Advertisement

Journal of Computational Social Science

, Volume 1, Issue 2, pp 277–294 | Cite as

Who sets the cyber agenda? Intermedia agenda-setting online: the case of Edward Snowden’s NSA revelations

  • Mario Haim
  • Gabriel Weimann
  • Hans-Bernd Brosius
Research Article

Abstract

Since the early introduction of the notion of agenda-setting, researchers have attempted to determine the factors that shape media agendas. One of the key sources of media agenda has been identified as intermedia flow, which various studies revealed in the offline-to-online-to-SNS media context. While most of them focused on the offline-online flow, the present study examines agenda-setting within the new online platforms in various countries, thus allowing for cross-country and cross-media comparisons. We applied time-series analysis to new online media and to traditional online media samples in the context of Edward Snowden’s NSA revelations. Our findings of intermedia agenda-setting effects show a moderate but consistent flow from new online media to traditional online media. This highlights the importance of studying these new directions of agenda flow. Apart from that, no profound agenda-setting patterns can be found elsewhere. Possible reasons and implications are discussed.

Keywords

Agenda-setting Internet Public sphere Comparative research 

Introduction

Like many other communication theories, the agenda-setting theory is being challenged by the dramatic growth of social networking sites platforms. The theory, formulated when traditional media entities owned the tools of item selection, content formation and content distribution, requires more integrative perspectives on traditional media outlets within today’s open and interactive media age [7, 17, 43, 44].

One such feature is the notion of intermedia flow, which refers to the influence of various media outlets’ agendas on one another [23]. Since the early introduction of the notion of agenda-setting, researchers have attempted to determine the factors that shape both the public and the media agenda. One of the key factors of shaping the media agenda has been identified as intermedia agenda-setting: media outlets watch each other carefully and devote attention to issues that have previously received exposure in other outlets (e.g., [9, 22]), which in turn leads to a large degree of consonance among the media. The fact that a certain outlet follows another outlet’s decision to cover a specific event, thereby considering it newsworthy, indirectly validates the first medium’s initial choice. Thus, intermedia agenda-setting is a mechanism that creates a common conception of newsworthiness (e.g., [38]). Another reason for intermedia agenda-setting is the competitive nature of the media market. Media outlets observe their competitors’ decisions and often follow one another’s selection of items once they see it as an advantage or fear that not covering certain items may be disadvantageous. This can also be assumed to apply online—since the media market has grown online, competition has increased as well [42]. Consequently, intermedia agenda-setting online can be expected alike, thereby causing different outlets’ selection and coverage of issues to follow similar patterns.

Evidence of the important role of intermedia agenda-setting has been provided by many studies that examined the process along a wide range of media platforms. Yet, the agenda-setting theory is being challenged by the dramatic growth of new media platforms, by audiences that are simultaneously both media users and producers, by the weakened role of traditional media gatekeepers, by a growing number of news suppliers on social networking sites (SNS), and by the more dynamic, faster, and interactive flow of information and news [5, 6, 10]. These changes have additional impact due to the growth of the online journalists’ community and their extensive use of SNS [12]. Intermedia agenda-setting solely among online media outlets has only scarcely been measured or analyzed (for notable exceptions, see for example [16, 17]). We thus look at online outlets only and ask, who is setting media agendas in the online environment, and how?

Recognizing the online environment’s interactive and bidirectional nature, we examine both traditional media outlets’ websites and SNS. Moreover, due to decreasing impacts of geographical distances and boundaries we also look at agenda-setting effects across multiple countries. For these issues to be addressed, the globally politicizing yet online-savvy topic of Edward Snowden’s revelations on the global NSA espionage affair constitutes an adequate subject of investigation as it also provides long-term data and similar eligibility terms (NSA, Snowden) across multiple languages. We build on a large third-party database of roughly 19.1 million news articles, blog posts, Tweets, and Facebook posts in seven languages from eleven countries over the course of more than 1 year. In this paper, we first look at prior evidence from intermedia agenda-setting research before presenting our study’s case as well as our database. We apply time-series analyses to answer our research questions. Finally, we discuss the findings in light of an emerging global public sphere.

Literature review

Earlier studies that examined intermedia agenda-setting among the offline mass media found strong effects between entities [21, 33], between quality and local newspapers [25], newspaper coverage and television news broadcasts [23, 34], between newspapers and wire services [21] and The New York Times and three American television evening news programs [13]. However, the emergence of online communication platforms, such as blogs, Facebook, or Twitter, has changed the media environment dramatically and, accordingly, the intermedia flow of news. The focus has now moved to the study of intermedia agenda-setting across and between online and offline platforms.

Offline news sources like newspapers, television, or radio, while they remain important, are giving way to emerging online platforms for conveying issues and providing perspectives on social polemics. So-called citizen journalists or “produsers,” as well as bloggers and online activists, are availing themselves of a wide spectrum of online platforms and are reaching huge audiences [5]. Almost half of all news consumers claim their main source of news to be online [30]; three-fourths of news consumers online said they received news through e-mail or SNS, and more than half used those means to share links to news [32]. Consequently, the interplay of traditional (i.e., legacy) and social (i.e., blogs, Facebook, Twitter) media has become a genre in itself [13, 16, 26, 27].

Several studies explored traditional-social intermedia agenda-setting, revealing how the various types of media platforms influence one another’s agendas. On the one hand, Groshek and Groshek [15] examined the intermedia influence of leading traditional online media (the online editions of the New York Times and CNN) and leading SNS, finding event-oriented reciprocal effects while Sweetser et al. [40] found strong correlations between the agendas of news media and campaign blogs during the 2004 election, suggesting a transfer of agendas from the media to the blogs. Similarly, Lee et al. [20] found higher correlations in support of newspaper influence on online bulletin board conversations in South Korea and Cornfield et al. [8] found a correlation of 0.78 for traditional media-to-blog influence as opposed to 0.65 for blog-to-traditional media influence.

On the other hand, Meraz [27] employed time-series analyses to reveal that blogs contributed to setting the agendas of offline elite media. Additional support for this social-to-traditional flow comes from studies on Twitter’s impact on traditional media which can be attributed to Twitter’s fast, easy-to-access, asynchronous and always-on nature, giving Twitter an “ambient” quality that offers “more complex ways of understanding and reporting on the subtleties of public communication” ([18]: 1). For example, a study on the relationship between The New York Times and Twitter found that Twitter was a viable source of what were considered entity-oriented topics with limited coverage in traditional media (Zhao et al. [47]). Similarly, Kwak et al. [19] compared Twitter’s trending topics to CNN headlines and Google trends. Although they found CNN was ahead in reporting more than half the time, they also saw evidence of what can be considered “focusing events” in social media agendas, noting that “some news broke out on Twitter before CNN and they are of live broadcasting nature” ([19]: 6). Wu et al. [46] investigated the influence of micro-blogs on major Chinese media in the immediate aftermath of a catastrophic railway accident. Their results suggest that alternative (i.e., not state-driven) online media played a decisive role in setting mainstream media agendas and providing a citizen forum on a sensitive issue that their conventional counterparts downplayed, ignored, or missed altogether.

More recently, Parmelee [31] studied how political tweets have shaped journalists’ coverage of events, the sources they interview, the quotes they use and the background information they rely on to decide how to cover an issue. Parmelee found that Twitter was more important for journalists than Facebook and other SNS. Some journalists, the findings highlight, actively seek out political tweets to find story ideas. In contrast, Maier [24] found that less than 30% of the top-story topics on blogs and SNS match the top issues on professional media. When their agendas do overlap, blog and SNS content is slightly more extensive, indicating a tendency to diverge from the professional news agenda.

Hypotheses and research questions

Most of the presented studies revealed intermedia agenda-setting in the offline-to-online-to-SNS media context. That is, most of them focused on the offline-online flow, but what about agenda-setting solely online? Classic offline news values such as periodicity (i.e., one decisive issue per day/week) and publicity (i.e., mass audiences) are waning online due to permanently changing websites and increased fragmentation [28]. Hence, news gathering and distribution today may not necessarily follow the classic stimulus–response model of communication. Following both Vonbun et al. [41] and Groshek and Groshek [15], intermedia relationships may be especially complex for events and topics. So who sets the cyber agenda?

The present study examines whether traditional mass media entities remain dominant as agenda setters with their new online platforms (so-called traditional online media) and observes intermedia agenda-setting (1) from traditional online media (TOM) to social network sites (SNS), (2) vice versa, and (3) within SNS (Twitter, Facebook, and blogs). In line with prior findings we expect TOM and SNS to primarily follow the stream of events. Moreover, due to Snowden’s revelations which were primarily offered to journalists, we expect TOM to affect SNS more strongly than vice versa.

Hypothesis 1 (H1)

Same-day correlations between TOM and SNS are stronger than any lagged effects.

H2

Intermedia agenda-setting effects from traditional online media (TOM) to social network sites (SNS) are stronger than vice versa?

In addition, we expect complex bidirectional dependencies within SNS (e.g., [29]), thus refraining from stating directional hypotheses at this point. Lastly, we expect geographical boundaries to be less limiting within the online environment (e.g., [11]). That is, we expect inter-country agenda-setting effects, both for TOM and SNS.

Research Question 1 (RQ1)

What are the intermedia agenda-setting effects within social network sites (SNS)?

RQ2

What are the inter-country agenda-setting effects, if any exists, from one country to another?

Method

Getting observational online data over time is costly. That is, as no centralized archives collect published online content regularly; considerable effort is required to capture the data independently. The database for this study comes from SAP’s Netbase, a commercial provider of archived online content. Netbase regularly scans and archives a wide variety of websites, enabling customers to conduct post-hoc searches for content using single or combined keywords. The search results yield the number of posts/articles per day for a predefined time period and a predefined set of websites (i.e., U.S. blog posts, German Facebook) or URLs (i.e., nytimes.com, theguardian.com). Importantly, most providers offer only aggregated numbers whereas Netbase offers insights into small sub-samples of raw articles (i.e., 100 articles per outlet) to prove the validity of their data. A manual inspection on a random basis proved the aggregation to be plausible.

To answer our research questions regarding intermedia agenda-setting we follow an issue-based approach focusing on one issue and its varying media coverage over time (i.e., [45]). We apply time-series analysis to social network sites (SNS) and traditional online media (TOM) samples with a lag of 1 day. Three reasons justify this lag. First, data availability as well as prior empirical evidence on intermedia agenda-setting between offline and online media suggests this [41]. Second, Edward Snowden’s NSA revelations were initially disseminated among traditional journalists. Our chosen time lag of 1 day thus represents traditional news-publication cycles. Third, 1 day represents a conservative lag, which is most likely not subject to overestimating any effects. That is, due to the event-driven character of issue-based time series [29] and the very fast distribution of online news items, we expect high zero-order correlations between time series at almost any time lag and, consequently, rather small correlations among lagged time series (i.e., H1). Given such data structures, a narrow lag may overestimate the effect due to usually logistic distribution patterns. Larger time lags, however, may underestimate actual effects since they only capture effects on some of their actual occurrences. In other words: If the most appropriate lag was 30 min, time-series analyses with a lag of 10 min might probably overestimate the effect due to non-linear and short-lived diffusion patterns. Similarly, time-series analyses with a lag of 1 day (i.e., 1440 min) would strongly underestimate the effect since they only capture actual agenda-setting every 48th (i.e., 1440 divided by 30 min) time.

The issue selected for this study was Edward Snowden’s revelations on the role of the NSA in global surveillance and espionage. The case of NSA was initiated by whistleblower Edward Snowden who fled from the U.S. to Hong Kong in late May of 2013. After contacting the journalists Laura Poitras and Glenn Greenwald, the revelations reached the public first on June 6th, when the Guardian published initial information on Verizon’s judicial obligation to provide the NSA with meta information on all U.S. citizens’ telephone calls. While the Guardian continued to publish stories on NSA software capable of global surveillance (PRISM, Boundless Informant), Snowden revealed his identity on June 9th through a YouTube video. Global attention increased drastically, again, by the end of July with revelations of another NSA software (XKeyscore), the end of August on the NSA’s espionage on various country leaders, and the end of October on the NSA’s surveillance of German chancellor Merkel’s private cellphone. We chose this topic due to its global political significance and media interest which presumably may be dominating or influencing the various agendas studied here (i.e., a “killer issue,” [4]). Moreover, this story remained “alive” and important with ongoing new revelations, thus allowing for a longitudinal research design. Ultimately, it raised interest across both various media outlets and a large online community.

To monitor the SNS agenda, we selected posts including “NSA” (case-sensitive) in all social media (Blogs, Facebook, Twitter) and all languages (English, French, German, Italian, Japanese, Portuguese, and Spanish) offered by Netbase. We retrieved data for the time period starting roughly 1 month before the revelations (on May 1, 2013) and ending roughly 1 year after the revelations (June 16, 2014) (N = 411 days). The number of posts in each media outlet is aggregated on a daily basis as provided by Netbase. It should be noted that a “day” is defined by the local time zone of the website in question.

The daily data can be depicted as time series. Like any other media data time series the number of posts ranges from zero to a maximum of posts per day with large variation over time. As journalists and other actors normally lose interest in a topic after a few days the time series regularly return to (nearly) zero posts before a new development brings a topic back onto the agenda.

The scan of the archive retrieved 19,131,665 posts on this issue in SNS. Table 1 presents the distribution of the posts in social network sites, classified by language (n = 7) and outlet (Blogs, Facebook, Twitter). For traditional online media (TOM), Netbase provides a more detailed, URL-based access. Due to Netbase’s available languages and owing to availability,1 we selected eleven countries and those two newspapers with the highest circulation in each country. There were three exceptions: (1) for Britain, we included The Guardian, since it was the medium that started revealing Snowden’s leaks; (2) for Switzerland, we included the Neue Zürcher Zeitung, as it is regarded a major quality paper with Europe-wide impact; and (3) for Germany, we selected the top five newspaper outlets to compare our findings to those of previous studies. Ultimately, this left 27 TOM time series from online newspaper outlets in 11 countries. Table 2 gives an overview of the database. The 11 countries can be sorted into the seven languages for SNS data (Table 1). The scan of the archive retrieved 73,172 posts on this issue in TOM.
Table 1

Number of posts in social network sites (SNS)

Language

Outlet

Total posts

Posts per day

n

M

SD

95% CI

English

Blogs

831,883

2019.1

1464.2

1877.3/2160.9

 

Facebook

463,771

1125.7

1965.1

935.3/1316.0

 

Twitter

14,163,475

34,377.4

36,602.7

30,832.5/37,922.2

French

Blogs

41,023

99.6

97.0

90.2/109.0

 

Facebook

5336

13.0

16.3

11.4/14.5

 

Twitter

443,160

1075.6

1580.5

922.6/1228.7

German

Blogs

467,251

1134.1

921.9

1044.8/1223.4

 

Facebook

11,023

26.8

30.2

23.8/29.7

 

Twitter

1,087,944

2640.6

2420.4

2406.2/2875.0

Italian

Blogs

15,892

38.6

38.9

34.8/42.3

 

Facebook

2665

6.5

8.3

5.7/7.3

 

Twitter

75,336

182.9

287.9

155.0/210.7

Japanese

Blogs

15,376

37.3

31.6

34.3/40.4

 

Facebook

288

0.7

1.5

0.6/0.8

 

Twitter

307,656

746.7

704.3

678.5/814.9

Portuguese

Blogs

29,034

70.5

66.0

64.1/76.9

 

Facebook

3355

8.1

18.8

6.3/10.0

 

Twitter

159,968

388.3

432.1

346.4/430.1

Spanish

Blogs

87,149

211.5

187.2

193.4/229.7

 

Facebook

7880

19.1

22.1

17.0/21.3

 

Twitter

912,200

2214.1

2499.7

1972.0/2456.2

Total

21

19,131,665

46,436.1

43,881.0

42,186.4/50,685.8

Table 2

Number of posts in traditional online media (TOM)

Language

Country

Outlet

Total posts

Posts per day

n

M

SD

95% CI

English

UK

Daily Mail

1073

2.6

4.1

2.2/3.0

  

The Guardian

35,382

85.9

101.7

76.0/95.7

  

The Mirror

119

0.3

0.9

0.2/0.4

 

USA

NYT

303

0.7

1.6

0.6/0.9

  

USA Today

2115

5.1

15.9

3.6/6.7

French

France

Le Figaro

675

1.6

2.8

1.4/1.9

  

Le Parisien

1152

2.8

4.1

2.4/3.2

German

Austria

Heute

307

0.7

3.2

0.4/1.1

  

Kronen Zeitung

1642

4.0

4.6

3.5/4.4

 

Germany

BILD

769

1.9

2.6

1.6/2.1

  

FAZ

2709

6.6

7.3

5.9/7.3

  

SpOn

5153

12.5

12.7

11.3/13.7

  

SZ

8264

20.1

26.5

17.5/22.6

  

taz

2218

5.4

5.8

4.8/5.9

 

Switzerland

Blick

1130

2.7

3.9

2.4/3.1

  

NZZ

1184

2.9

3.1

2.6/3.2

  

Tagesanzeiger

1813

4.4

4.9

3.9/4.9

Italian

Italy

Corriere della sera

349

0.8

1.7

0.7/1.0

  

la Repubblica

1037

2.5

5.3

2.0/3.0

Japanese

Japan

Asahi Shimbun

104

0.3

0.7

0.2/0.3

  

Yomiuri Shimbun

96

0.2

0.6

0.2/0.3

Portuguese

Brazil

Folha de S. Paulo

993

2.4

2.8

2.1/2.7

  

O Globo

1420

3.4

4.5

3.0/3.9

 

Portugal

Correio da Manhã

96

0.2

0.6

0.2/0.3

  

Jornal de Notícias

144

0.3

0.8

0.3/0.4

Spanish

Spain

El Mundo

1048

2.5

4.9

2.1/3.0

  

El País

1877

4.6

7.3

3.8/5.3

Total

11

27

73,172

177.6

154.8

162.6/192.6

Analysis

The different time series of the various traditional media within each country were correlated, yielding coefficients ranging from low (r = 0.12 for the traditional US media NYT and USA Today) to high (r = 0.63 for the traditional French media Le Figaro and Le Parisien). The overall average mean for traditional online media time-series correlation is r = 0.40. The three types of SNS (Blogs, Facebook, and Twitter) within each language were correlated, yielding high coefficients that range from r = 0.59 for Japanese SNS to r = 0.83 for English SNS (overall average r = 0.75). This indicates more variety between the traditional online media’s agendas than between the three types of SNS within each country.

To study the intermedia impact we analyzed the effects of each time series on the others. The 48 time series (21 SNS + 27 TOM series), each including 411 time points (i.e., days), were analyzed using cross-lagged correlations and causality checks based on linear regression time-series analysis of the various effects (see Fig. 1 for possible effects). We applied Granger causality tests, which assume that a time series Xt is influenced by Yt if “Yt contains information in past terms that helps in the prediction of Xt and if this information is NOT contained in any other series used in the predictor” ([14]: 430). For example, the intensity of posting on Twitter on a given day is dependent on the intensity of posting of the day before. Statistically speaking and under normal conditions, most of the variance of the time series of the dependent variable (Xt) is explained by its lagged variables (its past values Xt−1, Xt−2, …) yielding R X 2 . A second linear model includes the independent variable. Its lagged variables (Yt−1, Yt−2, …) explain an additional amount of variance (R Y 2 ). The difference between the two amounts of explained variances (R change 2 ) can be tested for significance. A significant contribution of time series Y means that Y causally influences X, if—and only if—the test of the reverse relationship is not significant.
Fig. 1

Explanation of analysis using Granger causalities. The diagram shows possible effects with a 1-day lag (t1 and t2). Zero-order correlations (a) between two time series without lag were calculated separately. Two models including SNS’s auto-regressive influence [Eq. 1; auto (a)] and TOM’s additional influence [Eq. 2; cross (a)] were tested and compared [by R²change in terms of additionally explained variance by cross (a)]. To exclude the possibility of interdependency between the two time series, two additional models were applied to exclude a causal effect of TOM on SNS’s impact

Granger analysis yields four regression models (Eqs. 14) which are described and explained in Fig. 1.
$$ y_{t0} = a1y_{t - 1} + a2y_{t - 2} + \cdots + c $$
(1)
$$ y_{t0} = a1y_{t - 1} + a2y_{t - 2} + \cdots + b1x_{t - 1} + b2x_{t - 2} + c $$
(2)
$$ x_{t0} = a1x_{t - 1} + a2x_{t - 2} + \cdots + c $$
(3)
$$ x_{t0} = a1x_{t - 1} + a2x_{t - 2} + \cdots + b1y_{t - 1} + b2y_{t - 2} + c $$
(4)

Results

According to the assumed speed of dissemination of online news (e.g., [37]), all 48 time series show high zero-order correlations with each other. The average correlation was r = 0.41. This common pattern of coverage and posting across countries and across media is interesting by itself. For our analysis we aggregated the time series by language (for SNS) and by nationality (for TOM). This procedure resulted in seven language-based SNS time series and 11 country-based TOM time series.

Intermedia agenda-setting effects (H1, H2)

To study intermedia agenda-setting effects, various models of effects were tested for pairs of two aggregated time series (SNS1 and TOM1, as presented in Fig. 1). In addition to simple Pearson-based correlations of the time series at t1, we tested two hierarchical regression models for each direction of influence. The dependent variable is always the lagged time series (SNS2, TOM2). The first model calculates the explained variance of the time series’ “past” (auto-regression, auto(a) and auto(b) in Fig. 1). In a second step the intermedia flow is added as an additional independent variable [cross(a) and cross(b)]. This yields results in an additional amount of explained variance compared to the first model. The difference of the explained variance between the two models can be tested for significance. If the intermedia flow adds an additional significant amount of variance, it can be concluded that it causes the dependent variable if, and only if, the reverse regression model yields no significant amount of explained variance.

Tables 3 and 4 show the results of Granger causality tests of the relationship between the traditional online media and SNS agenda. The first column depicts zero-order correlations. For all seven languages there are significant correlations between the SNS and the traditional online media agenda, indicating the event-driven character of the time series. However, the strength of the correlation varies markedly across languages. While in the English online media the correlation is only moderate (r = 0.25), the German online media follow each other almost perfectly (r = 0.87). One can assume that the correlation between traditional online media and SNS is a result of (a) a strong co-orientation of these websites. Even more likely (b) both types of media are influenced by the same sources such as press conferences, revelations, and leaks or statements of politicians and/or organizations. Hypothesis 1 expected time-series correlations to be stronger than any cross-media influences. This can thereby be confirmed.
Table 3

Intermedia effects of traditional online media on SNS

 

Zero-order correlation between SNS and TOM

First model (auto)

Second model (cross)

R total 2

r

β

R 2

β

R 2

English

0.25

0.82***

0.67

0.67

French

0.69

0.62***

0.38

0.39

German

0.87

0.78***

0.60

0.60

Italian

0.51

0.69***

0.48

0.48

Japanese

0.51

0.62***

0.38

0.38

Portuguese

0.65

0.57***

0.32

0.16***

0.01

0.33

Spanish

0.64

0.62***

0.39

0.18***

0.02

0.40

Mean across languages

0.62

0.68

0.47

0.05

0.00

0.47

SD

0.32

0.19

0.18

0.08

0.01

0.17

Mean values are Fisher-Z standardized. The first model calculates the stability of the new media agenda over time (i.e., the auto-correlation of the new media agenda). The second model calculates the additional influence of the past traditional media agenda on the new media agenda. The last column indicates how large the additional amount of explained variance is

*<.05; **<.01; ***<.001

Table 4

Intermedia effects of SNS on traditional online media

 

Zero-order correlation between TOM and SNS

First model (auto)

Second model (cross)

R total 2

R

β

R 2

β

R 2

English

0.25

0.76***

0.58

0.58

French

0.69

0.61***

0.37

0.18***

0.02

0.39

German

0.87

0.75***

0.56

0.46***

0.05

0.62

Italian

0.51

0.27***

0.07

0.27***

0.05

0.12

Japanese

0.51

0.52***

0.27

0.27***

0.05

0.32

Portuguese

0.65

0.65***

0.42

0.42

Spanish

0.64

0.76***

0.57

0.13***

0.01

0.58

Mean across languages

0.62

0.64

0.43

0.19

0.03

0.46

SD

0.32

0.26

0.24

0.17

0.02

0.24

Mean values are Fisher-Z standardized. The first model calculates the stability of the traditional media agenda over time (i.e., the auto-correlation of the traditional media agenda). The second model calculates the additional influence of the past new media agenda on the traditional media agenda. The last column indicates how large the additional amount of explained variance is

*<.05; **<.01; ***<.001

Hypothesis 2 predicted a stronger influence from TOM on SNS than vice versa. Looking at the time pattern of causal influences there are differences across languages: The English TOM and SNS show no effects indicated by a significant lagged influence. For the Spanish case, there is a mutual causal relationship from TOM onto SNS and vice versa. This makes it impossible to conclude any unequivocal causal effect. In the Portuguese case we find a slight causal effect of the traditional online media onto the SNS agenda, but the additional amount of explained variance is 1%. For the remaining four languages SNS seem to influence the TOM agenda with an additional amount of variance ranging from 2 to 5%. This unexpected pattern of results supports notions that the traditional media (be it online or offline) lose at least some of their power to raise and promote issues and to influence the perceived importance of issues by the public (e.g., [5]). Admittedly though, the strength of causal relationships is not very strong. Yet, this is probably based on the speed of online communication. Lags of 1 day are most likely too long to grab the interplay of different online actors. Hypothesis 2, assuming the existence of lagged effects from TOM on SNS, thus has to be rejected. This is also in slight contrast to findings such as Maier’s [24] as our results indicate a similar coverage of events in both TOM and SNS rather than divergence between the two (as Maier shows).

Intermedia agenda-setting effects within SNS (RQ1)

To test for intermedia effects within SNS, we ran analyses for Blogs, Facebook, and Twitter. This yielded 21 pairs of Granger causality regression models (three SNSs × seven languages). The results (Table 5) reveal no clear pattern of influence between the three types of SNS. Across all seven languages, zero-order correlations among the time series of Facebook, Blogs, and Twitter are strong, revealing a similar pattern of posts for all the time series included. The range of correlations vary between r = 0.39 and 0.90. SNS in this regard can be said to follow rather stable agenda patterns for the NSA issue. Lower correlations were found for Japanese and Spanish. These outliers are probably due to one SNS’s higher importance over another. For instance, in Japan Twitter is more relevant in daily communication than Facebook.
Table 5

Intermedia Effects between SNS Time Series

 

Blogs → FB

FB → Blogs

Blogs → Twitter

Twitter → Blogs

FB → Twitter

Twitter → FB

English

0.02

0.02

0.01

0.01

French

German

0.02

0.03

Italian

0.01

0.02

0.01

0.02

0.01

0.02

Japanese

0.02

0.01

0.01

0.01

Portuguese

0.01

0.02

Spanish

0.01

Z std. M

0.00

0.01

0.01

0.01

0.00

0.01

SD

0.01

0.01

0.01

0.01

0.01

0.01

Values represent R change 2 values. M-values are Fisher-Z standardized mean values

For all seven languages, the results of a Granger analysis showed very weak effects of one time series on the others. As seen in Table 5, a maximum of 3% additional variance can be explained by assuming that one SNS causally influences another. In addition to the small amount of explained variance, no clear pattern of causal influence emerges. None of the three platforms can be regarded as an agenda-setter for the other two platforms. If one excludes unequivocal relationships (e.g., the mutual influence of all three platforms in the Italian case) Facebook exerts an influence two times, blogs three times, and Twitter four times. While one could argue that the latter is due to Twitter’s fast-responding character, blogs—contrary to this pattern—seem more influential than Facebook. Research question 1 asking for any intermedia agenda-setting effects within SNS has thus to be neglected. Throughout the case of NSA no clear patterns of such effects could be identified.

Inter-country agenda-setting effects (RQ2)

To analyze inter-country effects we were restricted to those sets of countries that are located within the same time-zone. For example, the United States and England share the same language, but their time zones of New York and London differ by 5 h. An English “day” therefore comprises a different set of hours compared to an American day. In addition, SNS platforms were based on language, not on country. The English Twitter time series, therefore, consists of American and English feeds.

For the remaining cases European2 time series were analyzed. Every country’s SNS and TOM (as independent variable) may influence every other country’s SNS and TOM separately (as dependent variable). For every possible model, first, the auto regression of the dependent variable in question was determined, following again Granger causality with a lag of 1 day. In a second step the within-country effects were calculated (SNS or TOM, respectively). In a third step, the effects of each of the three other countries were calculated (Tables 6, 7). For example, the variance of coverage in traditional French online media (TOM) is explained mostly by its own past (37%), partly by French SNS (2%) and additionally by Spanish TOM (1%) as well as Italian SNS (1%, Table 7).
Table 6

Inter-country agenda-setting relations on social network sites (SNS)

 

R 2 auto (a)

Within-country cross (a) TOM R2

IV

FR

DE

IT

ES

French (FR)

0.38

TOM

 

   

SNS

 

German (DE)

0.60

TOM

 

0.01

   

SNS

 

Italian (IT)

0.48

TOM

0.01

 

0.01

   

SNS

0.03

 

Spanish (ES)

0.39

0.02

TOM

 
   

SNS

0.01

0.02

 

Z std. M

0.47

0.00

 

0.01

0.00

0.00

0.00

SD

0.13

0.01

 

0.01

0.00

0.01

0.00

Values in the table represent explained variances (R2). IV = independent variable; auto = auto-regressive influence (first model in Table 3); within-country TOM R2 = cross influence (second model in Table 3). M values are Fisher-Z standardized mean values

Table 7

Inter-country agenda-setting relations on traditional online media (TOM)

 

R 2 auto (b)

Within-country cross (b) SNS R2

IV

FR

DE

IT

ES

French (FR)

0.37

0.02

TOM

 

0.01

   

SNS

 

0.01

German (DE)

0.56

0.05

TOM

0.01

 

0.01

   

SNS

 

0.01

Italian (IT)

0.07

0.05

TOM

0.02

0.02

 

0.02

   

SNS

0.01

 

Spanish (ES)

0.57

0.01

TOM

0.01

 
   

SNS

0.01

0.01

 

Z std. M

0.41

0.03

 

0.01

0.00

0.00

0.00

SD

0.25

0.02

 

0.01

0.01

0.01

0.01

Values in the table represent explained variances (R2). IV = independent variable; auto = auto-regressive influence (first model in Table 4); within-country SNS R2 = cross influence (second model in Table 4). M values are Fisher-Z standardized mean values

Results show that except for Italian TOM, the time series of both SNS and TOM are best predicted by their own past coverage of the issue. Moreover, all other effects are weak again. The contribution of new media to explain the variance of traditional media does not exceed 5% (for Germany and Italy) while the contribution of traditional media to explain the variance of new media does not exceed 2% (in the case of Spain). Cross-country effects do not exceed 3% (the influence of French new media on Italian new media). Hence, also research question 2 has to be neglected. We simply could not find any intermedia agenda-setting effect patterns across countries.

Discussion

Various conclusions can be drawn from these findings. First, zero-order correlations within all the time series—in all countries and for all languages—are moderate for traditional online media (TOM) and high for social network sites (SNS). That is, all media outlets under investigation have similar selection patterns of the NSA case on their websites. One can conclude that the coverage is mostly event-driven.

Second, our results on intermedia agenda-setting effects show a moderate but consistent flow from SNS to TOM. Although Snowden revealed information exclusively to mainstream journalists, SNS seem to have a stronger effect on TOM than vice versa.

Third, apart from zero-order correlations no profound agenda-setting patterns can be found elsewhere—neither within SNS nor between countries. This is a surprising result. We can only speculate about the reasons for missing agenda-setting effects. One might be that the rapid and dynamic changes of online media are still going on and that our data reflect a transition stage that is subject to further development. Another reason could be the time interval of 1 day which might not be adequate to reveal the rapid co-orientation of different media in the online world. To completely mask intermedia agenda-setting effects in our study, global co-orientation would need to be entirely completed within 24 h.

The findings also re-emphasize the importance of events termed by Brosius and Kepplinger [4] as “killer issues.” Such killer issues depict certain topics of global importance which clearly shape the global news agenda and thus the global arena from which (perceived) public opinion arises. Rusciano and Fiske-Rusciano ([36]: 320) describe this global formation of public opinion as “world opinion”, that “refers to the moral judgments of observers which actors must heed in the international arena, or risk isolation as a nation.” That is, for globally dominating and polarizing topics, such as Edward Snowden’s NSA revelations, nations might risk partial isolation if their national media agenda does not include any such references. While this concept strongly adheres to Noelle-Neumann’s concept of the Spiral of Silence, Rusciano [35] argues that social media might both foster and complicate the formation of a world public opinion, depending on the intention and influence of powerful nations. While such assumptions remain vague, our findings are in line with prior empirical evidence of SNS affecting TOM more strongly than vice versa (e.g., [19, 27, 29]).

In the case of NSA, the nations’ power appears weakened, which in turn would promote the agenda-setting power of SNS. Four explanations seem reasonable for this pattern. First, SNS might have influenced TOM more strongly due to the bottom-up nature of whistleblowing. Second, media attention in the U.S., the scandal’s originating country, was low. This partial absence of a powerful nation’s voice might have lowered other nations’ voices similarly. Third, the tech-savvy topic of online espionage and surveillance might be more appealing to (Internet) users than to traditional outlets. Fourth, the current findings could also be a strong indicator of media outlets’ audience orientation. That is, if the captured SNS posts largely depicted media outlets’ shared articles, this would render the findings in slightly new lights. Yet, given our inclusion of blogs, we do not consider this to be the sole determining factor.

The Internet and its numerous online platforms have become a global arena for the formation, distribution, and reception of world news, opinions, and agendas. Thus, one would have expected the emergence of a global news system serving as the public sphere. As a consequence, journalism is becoming more oriented towards news of global importance from a global perspective. Issues of global importance, such as the NSA case, might reveal different agenda-setting patterns. Most of the news, however, are still treated from a national or local perspective, while “such clear-cut cases of global journalism are rarely observed” ([3]: 854). With the growing popularity and growth of social network sites, these assumptions of a world public sphere gain renewed relevance (e.g., [1]).

In terms of agenda-setting these changes call for a new model of the agenda-setting process. The classic approach follows a stimulus–response-like model where the media are capable of shaping the public’s agenda under certain circumstances. This is probably not the case in today’s fragmented media environment (e.g., [29]). More likely, the above mentioned globalization of public opinion formation gives rise to a more differentiated and flow-oriented model of news. That is, issues and opinions spread differently, with their reach and impact depending on various characteristics, actors, and influences. This diffusion of news already raises significant scientific interest (e.g., [2, 39]). The partial results of the present study and those of other studies highlight the need of both theoretical reformulation of the agenda-setting process as well as new methodological approaches and empirical validation.

The present study focused on intermedia and inter-country agenda-setting effects on a single issue, the NSA scandal, thus blanking out surrounding issues on the media and public agendas. While this is a common limitation of many agenda-setting studies [4], it may also have some advantages as it grants unequivocal empirical access to a global issue without the necessity to take domestic news influences into account. A stronger limitation of the study is related to our database: we had no access to the complete original raw data but only to aggregated sums of articles and posts per day as well as very small samples of raw articles. Consequently, we are not capable of adequately validating the data quality but rather rely on the provider’s (Netbase) claims. While this might be a flaw of scientific principles commercial data providers offer access to what would otherwise be hidden to our field of scientific interest. The present findings support this workaround but nevertheless highlight the necessity for adequate data collections in the online environment.

Finally, our lag of 1 day is both the shortest interval of available data (from Netbase) and very likely a too long time lag to capture intermedia effects within the online environment. In the fast and dynamic flow on new online platforms, very short time lags seem to be more realistic and valid. Assuming, for instance, a lag of 1 h will result in a blurring of effects 23 times a day with only one hourly lag applicable on midnight (i.e., when the lag of 1 day applies). Such database and analysis will require a finer granulated data and huge databases that could be transformed into time series. Time series analysis in these regards seems to be the only applicable analytical approach as it can test various models of causality in time-based data. The present study, with its 1 day lag, is probably revealing only conservative estimates for agenda-setting effects in the online media.

Footnotes

  1. 1.

    The British daily The Sun had to be excluded as it was not accessible through Netbase. Instead, “The Mirror” was chosen.

  2. 2.

    Although being European, England and Portugal had to be excluded because of their common language with the U.S. and Brazil, respectively.

References

  1. 1.
    Althaus, S. L., & Tewksbury, D. (2002). Agenda setting and the “new” news patterns of issue importance among readers of the paper and online versions of the New York Times. Communication Research, 29, 180–207.  https://doi.org/10.1177/0093650202029002004.CrossRefGoogle Scholar
  2. 2.
    Bakshy, E., Rosenn, I., Marlow, C., & Adamic, L. (2012). The role of social networks in information diffusion. Proceedings of the 21st International Conference on World Wide Web (pp. 519–528). New York: ACM.Google Scholar
  3. 3.
    Berglez, P. (2008). What is global journalism? Journalism Studies, 9, 845–858.  https://doi.org/10.1080/14616700802337727.CrossRefGoogle Scholar
  4. 4.
    Brosius, H.-B., & Kepplinger, M. (1995). Killer and victim issues: issue competition in the agenda-setting process of German television. International Journal of Public Opinion Research, 7, 211–231.  https://doi.org/10.1093/ijpor/7.3.211.CrossRefGoogle Scholar
  5. 5.
    Bruns, A. (2005). Gatewatching: Collaborative online news production. New York: Peter Lang.Google Scholar
  6. 6.
    Chadwick, A. (2013). The hybrid media system: Politics and power. Oxford: University Press.CrossRefGoogle Scholar
  7. 7.
    Chaffee, S. H., & Metzger, M. J. (2001). The end of mass communication? Mass Communication and Society, 4, 365–379.  https://doi.org/10.1207/S15327825MCS0404_3.CrossRefGoogle Scholar
  8. 8.
    Cornfield, M., Carson, J., Kalis, A., & Simon, E. (2005). Buzz, blogs, and beyond: The internet and the national discourse in the fall of 2004. Pew Research Center. Retrieved May 25, 2015, from http://www.pewtrusts.org/~/media/legacy/uploadedfiles/wwwpewtrustsorg/news/press_releases/society_and_the_internet/PIPBlogs051605pdf.pdf.
  9. 9.
    Denham, B. E. (2014). Intermedia attribute agenda setting in the New York Times: The case of animal abuse in U.S. horse racing. Journalism & Mass Communication Quarterly, 91, 17–37.  https://doi.org/10.1177/1077699013514415.CrossRefGoogle Scholar
  10. 10.
    Esser, F. (2013). Mediatization as a challenge: Media logic versus political logic. In H. Kriesi, S. Lavenex, F. Esser, J. Matthes, M. Bühlmann, & D. Bochsler (Eds.), Democracy in the age of globalization and mediatization. New York: Palgrave Macmillan.Google Scholar
  11. 11.
    Fletcher, R., & Nielsen, R. K. (2017). Are news audiences increasingly fragmented? A cross-national comparative analysis of cross-platform news audience fragmentation and duplication. Journal of Communication.  https://doi.org/10.1111/jcom.12315.Google Scholar
  12. 12.
    Gleason, S. (2010). Harnessing social media. American Journalism Review, 32, 6–7.Google Scholar
  13. 13.
    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, 323–334.  https://doi.org/10.1080/14616700500533643.CrossRefGoogle Scholar
  14. 14.
    Granger, C. W. J. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica, 37, 424–438.CrossRefGoogle Scholar
  15. 15.
    Groshek, J., & Groshek, M. K. (2013). Agenda trending: Reciprocity and the predictive capacity of social networking sites in intermedia agenda setting across topics over time. Media and Communication, 1, 15–27.  https://doi.org/10.2139/ssrn.2199144.CrossRefGoogle Scholar
  16. 16.
    Guo, L., & Vargo, C. J. (2017). Global intermedia agenda setting: A big data analysis of international news flow. Journal of Communication, 67(4), 499–520.  https://doi.org/10.1111/jcom.12311.CrossRefGoogle Scholar
  17. 17.
    Harder, R. A., 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, 22(3), 275–293.  https://doi.org/10.1177/1940161217704969.CrossRefGoogle Scholar
  18. 18.
    Hermida, A. (2010). Twittering the news: The emergence of ambient journalism. Journalism Practice, 4, 1–12.  https://doi.org/10.1080/17512781003640703.CrossRefGoogle Scholar
  19. 19.
    Kwak, H., Lee, C., Park, H., & Moon, S. (2010). What is Twitter, a social network or news media? In Proceedings of the 19th International Conference on World Wide Web, Raleigh, NC, USA, 26–30 April 2010. Retrieved June 30, 2014, from http://an.kaist.ac.kr/~haewoon/papers/2010-www-twitter.pdf.
  20. 20.
    Lee, B., Lancendorfer, K., & Lee, K. J. (2005). Agenda-setting and the internet: The intermedia influence of internet bulletin boards on newspaper coverage of the 2000 general election in South Korea. Asian Journal of Communication, 15, 57–71.  https://doi.org/10.1080/0129298042000329793.CrossRefGoogle Scholar
  21. 21.
    Lim, J. (2006). A cross-lagged analysis of agenda setting among online news media. Journalism and Mass Communication Quarterly, 83, 298–312.  https://doi.org/10.1177/107769900608300205.CrossRefGoogle Scholar
  22. 22.
    Lim, J. (2011). Intermedia agenda setting and news discourse. Journalism Practice, 5, 227–244.  https://doi.org/10.1080/17512786.2010.509184.CrossRefGoogle Scholar
  23. 23.
    Lopez-Escobar, E., Llamas, J. P., McCombs, M., & Lennon, F. (1998). Two levels of agenda setting among advertising and news in the 1995 Spanish elections. Political Communication, 15, 225–238.  https://doi.org/10.1080/10584609809342367.CrossRefGoogle Scholar
  24. 24.
    Maier, S. (2010). All the news fit to post? Comparing news content on the web to newspapers, television, and radio. Journalism & Mass Communication Quarterly, 87, 548–562.  https://doi.org/10.1177/107769901008700307.CrossRefGoogle Scholar
  25. 25.
    Mathes, R., & Pfetsch, B. (1991). The role of the alternative press in the agenda-building process: Spill-over-effects and media opinion leadership. European Journal of Communication, 6, 33–62.  https://doi.org/10.1177/0267323191006001003.CrossRefGoogle Scholar
  26. 26.
    McCombs, M. E. (2014). Setting the agenda: The mass media and public opinion (2nd ed.). Cambridge: Malden.Google Scholar
  27. 27.
    Meraz, S. (2011). The fight for ‘how to think’: Traditional media, social networks, and issue interpretation. Journalism, 12, 107–127.  https://doi.org/10.1177/1464884910385193.CrossRefGoogle Scholar
  28. 28.
    Napoli, P. M. (2010). Audience evolution: New technologies and the transformation of media audiences. New York: Columbia University Press.Google Scholar
  29. 29.
    Neumann, R. W., Guggenheim, L., Jang, S. M., & Bae, S. Y. (2014). The dynamics of public attention: Agenda-setting theory meets big data. Journal of Communication, 64, 193–214.  https://doi.org/10.1111/jcom.12088.CrossRefGoogle Scholar
  30. 30.
    Newman, N., Fletcher, R., Kalogeropoulos, A., Levy, D. A. L., & Nielsen, R. K. (2017). Digital news report 2017. Oxford. Retrieved June 26, 2017, from https://reutersinstitute.politics.ox.ac.uk/sites/default/files/Digital%20News%20Report%202017%20web_0.pdf. Accessed 26 June 2017
  31. 31.
    Parmelee, J. (2014). The agenda-building function of political tweets. New Media & Society, 16, 434–450.  https://doi.org/10.1177/1461444813487955.CrossRefGoogle Scholar
  32. 32.
    Purcell, K., Rainie, L., Mitchell, A., Rosenstiel, T., & Olmstead, K. (2010). Understanding the participatory news consumer. Pew Internet & American Life Project. Retrieved July 4, 2014, from http://www.pewinternet.org/Reports/2010/Online-News.aspx.
  33. 33.
    Reese, S. D., & Danielian, L. H. (1989). Intermedia influence and the drug issue: Converging on cocaine. In P. J. Shoemaker (Ed.), Communication Campaigns about Drugs: Government, Media and the Public (pp. 47–66). Hillside: Lawrence Erlbaum Associates.Google Scholar
  34. 34.
    Roberts, M., & McCombs, M. (1994). Agenda setting and political advertising: Origins of the news agenda. Political Communication, 11, 249–262.  https://doi.org/10.1080/10584609.1994.9963030.CrossRefGoogle Scholar
  35. 35.
    Rusciano, F. L. (2013). World Public Opinion. In W. Donsbach, C. T. Salmon, & Y. Tsfati (Eds.), The Spiral of silence: New perspectives on communication and public opinion (pp. 179–186). New York: Routledge Chapman & Hall.Google Scholar
  36. 36.
    Rusciano, F. L., & Fiske-Rusciano, R. (1990). Towards a notion of “world opinion”. International Journal of Public Opinion Research, 2, 305–322.  https://doi.org/10.1093/ijpor/2.4.305.CrossRefGoogle Scholar
  37. 37.
    Schäfer, M. S., Ivanova, A., & Schmidt, A. (2014). What drives media attention for climate change? Explaining issue attention in Australian, German and Indian print media from 1996 to 2010. International Communication Gazette, 76, 152–176.  https://doi.org/10.1177/1748048513504169.CrossRefGoogle Scholar
  38. 38.
    Shoemaker, P. J., & Reese, S. D. (2013). Mediating the message in the 21st century: A media sociology perspective (3rd ed.). New York: Routledge.CrossRefGoogle Scholar
  39. 39.
    Susarla, A., Oh, J.-H., & Tan, Y. (2012). Social networks and the diffusion of user-generated content: Evidence from YouTube. Information Systems Research, 23, 23–41.CrossRefGoogle Scholar
  40. 40.
    Sweetser, K., Golan, G., & Wanta, W. (2008). Intermedia agenda setting in television, advertising, and blogs during the 2004 election. Mass Communication and Society, 11, 197–216.  https://doi.org/10.1080/15205430701590267.CrossRefGoogle Scholar
  41. 41.
    Vonbun, R., Königslöw, K. K., & Schoenbach, K. (2016). Intermedia agenda-setting in a multimedia news environment. Journalism, 17(8), 1054–1073.  https://doi.org/10.1177/1464884915595475.CrossRefGoogle Scholar
  42. 42.
    Webster, J. G. (2014). The marketplace of attention: How audiences take shape in a digital age. Cambridge: MIT.Google Scholar
  43. 43.
    Weimann, G., & Brosius, H. (2016). A new agenda for agenda-setting research in the new media era. In G. Vowe & P. Henn (Eds.), Political communication in the online world. Theoretical approaches and research designs (pp. 26–44). New York, NY: Routledge.Google Scholar
  44. 44.
    Weimann, G., Weiss-Blatt, N., Mengistu, G., Tregerman, M. M., & Oren, R. (2014). Reevaluating “The end of mass communication?”. Mass Communication and Society, 17, 803–829.  https://doi.org/10.1080/15205436.2013.851700.CrossRefGoogle Scholar
  45. 45.
    Winter, J. P., Eyal, C. H., & Rogers, A. H. (1982). Issue-specific agenda-setting: The whole as less than the sum of the parts. Canadian Journal of Communication, 8(2), 1–10.CrossRefGoogle Scholar
  46. 46.
    Wu, Y., Atkin, D., Mou, Y., Lin, C. A., & Lau, T. (2013). Agenda setting and Micro-blog use: An analysis of the relationship between Sina Weibo and newspaper agendas in China. The Journal of Social Media in Society, 2(2), 8–25.Google Scholar
  47. 47.
    Zhao, W. X., Jiang J., Weng J., He J., Lim E., Yan H., & Li X. (2011). Comparing Twitter and traditional media using topic models. In Advances in Information Retrieval, Proceedings of the 33rd European Conference on IR Research, ECIR, Dublin, Ireland, April 18–21, 2011. Berlin: Springer.Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  • Mario Haim
    • 1
  • Gabriel Weimann
    • 2
  • Hans-Bernd Brosius
    • 1
  1. 1.Department of Communication Studies and Media ResearchLMU MunichMunichGermany
  2. 2.Department of CommunicationUniversity of HaifaHaifaIsrael

Personalised recommendations