverse range of actors and detaches itself from the predominance of journalistic and political elites (Chadwick, 2011). The hybrid media system with its different media logics and platforms’ specificities offers new
opportunities for non-elite actors to shape information flows. While the horizontal nature of Twitter makes it a good platform for activist groups to discuss, coordinate and mobilize, the presence in traditional media is (still) essential to target policy elites, legitimize a movement and increase the movement’s visibility for a wider audience (Shehata & Strömbäck, 2013). Although online media help to circumvent the gatekeeping function of traditional mass media, studies found especially broadcast media to be powerful actors in steering public opinion (Nielsen & Schrøder, 2014).
So far, studies on intermedia agenda-setting in a hybrid media system have produced inconsistent results. Whereas some see Twitter as a filter and amplifier of interesting news taken from the traditional media (e.g. Asur, Huberman, Szabo, & Wang, 2011; Reese, Lou, Kideuk, & Jaekwan, 2007), Rogstad (2016) found that Twitter gives attention to issues that do not attract much interest from traditional media such as environmental challenges. Taking up this point, Su and Borah (2019) studied the intermedia agenda-setting effect between Twitter and newspapers before and after the US announcement to withdraw from the Paris
Agreement. They found Twitter to be more influential in terms of breaking news, whereas newspapers had a more long-lasting impact during times of ongoing debates. Neuman et al. (2014) concluded that social media emphasize social and public order issues more strongly than traditional media. Araujo and van der Meer (2018) found that tweets about organizations had a positive influence on their traditional media coverage, indicating a ‘reversed’ agenda-setting effect. All these studies emphasize the changing role of timing in news production and the development of an ever-accelerating news cycle driven by social media. According to Sevenans and van Aelst (2017), the contemporary 24/7 news ecology is characterized by immediacy, which has become a production norm for journalists and, at the same time, a consumer expectation. Due to their fixed publication schedule, it is sometimes impossible for broadcast media and print newspapers to meet these requirements, whereas social media has no restrictions on time and space.
Since there is no definite conclusion about the relationship between social media and mainstream media in the context of agenda-setting, it is of necessity to reach a higher degree of understanding in this matter. For this study, intermedia agenda-setting theory is used to assess the flow of influence between the online presence of traditional, major German newspapers and German tweets on the climate movement ‘Fridays for Future’. Intermedia agenda-setting will be investigated in the context of a people driven political cause. Given the bottom-up nature of a movement, the direction of influences as well as the order of cause and effect is even more volatile and ambiguous (Neuman et al., 2014). The interdependencies between older and more recent media and the online accessibility of its data make it
imperative to re-test and revisit the classical model of intermedia agenda-setting. In this study, intermedia agenda-setting is not understood as a mechanical causal linkage from one media to the other, but as a dynamic interaction in a hybrid media environment. Accordingly, the first research question is:
RQ 1: How do Twitter and newspaper influence each other’s agenda about ‘Fridays
for Future’ in a hybrid media system?
While many agenda-setting studies have used Pearson correlations or rank-order analyses to investigate overtime dynamics between media agendas (Wanta & Ghanem, 2007), this study will use the more sophisticated approach of time-series analysis. Since no clear direction of influence can be derived from the theory, the Vector Autoregression (VAR) model which treats Twitter’s and newspapers’ agenda levels as dependent and independent variables at the same time, is chosen for the analysis. Thus, the approach is exploratory and no hypothesis is put forward.
In addition to the intermedia dynamics regarding the issue salience of 'Fridays for Future', the news items will be examined on a content level using news value theory.
News Value Theory
Like the origins of agenda-setting theory, the beginnings of news value research can also be traced back to Lippmann’s ‘Public Opinion’ (1922), where he first identified a set of ‘newsworthy’ attributes. But it was several decades later when Galtung and Ruge (1965) laid the foundation of news value research by providing the first systematic list of news factors1. They proposed twelve hypothetical news factors to help explain the reporting of three major foreign crisis in the Norwegian press. Since the publication of this landmark study, many researches have undertaken content analyses for many different news outlets, across time and countries (Harcup & O'Neill, 2001; for a summary see Caple & Bednarek, 2016).
This study follows Eilders’ (2006) argumentation that news values do not only guide journalistic work but also how the audience selects the news it consumes. In her view, news values should be conceptualized as general relevance indicators helping consumers of news to process information and direct attention to the meaningful. Reception studies showed that journalists and recipients process events in very similar ways. Theories explaining how news factors impact the news selection by audiences range from biological evolution to cognitive and social psychology (Shoemaker & Cohen, 2006). The conceptualization of news factors as relevance indicators also links back to Galtung and Ruge (1965, p. 71) who argue that every participant in the news process is guided by the same principles. Following this
argumentation, social media users apply the same relevance criteria in their role as consumers and producers of news. This allows a comparison of news published on social media with professional journalism in traditional media.
Causal vs. Functional Model.
In general, news values can be studied from two theoretical perspectives: the causal and the functional model. This study follows Staab’s (1990) functional model by analyzing how journalists and social media users construct newsworthiness. According to the functional perspective, news values are not inherent event characteristics, but the outcome of journalistic decisions. However, the beginnings of news value research were subject to the assumption of the causal model that news values of an event determine whether and how extensively and prominently a story will be covered: The more news values are present in an event, the more likely it will pass the journalistic gates and become news (Boukes & Vliegenthart, 2017). This causal model leaves the journalist with a rather passive role that does not correspond to
journalists’ autonomy in the production of news. It is much more likely that news values tell more about how instead of why a story has been covered. In that regard, newsworthiness is considered a discursive construct (Caple & Bednarek, 2016) used by journalists to stress certain aspects in order to sell a story as ‘news’ to the audience. Correspondingly, news values like ‘personification’, ‘controversy’ or ‘prominence’ are not intrinsic qualities of a story but can easily be used to make the story newsworthy and legitimize its reporting.
Another reason to understand news values as outcome rather than cause of journalistic decisions, is that content analysis, the method mostly used to investigate news values, only analyzes characteristics of published news (Boukes & Vliegenthart, 2017). To measure the causal model, events that were selected for coverage must be compared to those which were not. Consequently, analyzing news values by conducting content analysis reveals more about news treatment than news selection (O'Neill & Harcup, 2009).
News Values on Social Media.
News values are not stable but have to be updated due to social, cultural and economic changes (O'Neill & Harcup, 2009), the most influential one being the emergence of social
media and the dissolving boundaries between journalists as producers and citizens as consumers of news. Via social media, citizens can take over the role of traditional media: transmitting information, criticizing and controlling the political system. If tweets are concerned with this mediating function and deal with topics of general relevance and topicality, they compete with professional journalism. Thus, it can be assumed that social media users apply news values to construct newsworthiness just as journalists. It is reasonable to expect, that reality constructions by non-journalistic users are not free of media logic (Eilders et al., 2010). A social media user’s decision to send a tweet is comparable to the decision of a journalist to publish a text, both try to anticipate if the audience or network will be interested in a story. Nevertheless, there are also clear differences between journalists and social media users. Unlike journalists, social media users are not bound by media
organizations and routine practices (Shoemaker & Reese, 1991). They can choose topics and assign news values to a story according to their individual preferences or, to go one step further, their ideologies. As traditional and social media aim at a different audience and have different attentional dynamics, the news values in each medium could diverge accordingly. The study investigates whether professional journalists and social media users use different news values when covering a protest movement. Hence, the second research question to be answered within this paper is:
RQ2: Which news values are used (a) by journalists and (b) by Twitter users when
reporting or posting about ‘Fridays for Future’?
A content analytical approach is necessary to examine the differential use of news values and different ways of storytelling in tweets and newspaper articles. There has been a lot of research on audiences’ news values preferences on social networks. Studies about sharing behavior have found that high-arousal content with strong positive or negative sentiment is more likely to be shared (Berger & Milkman, 2012). While Kim (2015) found
‘controversy’ and ‘emotions’ to be key driver for virality, García-Perdome, Salaverría, Kilgo and Harlow (2018) concluded that sharing and interacting was mainly triggered by the news values ‘human interest’, ‘conflict’ and ‘controversy’. Trilling, Tolochko and Burscher (2017) transferred the concept of ‘newsworthiness’ to the digital news system of social
recommendations and coined the term ‘shareworthiness’. In their analysis of sharing behavior in the Netherlands, the news values ‘proximity’, ‘conflict’ and ‘human interest’ have proven to be the most important ones. Knowing how social recommendations in a networked audience work, is key to understand which and how news stories are covered. Unlike social media users, journalists must adhere to the principle of objectivity and meet certain quality criteria for reporting. This leads to the following hypothesis:
H1: The news values ‘emotionality’, ‘controversy’, ‘negative valence’ and ‘positive
valence’ are more prominent in tweets than in newspaper articles.
Method
Data Collection
For the first sample, newspaper articles and tweets were collected in the period from 11th September 2019 to 10th October 2019. This four-week period was chosen since it includes important events related to the movement, such as the worldwide climate strike on 20th September organized by ‘Fridays for Future’. Additionally, data was collected in a second period of investigation from 6th to 26th November 2019 2, but the following explanations and numbers in the method section refer to the first and main investigation period.
2
Due to time constraints, no content analysis of news values could be performed for the second sample. Therefore, the second research period can only serve as material for the investigation of RQ1.
Newspaper.
Articles from eight well established German quality newspapers and magazines with a significant reach were selected to examine the above-mentioned research questions. The selection was based on three criteria: (1) the newspapers had to be a nation-wide outlet with a broad audience, (2) the online editions of the newspapers had to be large enough to cover the ‘Fridays for Future’ movement sufficiently and (3) a political balance had to be maintained. The exact distribution of articles per news outlet can be found in the appendix (Appendix A). The newspaper articles were collected and stored within the Infrastructure for Automated
Content Analysis-tool INCA (Trilling, et al., 2018). To filter for the relevant articles within
the period of analysis, all articles published in one of these outlets that contained the phrase ‘Fridays for Future’ in the text body were retrieved.3 This method concluded a sample of N = 299 articles.
Twitter.
All tweets containing the key phrase ‘Fridays for Future’ were collected via DMI-TCAT, a tool set up to retrieve and collect tweets (for more information see Borra & Rieder, 2014). During the data preparation process, the German tweets were filtered, and all retweets were removed. After checking for duplicates a final sample of 4720 tweets was left. In order to diminish the dataset for the sake of feasibility, a subset was constructed with every second tweet sent, leaving a coding sample of 2360.
Content Analysis
In order to find out the differences between both types of media in terms of the use of news values (RQ2), a manual quantitative content analysis of articles and tweets was
3 Searches with other keywords (e.g. ‘Klimastreik’) did not result in a significant increase of data and
conducted. During the coding process, irrelevant newspaper articles were removed that did not make an explicit, meaningful reference to the ‘Fridays for Future’ movement in the title or teaser of the article. Furthermore, all duplicated tweets that were not recognized before were filtered out manually. After coding, the final sample consisted of 1794 tweets and 161 newspaper articles.4
News Values.
The paper investigates how the presence of eleven news factors – (1) controversy, (2) continuity, (3) emotions, (4) facticity, (5) geographical proximity, (6) influence, (7)
personification, (8) prominence, (9) reach, (10) negative valence and (11) positive valence - differs between newspapers and tweets. Each one of the variables was coded binary as 1 for present and 0 for absent. The list and coding instructions for news values were developed on a theory-driven approach and then revised based on the empirical material. Eilders’ (2006) news value list and its further development by Araujo and van der Meer (2018) served as a starting point. The codebook with further information regarding the coding process can be found in the appendix (Appendix B).
Intercoder Reliability Test.
After developing and testing the codebook, intercoder reliability was assessed through a reliability sample of 40 newspaper articles (5 per outlet) and 100 randomly selected tweets that were coded by two coders. Two indices were used to calculate reliability scores: percent agreement and Krippendorff’s Alpha (see Table 9 in Appendix B). Krippendorff’s alpha level of at least .67 was regarded as satisfactory (Riffe, Lacy, & Fico, 2013). All variables provided satisfying results on intercoder reliability, with the lowest score on the news value ‘influence’ (.77) and the highest on ‘geographical proximity’ (.95).
4
Construction of the Dataset
After aggregating the daily number of publications, a time-series on each type of media was created, presenting the coverage during the four weeks of the first investigation period. The amount of publications per day measured how high the topic was on the media’s agenda. The Python module ‘pandas’ was used to aggregate and merge the online newspaper data with Twitter data (McKinney, 2010).
Data Analysis
Time-Series Analysis.
In the field of communication science, time-series analysis has mostly been used as a method to make causal inferences about political or intermedia agenda-setting. In an effort to understand the attention dynamics between Twitter and newspapers (RQ1), the Vector Autoregression (VAR) model was chosen for this study since the direction and delay of effects can not be deduced from theory. Both newspaper and Twitter variables are considered as dependent and independent variables at the same time in the estimation model. Linear relationships were developed that regressed each of the two aggregated media agenda levels based on both their own past and the past agenda levels of the other media (Meraz, 2011). This is in line with the idea that, on the one hand, news are news because they were news yesterday (Hollanders & Vliegenthart, 2008) and, on the other hand, that news from one media are influenced by other media’s agenda. Inferences about the direction of intermedia effects can be made by using Ordinary Least Squares (OLS) regressions to estimate Granger causality. Granger causality is a useful methodological approach to answer the question of agenda-setting while permitting mutual and reciprocal causality. This analysis followed the procedure recommended by Vliegenthart (2014) and the tutorial by Strycharz, Strauss and Trilling (2018). The Python package ‘statsmodels’ was used to conduct time-series analysis (McKinney, Perktold, & Seabold, 2011).
Results
The Dynamics of Intermedia Agenda-Setting
In this study, the dynamics of intermedia agenda-setting between Twitter and newspapers (RQ1) were investigated over a period of 30 days. The time-series can be interpreted as the saliency of ‘Fridays for Future’ in two different media. Figure 1 shows the number of daily posted tweets5 and published newspaper articles about ‘Fridays for Future’ from 11th September 2019 to 9th October 2019. The left y-axis shows the number of
newspaper articles and the right one the number of tweets. The time-series depicts a clear peak around the 20th September, when both tweets and newspaper articles reached their highest level. This spike is probably caused by an exogenous factor. On 20th September ‘Fridays for Future’ organized and called for a global climate strike. Millions of people worldwide followed this call and participated in demonstrations. Furthermore, on 21st September the UN Youth Climate Summit took place in New York, as part of a weekend of events leading up to the UN Secretary-General’s Climate Actions Summit on Monday, 23rd September. This eventful weekend is reflected in the greatly increased attention for the issues surrounding ‘Fridays for Future’. A control variable for a subperiod of three days (19th -21st September), the ‘worldwide climate strike’ dummy, is included in the model. Apart from the spike in the time-series, there are only minor increases and decreases in coverage. Due to the short period of investigation, no global trend or seasonality were observed.
5 Since the used dataset only contains every second tweet, the results and conclusions only refer to half
of all tweets sent on the topic. For reasons of clarity and accuracy, it was decided not to double the values for tweets but point out and accept this limitation. This restriction applies to all the following explanations of the first investigation period.
Figure 1. First dataset: Twitter and newspaper agenda about ‘Fridays for Future’ over time
Source: Own production.
VAR Analysis.
As a first step in the VAR analysis, the series were tested for stationarity. Since the Dickey-Fuller-Test failed, all series were differenced to achieve stationarity. The second step was to specify the number of lags. Theoretically, there are at most a couple of days until the influence in agenda-setting from one media to the other can be expected, especially with social media working at a higher pace. Looking at the fit-statistics, both AIC and BIC suggest that two days represent an optimal lag for the analysis.
Then, a VAR model with the dummy control variable ‘worldwide climate strike’ is estimated, resulting in the following regression summary (Table 1). Before interpreting the regression summary, several tests were performed to check the robustness of the VAR model. The Portmanteau (Q) test was used to check for autocorrelation in the residuals and
heteroscedasticity in the squared residuals. The results indicate a well-specified and robust model.
The upper part of Table 1 shows that the number of tweets is estimated using the control variable and both its own past agenda levels (L1/L2 Twitter) and the newspapers’ past agenda levels (L1/L2 Newspaper) as predictors. The Twitter agenda from two days ago (L2
Twitter) is a significant predictor, B = -1.19, t = -4.76, p <.001. The predictor L2 Newspaper
is significant as well, B = 13.47, t = 3.59, p < .001, indicating that one more newspaper article two days ago leads to an average increase of 13.47 tweets. The coefficients show that Twitter agenda levels from one and two days ago (L1 & L2 Twitter) and the newspaper agenda levels from one day ago (L1 Newspaper) negatively influence the Twitter agenda level.
The lower part of Table 1 displays the regression summary for predicting the number of newspaper articles. The newspaper agenda level from one day ago (L1 Newspaper) turns out to be a significant predictor, B = -0.80, t = -2.82, p = .005, as well as the Twitter agenda level from two days ago (L2 Twitter), B = -0.06, t = -2.88, p < .005.
Table 1. First dataset: Regression summary for predicting number of tweets and number of newspaper articles with one exogenous variable and two previous time lags
B SE B t p
Climate Strike Dummy 74.98 40.73 1.84 .066
L1 Twitter -0.05 0.25 -0.18 .856
L1 Newspaper -3.98 3.46 -1.15 .251
L2 Twitter -1.19 0.25 -4.76 .000
L2 Newspaper 13.47 3.75 3.59 .000
Newspaper
Climate Strike Dummy -1.10 3.32 -0.33 .741
L1 Twitter 0.03 0.02 1.33 .183
L1 Newspaper -0.80 0.28 -2.82 .005
L2 Twitter -0.06 0.02 -2.88 .004
L2 Newspaper 0.53 0.31 1.74 .082
Source: Own production.
But these coefficients should be interpreted cautiously due to the high levels of collinearity between the multiple lags of a variable. To obtain a good picture of the nature of relationships between the endogenous variables, Granger causality is estimated. As shown in Table 2, the Twitter agenda level Granger-causes the Newspaper agenda level (F = 5.742, p <
.01). This means that the prediction of newspapers’ agenda level can be improved by considering the Twitter agenda level. The other way around, the newspapers’ agenda level also Granger-causes the Twitter agenda level (F = 0.961, p < .001). It seems that agenda-setting works in both ways, from traditional to new media and vice versa.
Table 2. First dataset: Granger causality tests and decomposition of forecast error variance (FEVD)
Newspaper Twitter Newspaper Granger FEVD 10.961*** 4.02% Twitter Granger FEVD 5.742** 47.4% Portmanteau Q (20) 17.74 18.80 Squared Portmanteau Q (20) 21.39 24.79 AIC 10.862 BIC 11.345 ** p < .01, *** p < .001 Source: Own production.
The decomposition of error variance forecast (FEVD) gives additional insight in the overall size of the effects: It can be concluded that 4.02 percent of variation in Twitter attention can be explained by newspapers’ agenda level. On the other hand, 47.4 percent of variation in newspaper coverage can be explained by Twitter’s past agenda level. The low FEVD value (4.02 percent) indicates that the effect from newspaper agenda on Twitter agenda is not substantial. Due to the large difference in FEVD values, a unidirectional effect direction can be assumed.
Scrutinizing the impulse response function helps to visualize the influences that were found to be significant by the Granger causality test. Figure 2 shows how a one-unit increase in the impulse variable at time t = 0 leads to an over-time change in the response variable. One additional article published results in a significant increase in change of 12.48 on the Twitter agenda level (Figure 2a). The other way around, one additional tweet results in a decrease in change of 0.05 on the newspaper agenda (Figure 2b). Since the model is based on differenced series, the impulse response does not indicate absolute numbers and is therefore
difficult to interpret. However, it shows whether there is a change and whether it is positive or negative.
Figure 2. Impulse response analysis: a) impulse newspaper, response Twitter; b) impulse Twitter, response newspaper
Source: Own production
Finally, there is a high contemporaneous correlation (0.77) of the residuals from the VAR analysis. Based on the previous tests, it can be ruled out that the high value is due to a trend, seasonality or autocorrelation. Therefore, it might indicate that traditional and new media agenda levels are affected by the same external factors, such as events, statements and strikes that trigger similar changes in the series. To not overshoot the scope of this paper, the model only includes one exogenous variable to control for an external event, but there might be more influences that remain uncontrolled for. Furthermore, it could be possible that
intermedia agenda-setting works even faster and takes place within minutes and hours instead of days (Vliegenthart, 2014). Future research is invited to further investigate these limitations.
In order to verify the results, a second time-series analysis was carried out with a data sample scraped during a second time period, from 6th November to 26th November 2019. The second plot (Figure 3) also shows parallel dynamics in the agenda levels of both types of media but differs substantially from the graph of the first period (Figure 1). This time, the plot
does not show a peak, but a more constant time-series with regular, smaller ups and downs. Therefore, no control variable was included in the model.
Figure 3. Second dataset: Twitter and newspaper agenda about ‘Fridays for Future’ over time
Source: Own production.
The Dickey-Fuller Test showed that both series of the second investigation period are stationary. Thus, the series are not differenced for the VAR analysis. In the second step of selecting lags, the fit-statistics suggested no lags. Nevertheless, a model with one time lag can be theoretically justified, but the results must be interpreted with caution. Table 3 displays the regression summary of the estimated VAR model. The upper part shows that only the
intercept is a significant predictor of the number of tweets, B = 44.27, t = 2.43, p < .05. In order to predict the number of newspaper articles (see lower part of Table 3) the intercept, B = 5.04, t =3.69, p < .001, as well as the past Twitter agenda level (L1 Twitter), B = -0.04, t=-2.04, p < .05, have proven to be significant predictors.
Table 3. Second dataset: Regression summary for predicting number of tweets and number of newspaper articles with one previous time lag
B SE B t p Twitter Constant 44.27 18.19 2.43 .015 L1 Twitter 0.29 0.27 1.06 .288 L1 Newspaper -1.51 3.31 -0.46 .649 Newspaper Constant 5.04 1.36 3.69 .000 L1 Twitter -0.04 0.02 -2.04 .041 L1 Newspaper 0.5 0.25 0.59 .055
Source: Own production.
The further analysis shows that the Twitter’s agenda level Granger-causes the newspapers’ agenda level but not the other way around (Table 4). Thus, Twitter’s agenda level influences newspapers’ agenda level unidirectionally and the agenda-setting dynamic moves from social media to traditional media. The forecast error variance decomposition (FEVD) indicates that 23.4 percent of variation in newspaper agenda level on ‘Fridays for Future’ within the second sample can be explained by the Twitter agenda level. These findings are in line with the results of the first investigation period.
Table 4. Second dataset: Granger causality tests and forecast error variance decomposition (FEVD)
Newspaper Twitter Newspaper Granger FEVD 0.208 1,2% Twitter Granger FEVD 4.163* 23.4% Ljung-Box Q (20) 19.86 14.97 Lagrange-M test (20) 8.20 11.47 AIC 7.848 BIC 8.147 * p = .05, ** p < .01, *** p < .001 Source: Own production.
To answer RQ1, evidence from time-series analysis clearly runs from Twitter’s agenda level to newspapers’ agenda level. In the first sample, Granger causality and FEVD
established a strong social to traditional media direction, with nearly half of variation in the newspaper agenda level (47.7 percent) being explained by Twitter’s agenda level. While the traditional-to-social agenda-setting effect was significant but not substantial in the first
investigation, the second period did not indicate a significant effect. In summary, only a unidirectional effect can be determined with social media being identified as the dominant agenda-setter in this case of protest coverage.
The Use of News Values on Twitter and in Newspapers
RQ2 aimed to find the relationship between the use of news values and the type of media. Therefore, a Chi-Square test of independence was conducted to determine if there is a significant relationship between two categorical variables. The null hypothesis for this test states that there is no relationship between news values and type of media. Table 5 provides the percentage of newspaper articles and tweets in which a news factor was present and the respective Chi-Square value. The results reveal a significant relationship between both groups in ten out of eleven news values tested. In nine of these ten cases, the news values can be found significantly more often in newspaper articles than in tweets. For example, more newspaper articles (46.6 percent) than tweets (9.9 percent) contain the news value ‘personification’ (X²=177.390, df=1, p<.01).
Table 5. News values by type of media, in percent
*p < 0.05; **p <0.01
Source: Own production.
Newspaper Twitter X² Controversy 32.9 (N = 53) 17.3 (N = 311) 23.678, df = 1** Continuity 13.0 (N = 21) 6.0 (N = 108) 11.825, df = 1* Emotions 18.6 (N = 30) 6.4 (N = 114) 32.647, df = 1** Facticity 83.9 (N = 135) 40.1 (N = 720) 114.747, df = 1 ** Geographical Proximity 37.9 (N = 61) 26.2 (N = 470) 10.205, df = 1* Influence 18.0 (N = 29) 4.2 (N = 75) 56.120, df = 1** Personification 46.6 (N = 75) 9.9 (N = 177) 177.390, df = 1** Prominence 47.8 (N = 77) 14.9 (N = 268) 109.954, df = 1** Reach 18.6 (N = 30) 9.2 (N = 165) 14.650, df = 1** Negative Valence 9.9 (N = 16) 24.5 (N = 440) 17.581, df = 1** Positive Valence 29.8 (N = 48) 33.0 (N = 592) 0.681, df = 1
A time-series analysis was conducted for the news value ‘personification’, because large differences in the distribution were found and the respective Ns were large enough for both media. A plot over time and regression summaries can be found in the appendix (Appendix C). VAR analysis did not find a Granger causality between the news value ‘personification’ on Twitter and in newspaper articles. The time-series of ‘personification’ shows peaks in the traditional medium that do not correspond to the appearance of that news value on Twitter, e.g. in the run-up to the UN climate summit. This indicates that reporting ahead of an event might work differently in traditional media than on social media.
Figure 4. Presence of all news values by type of media
Source: Own production.
Looking at the news values ‘negative valence’ and ‘positive valence’ an interesting picture emerges (Figure 4). ‘Negative valence’ is the only news value that is significantly more likely (X²=17.581, df=1, p<.01) to be present in tweets (33.0 percent) than in newspaper articles (9.9 percent). On the other hand, ‘positive valence’ is the only news value without a
32,9% 13,0% 18,6% 83,9% 37,9% 18,0% 46,6% 47,8% 18,6% 9,9% 29,8% 17,3% 6,0% 6,4% 40,1% 26,2% 4,2% 9,9% 14,9% 9,2% 24,5% 33,0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Newspaper Twitter
significant result on the Chi-Square test (see Table 5), as the distribution in newspaper articles (29.8 percent) is not significantly different from tweets (33.0 percent).
However, a closer look at the most important news values per medium reveals that ‘positivity’ and ‘negativity’6 play a much more important role on Twitter than in newspapers. While the four most commonly used news values by journalists (RQ2a) are ‘facticity’ (83.9 percent), ‘prominence’ (47.8 percent), ‘personification’ (46.6 percent) and ‘geographical proximity’ (37.9 percent), Twitter users (RQ2b) are most likely to use ‘facticity’ (40.1 percent), ‘positive valence’ (33.0 percent), ‘geographical proximity’ (26.2 percent) and ‘negative valence’ (24.5 percent). The high relevance of the news factors ‘facticity’ and ‘geographical proximity’ in both newspapers and tweets emphasizes that news are, first and foremost, influenced by real world events. The frequent use of ‘positivity’ and ‘negativity’ in tweets suggests that social media are frequently used to express personal opinions. Our results indicate that negative reporting about ‘Fridays for Future’ is rare in traditional media.
Contrary to H1, the news values ‘controversy’ and ‘emotionality’ do not play a big role in the examined tweets. More newspaper articles (18.6 percent) than tweets (6.4 percent) contain the news value ‘emotionality’ (X²=32.647, df=1, p<.01). Furthermore, the news value ‘controversy’ is more likely to be found in newspaper articles (32.9 percent) than in tweets (17.3 percent) (X²=23.678, df=1, p<.01).
Thus, the results do not lead to a definite answer to H1. Contrary to H1, ‘emotionality’ and ‘controversy’ are more prominent in newspaper articles than in tweets. While, in line with H1, ‘negative valence’ is more present in tweets than in newspaper articles, ‘positivity’ is not used significantly different between the two types of media.
6
‘Positive valence‘ and ‘positivity’ as well as ‘negative valence’ and ‘negativity’ are used interchangeably.
Discussion
This study examines the differences in news values between Twitter and major
German newspapers concerning the topic ‘Fridays for Future’ (RQ2). The partial rejection of H1 requires some attempts at explanation. The high use of the news value ‘positivity’ in newspaper articles, which led to insignificant results between both types of media, is surprising as journalists are obliged to be objective. Reasons may lie in the research design (no distinction between different types of articles, e.g. comments) and the operationalization of the variable. A possible explanation for the high amount of ‘negativity’ in tweets may be, that unlike professional journalists, Twitter users publish their subjective opinions and actively use Twitter as an alternative platform to oppose the positively dominated traditional media coverage. Unfortunately, the sample size is too small to investigate this effect using a time-series analysis. The low use of the news values ‘controversy’ and ‘emotionality’ may be attributed to the fact that only the text of the tweet and not the link to a further text or video were considered as coding material. These news values may not be revealed in the tweet itself but can be found in the content attached to it. Nevertheless, the low level of ‘emotionality’, ‘controversy’, ‘prominence’ and ‘personalization’, which are frequently criticized factors of traditional media, may at least partly indicate that media logic of dramatization has been suspended.
The results of the time-series analyses (RQ1) indicate a unidirectional relationship, where Twitter significantly influences the agenda level of newspapers. The influence from social media to traditional media was found in the analyses of both samples and proved to be quite strong as measured by FEDV. As Neuman et al. (2014) previously stated, this finding of reversed agenda-setting contradicts a large part of the established literature that for a long time propagated the one-way media agenda-setting from traditional media. Reasons could be that Twitter users respond more immediately to events, since they don’t follow a journalistic
routine. Also given the subject of a civil movement, it is reasonable that social media presents the decisive agenda-setter. ‘Fridays for Future’ is a movement that extensively uses social media for mobilization and is used specifically to research the dynamics within an issue driven, bottom-up movement. Since the study focused on that particular topic, results cannot be generalized to all Twitter usage and no statements can be made about the extent to which social media contributes to put different topics on the agenda.
Overall, it appears that traditional and social media respond similarly to external factors. Figure 1 indicates a parallel dynamic responsiveness of Twitter users and professional journalists to events and public statements. In accordance with this logic, speaking of Granger causality can be misleading since temporal precedence does not necessarily signify a causal link. It is not certain that one medium ‘causes’ the other to set a topic on the agenda.
Sometimes it may be the case that journalists jump on a topic that has emerged and became big on social media, but it appears that mostly both media types share the same perception whether an event or statement is significant and meaningful (Neuman et al., 2014). The uniform course of both agendas seen in the plots can also be explained by the nature of the media chosen for this study. Twitter and the online representation of newspapers are relatively fast pace publishing engines, as they have no fixed schedule. Contrarily, television news and print newspaper have fewer occasions to publish and take more time to cover news (Sevenans & Van Aelst, 2017). Although traditional media might not be the most dominant agenda-setter anymore due to their speed, their impact can still be large. Even though, tweets by e.g. US President Trump are seen by millions of people and can keep the media busy for days, most tweets are seen by far fewer people than the cover page of a major newspaper or a report on the evening news.
Limitations and Implications
Despite all the contributions, the study has limitations, which in turn raise new questions for future studies to address. The insightful examination of news values
unfortunately was not able to include all relevant aspects of the news creating process. Due to the study design that treated all news items (article or tweet) equally, some important
distinctions could not be included. Decisive characteristics such as type, length and position of articles remain unnoticed as well as the differences between traditional news content (such as a newspaper article) shared online and native online content that requires more effort and thoughts. Furthermore, with the coding scheme at hand, no differentiation was made between tweets, which are intended to only inform or those sent to express one’s opinion. Especially concerning this topic of intense public debate, ideologies and agitation play an important role. The amount of links referring to alternative, right-wing media was conspicuous during the coding analysis. More in-depth coding of tweet content characteristics, including pictures and videos linked in the tweet, could help to understand what kind of information Twitter users disseminate and to uncover ways of opinion making on social media.
For the time-series analysis, the operationalization of time represents one of the greatest limiting factors. Firstly, the investigation period was too short for long-term developments. Secondly, a 24-hour day as the temporal unit of analysis dismisses all
dynamics that take place in a matter of minutes or only several hours. Future studies need to make greater use of the available digital data and customize the methodology to the changing conditions, e.g. time frames, of the hybrid media system.
The general approach of this study to refer to the aggregated data of a platform as ‘the’ agenda level of the respective medium has its downfalls. While it might be appropriate for traditional media to treat its content as a single product of one specific actor group (journalists and editors), this equation is hardly applicable to social media (Sevenans & Van Aelst, 2017).
On Twitter a multitude of actors like citizens, journalists and politicians interact and share all different kind of information, with some users being more influential than others. However, the method used in this study could not disentangle different Twitter publics. The Twitter public for instance is strongly dominated by journalists and politicians who use the platform to disseminate information and content (Sevenans & Van Aelst, 2017). Therefore, no inferences can be made from this study to social media in general. Other social media platforms like Facebook for example have different dynamics in terms of content production and flow.
This study does not answer the question whether the observed Granger causality from Twitter’s agenda to newspapers’ agenda is caused by the citizens’ Twitter usage or by
journalistic publications of content on Twitter. To consider Twitter as a platform of citizen journalism, future research should distinguish between different users.
Given not only the high user rate of Twitter but also the increasing importance and sometimes far-reaching consequences of tweets, it is important to continue this endeavour to better understand a fast-changing digital public sphere and its implications on the broader news agenda. Since Granger causality permits mutual and reciprocal causality, as it was found in the first analysis, it is an adequate and promising method to test and refine mass
communication theory. Future pathways for intermedia agenda-setting in the digital era should concentrate on finding the conditions that moderate agenda-setting effects (e.g. type of issue, type of news, timing).
While this study has elaborated on mechanisms of intermedia agenda-setting, future research should examine the dynamics of public agenda-setting that may have influenced the public perception of ‘Fridays for Future’. In that context, it would be interesting to examine the long-term issue developments to explore potential shifts in public discourse over time.
Conclusion
This study explored the applicability of the news value theory to a hybrid media environment. The results show that tweets are less characterized by news values than
newspaper articles. Since Twitter users make much less use of news values, one can conclude that their use is mainly an attribute of professional news treatment. This became particularly clear in the news value ‘personification’. While the newspapers spin most stories about the movement's leader figure Greta Thunberg, Twitter users distance themselves from this, by addressing the movement as a whole. Since the most prominent news values were the same in both media, tweets are not subject to fundamentally different presentation criteria than
journalistic work. Nevertheless, Twitter deviates in some cases from the often criticized media logic. The findings suggest, that even though Twitter users experience the news
treatment and use of news values by the media in their role as news consumers, they have not internalized it (due to a lack of socialization and training) or actively reject it as producers of news. The underlying mechanisms cannot be revealed but should be subject for further qualitative analyses. The findings show that it is necessary to further refine the application of the news value theory to social media and elaborate more on which news values are the product of professional news treatment and which can serve as general relevance indicators.
This study brought new evidence to the question whether traditional or social media play the role of the dominant agenda-setter. Due to the prominence and strength of the ‘reversed’ agenda-setting effect, from social to traditional media, the study confirms the results by Neuman et al. (2014) and Araujo and van der Meer (2018). The research has shown that the pattern of one-way agenda-setting by traditional media must finally belong to the past. Twitter has become a mainstream news information distributor just as traditional media, which have lost the privilege of being the undisputed agenda-setter as they find themselves in an ever-evolving hybrid media sphere. Our study extends prior research by pointing out that
media coverage about a protest movement is dominated by social media. In general, the flow of agenda has been more straightforward and the effect stronger than expected. These findings encourage to further investigate whether the era of the hybrid media system has made social media the new dominant agenda-setter. If further studies confirm these results, the influence of social media on traditional media should no longer be called ‘reversed’ agenda-setting. This would open a new chapter of agenda-setting theory.
References
Araujo, T., & van der Meer, T. (2018). News values on social media: Exploring what drives peaks in user activity about organizations on Twitter. Journalism, pp. 1-19.
Asur, S., Huberman, B., Szabo, G., & Wang, C. (2011). Trends in social media: Persistence
and decay. Barcelona, Spain: Paper presented at the Fifth International AAAI
Conference on Weblogs and Social Media.
Berger, J., & Milkman, K. (2012). What Makes Online Content Viral. Journal of Marketing
Research, 49 (2), pp. 192-205.
Borra, E., & Rieder, B. (2014). Programmed method. Developing a toolset for capturing and analyzing tweets. Aslib Journal of Information Management, 66 (3), pp. 262-278.
Boukes, M. (2019). Social network sites and acquiring current affairs knowledge: The impact of Twitter and Facebook usage on learning about the news. Journal of Information
Technology & Politics, 16 (1), pp. 36-51.
Boukes, M., & Vliegenthart, R. (2017). A general pattern in the construction of economic newsworthiness? Analyzing news factors in popular, quality, regional, and financial newspapers. Journalism, pp. 1-22.
Bucher, T., & Helmond, A. (2017). The Affordances of Social Media Platforms. In J.
Burgess, T. Poell, & A. Marwick, The SAGE Handbook of Social Media. London and New York: SAGE Publications Ltd.
Caple, H., & Bednarek, M. (2013). Delving into the Discourse. Oxford: Reuters Institute for
the Study of Journalism.
Caple, H., & Bednarek, M. (2016). Rethinking news values: What a discursive approach can tell us about the construction of news discourse and news photography. Journalism,
17(4), pp. 435-455.
Chadwick, A. (2011). The Political Information Cycle in a Hybrid News Syytem: the British Prime Minister and the "Bullygate" Affair. The International Journal of Press/Politics
16(1), pp. 3-29.
Djerf-Pierre, M., & Shehata, A. (2017). Still an agenda setter: Traditional news media and public opinion during the transition from low to high choice media environments.
Journal of Communication, 67 (5), pp. 733-757.
Earl, J., McKee-Hurwitz, H., Messinas, A., Tolan, M., & Arlotti, A. (2013). This protest will be tweeted: Twitter and protest policing during the Pittsburgh G20. Information,
Communication & Society, 16(4), pp. 459-478.
Eilders, C. (2006). News factors and news decisions. Theoretical and methodological advances in Germany. Communication, 31, pp. 5-24.
Eilders, C., Geißler, S., Hallermayer, M., Noghero, M., & Schnurr, J.-M. (2010).
Zivilgesellschaftliche Konstruktionen politischer Realität. Eine vergleichene Analyse zu Themen und Nachrichtenfaktoren in politischen Weblogs und professionellem Journalismus. Medien & Kommunikationswissenschaft, 58(1), pp. 63-82.
European Parliament. (2019, July 02). Ergebnisse der Europawahl. Retrieved from https://www.europarl.europa.eu/election-results-2019/de/deutschland/
Galtung, J., & Ruge, M. (1965). The Structure of Foreign News: The Presentation of the Congo, Cuba and Cyprus Crisis in Four Norwegian Newspapers. Journal of
International Peace Research 2, pp. 64-90.
García-Perdomo, V., Salaverría, R., Kilgo, D., & Harlow, S. (2018). To Share or Not to Share. Journalism Studies, 19 (8), pp. 1180-1201.
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), pp. 323-333.
Harcup, T., & O'Neill, D. (2001). What is news? Galtung and Ruge revisited. Journalism
Studies, 2(2), pp. 261-280.
Hollanders, D., & Vliegenthart, R. (2008). Telling what yesterday's news might be tomorrow: Modelling media dynamics. Communications, 33, pp. 47-68.
Jungherr, A., & Jürgens, P. (2014). Through a glass, darkly: Tactical support and symbolic association in Twitter messages commenting on Stuttgart 21. Social Science Computer
Review, 32(1), pp. 74-89.
Kim, H. (2015). Attracting Views and Going Viral: How Message Features and Newssharing Channels Affect Health News Diffusion. Journal of Communication 65, pp. 512-534.
Kwak, H., Lee, C., Park, H., & Moon, S. (2010). What is Twitter, a social network or a news media? Paper presented at the Proceedings of the 19th International Conference on
World Wide Web, (pp. 591-600). Raleigh (NC).
McCombs, M. (2014). Setting the agenda: the mass media and public opinion. Camridge, UK; Malden, MA, USA: Polity Press, Second edition.
McCombs, M., & Shaw, D. (1972). The agenda-setting function of the mass media. Public
Opinion Quarterly, 36, pp. 176-187.
McKinney, W. (2010). Data structures for statistical computing in Python. Proceedings of the 9th Python in science conference, Austin, Texas (Vol. 445, pp. 51-56).
McKinney, W., Perktold, J., & Seabold, S. (2011). Timer Series Analysis in Python with statsmodels. Proceedings of the 10th Python in Science Conference, pp. 96-102.
Meraz, S. (2011). Using Time Series Analysis to Measure Intermedi agenda-setting Influence in Traditional Media and Political Blog Networks. J&MC Quarterly, 88 (1), pp. 176-194.
Neuman, W., Guggenheim, L., Jang, S., & Bae, S. (2014). The Dynamics of Public Attention: Agenda-Setting Theory Meets Big Data. Journal of Communication, 64, pp. 193-214.
Nielsen, R., & Schrøder, K. (2014). The Relative Importance of Social Media for Accessing, Finding, and Engaging with News. Digital Journalism, 2(4), pp. 472-489.
O'Neill, D., & Harcup, T. (2009). News Values and Selectivity. In K. Wahl-Jorgensen, & T. Hanitzsch, The Handbook of Journalism Studies (pp. 161-174). New York: Routledge.
Reese, S., Lou, R., Kideuk, H., & Jaekwan, J. (2007). Mapping the blogosphere: Professional and citizen-based media in the global news arena. Journalism, 8(3), pp. 235-261.
Riffe, D., Lacy, S., & Fico, F. (2013). Analyzing Media Messages. Quantitative content
analysis research (3rd ed.). London: Routledge.
Rogstad, I. (2016). Is Twitter just rehashing? Intermedia agenda setting between Twitter and mainstream media. Journal of Information Technology & Politics, 13(2), pp. 142-158.
Segesten, D., & Bossetta, M. (2016). A Typoligy of Political Participation Online: How Citizens used Twitter to Mobilize during the 2015 British General Elections.
Information, Communication & Society, pp. 1625-1643.
Sevenans, J., & Van Aelst, P. (2017). Intermedia Agenda Setting in the Social Media Age: How Traditional Players Dominate the News Agenda in Election Times. The
International Journal of Press/Politics, pp. 1-19.
Shehata, A., & Strömbäck, J. (2013). Not (yet) a new era of minimal effects: A study of agenda setting at the aggregate and individual levels. The International Journal of
Press/Politics, 18(2), pp. 234-255.
Shoemaker, P., & Cohen, A. (2006). News around the World. Content, Practitioners, and the
Public. New York: Routledge.
Shoemaker, P., & Reese, S. (1991). Mediating the message. White Plains, NY: Longman.
Staab, J. (1990). The role of news factors in news selection: A theoretical consideration.
European Journal of Communication, 5, pp. 423-443.
Strycharz, J., Strauss, N., & Trilling, D. (2018). The role of media coverage in explaining stock market fluctuations: Insights for strategic financial communication.
International Journal of Strategic Communication, 12 (1), pp. 67-85.
Su, Y., & Borah, P. (2019). Who is the agenda setter? Examining the intermedia agenda-setting effect between Twitter and newspapers. Journal of Information Technology &
Politics, pp. 236-249.
Trilling, D., Tolochko, P., & Burscher, B. (2017). From Newsworthiness to Shareworthiness: How to Predict News Sharing Based on Article Characteristics. Journalism & Mass
Trilling, D., Van de Velde, B., Kroon, A., Löcherbach, F., Araujo, T., Strychar, J., &
Jonkman, J. (2018). INCA: Infrastructure for content analysis. Paper presented at the
IEEE eScience 2018, Amsterdam, the Netherlands.
Vliegenthart, R. (2014). Moving up. Applying aggregate level time series analysis in the study of media coverage. Quality & Quantity, pp. 2427-2445.
Waldherr, A. (2018). Modelling issue-attention dynamics in a hybrid media system. In P. Vasterman, From Media Hype to Twitter Storm. News Explosions and Their Impact on
Issues, Crises and Public Opinion (pp. 291-311). Amsterdam: Amsterdam University
Press.
Wanta, W., & Ghanem, S. (2007). Effects of agenda setting. In R. Preiss, B. Gayle, N. Burrell, M. Allen, & J. Bryant, Mass media effects research: Advances through
Appendix
Appendix A: Distribution of newspaper articles in different outlets
Table 6. First dataset: Distribution of newspaper articles per outlet
Newspaper Number
Der Tagesspiegel 47 Die Tageszeitung 42
Die Welt 58
Frankfurter Allgemeine Zeitung 31
Focus 30
Spiegel 55
Zeit 26
Wirtschaftswoche 26
All 299
Source: Own production.
Table 7. Second dataset: Distribution of newspaper articles per outlet
Newspaper Number
Der Tagesspiegel 13 Die Tageszeitung 13
Die Welt 15
Frankfurter Allgemeine Zeitung 0
Focus 4
Spiegel 16
Zeit 7
Wirtschaftswoche 2
All 70
Source: Own production.
Appendix B: Codebook and intercoder reliability test
Table 8. Codebook
News Value Definition Coding Instruction
Controversy Presence of different opinions or indications of disagreement
Disagreement between parties, individuals, groups, organizations, etc. or reporting about an event from two or more sides (1)
Continuity An event that is already defined as a news item has a high chance of continuing to be noticed by the media
Frequency of a topic/event is mentioned (1)
Emotionality Presence of emotions associated with the topic
Display of emotions like joy, anger, fear (1) Facticity Reference to concrete events/ actions/
announcements/ decisions/ protests rather than statements or journalistic explanations of backgrounds, can also be future events
Article/tweet mentions concrete facts and/or deals primarily with the actions of concretely identifiable persons, groups or institutions (1)
Geographical Proximity
Events that take place close to our homes are more likely to have an effect than events that take place far away
Measured by whether the story takes place in Germany (1) or somewhere else (0)
Influence The high significance of a story/an event in terms of its effects and/or consequences
Explicit reference to the economic or societal impact the story has, has had, or may have on individuals, companies, a group or a country (1) Personification Meaning of individuals in an event/
context/ group or highlighting the ‘human face’ of an event
Personal example used to illustrate a trend (mostly Greta Thunberg or other FFF-activists) (1)
Prominence Significance of single prominent people within a certain event
Presence of prominent individuals or individuals of high status (from all sectors) (1)
Reach Reach refers to the number of people affected by an event, level of social relevance
Mentioning of large range of affected persons, social (sub-)groups, companies or all citizens (1) Negative
Valence
Whether the topic is negative toward the organization or the cause is questioned
Reporting about an event opposing the cause of FFF (1)
Positive Valence
Whether the topic is positive toward the organization or the cause is supported
Reporting about an event supporting the cause of FFF (1)
Source: Own production.
Table 9. Intercoder reliability tests, measured in percent agreement and Krippendorff’s alpha
Variable name % KALPHA
Topic Fridays for Future 100 1
Controversy 95.8 .82 Continuity 95.8 .84 Emotions 93.3 .83 Facticity 93.3 .86 Geographical proximity 98.3 .95 Influence 95.0 .77 Personification 97.5 .90 Prominence 95.0 .84 Reach 97.5 .84 Negative Valence 95.8 .89 Positive Valence 95.8 .85
Notes. NArticles= 161, NTweets= 1794. The test was performed by two coders.
Appendix C: Complementary tables for time-series analysis
Figure 5. Personification in tweets and newspaper articles about Fridays for Future over time (during the first investigation period)
Source: Own Production.
Table 10. Regression Summary for Predicting Personification in Tweets and Newspaper Articles with control variable and two previous time lags
B SE B t p Personification Twitter Control 1.18 3.38 0.35 .726 L1 Twitter -0.42 0.25 -1.66 .097 L1 Newspaper -0.28 0.32 -0.88 .378 L2 Twitter -0.32 0.23 -1.38 .166 L2 Newspaper 0.05 0.32 0.17 .864 Personification Newspaper Control -3.75 2.23 -1.68 .092 L1 Twitter 0.11 0.17 0.64 .522 L1 Newspaper -0.65 0.21 -3.13 .002 L2 Twitter 0.03 0.15 0.21 .834 L2 Newspaper -0.48 0.21 -2.30 .022 Source: Own Production.
