2

new media & society 2016, Vol. 18(9) 1916 –1934

© The Author(s) 2015 Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1461444815616226

nms.sagepub.com

Agenda setting in the twenty-first century

GR Boynton The University of Iowa, USA

Glenn W Richardson Jr Kutztown University of Pennsylvania, USA

Abstract The revelation of the surveillance practices of the US National Security Agency (NSA) and the UK Government Communications Headquarters (GCHQ) serves as a compelling case study of the new agenda-setting capacities in the 21st century. We use a stream of 14 million messages on Twitter to note how its reach equals that of television. Next, we demonstrate the interactive communication characteristics of social media in the use of hashtags, retweets, and shared URLs. Finally, we draw upon data from Lexis-Nexis on newspaper stories and broadcast transcripts to emphasize the different ways social and mainstream media constructed the story about surveillance and interacted with each other. We conclude agenda setting can no longer be understood as a monopoly of the mainstream media, indexed to the actions of political elites. Social media, through its reach, interaction, and broadening of ideas brought into the discussion, emerge as a distinctive mode of large-scale communication. Giving voice to the people introduces an entirely new dimension to agenda setting.

Keywords Administration, agenda setting, mainstream media, national security, Snowden, new media, Twitter

In 1948, Harry Truman traveled from coast to coast by train. At every stop, he spoke to those who had assembled to participate in the public discussion of the politics of the nation. In December 2014, Jeb Bush announced that he was giving serious consideration

Corresponding author: GR Boynton, Department of Political Science, The University of Iowa, Iowa City, IA 52242, USA. Email: [email protected]

616226 NMS0010.1177/1461444815616226new media & societyBoynton and Richardson research-article2015

Article

Boynton and Richardson 1917

to seeking the Republican nomination for the presidency. He did not take the train. He did not speak to assembled news media personnel before TV cameras. He posted his announcement on Facebook. Then in March 2015, Ted Cruz announced by posting a message to Twitter, and he was quickly followed by Hillary Clinton. Those events bracket the television age of politics.

Within a decade of Truman’s “whistle stop tour,” television had become the center of public discussion about politics because of its reach. Whistle stopping across the country requires quite a lot of time to talk with citizens numbered in the thousands. With TV, one can speak once and reach something approaching the entire population of the country. Politicians still travel to Iowa to talk with crowds of hundreds or thousands. However, the bulk of their effort is spent raising money so they can purchase TV time which will let them address citizens by the millions.

By 2013, television’s reach was no longer unique. Social media entered the scene in roughly 2000 and by 2013 its reach equaled or surpassed that of television. Thus, a sepa- rate stream of communication had become a part of national discussion of political affairs. Just as TV did not completely replace interpersonal communication on whistle stop tours, the communication via social media does not completely replace television. It is an independent stream of communication that supplements in its own way the personal communication of traveling to visit citizens and the broadcast communication available via television.

The character of social media as an independent domain for public discussion of poli- tics is the focus of this article. We have chosen agenda setting as a particularly apt con- ceptual framework for comparing mainstream media with social media as a domain of political discussion. We will summarize research on agenda setting later, but in very broad outline it is: the reach of mainstream media especially television, led by influential media interpreters such as the New York Times that have circumscribed their focus of attention in terms of the political arguments taking place among political leaders. To make our point as clearly as possible, we have chosen a case that is about as different from that broad outline as possible.

The case is the revelation of the surveillance practices of the US National Security Administration (NSA) and the UK Government Communications Headquarters (GCHQ). A single individual, Edward Snowden, left his position in the “intelligence industry” with a very large volume of documents that revealed to the world how the agencies were attempting to capture every piece of electronic communication produced on the planet. No one was to be ignored, including the leaders of most governments. A single journalist, who could not be characterized as having been an influential media interpreter, wrote the stories based on the documentation provided by Snowden. The vehicle of publication was not the New York Times, it was not major television, it was the Guardian; a news publisher with a substantial history but well down the list of globally influential media outlets. In addition, in the 10 months from June 2013 to April 2014, there were 20 million messages posted to Twitter mentioning either the NSA or Snowden. No politician took the lead in this discussion. Most talked about it as little as they could. The mainstream media had to wait for the stories published by the Guardian before they could report the story at all, and the 20 million tweets were an important part of the discussion by spread- ing the news as it was being reported and keeping the story from a quick death. Much

1918 new media & society 18(9)

news is episodic; here today and gone tomorrow. This news focus continued for month after month.

After reviewing the agenda setting literature and describing the methods for data col- lection, the article will proceed in three steps. First, the message stream mentioning NSA and Snowden will be traced. The point of the section is demonstrating the extent of com- munication that was happening in social media and how it might equate to the reach of television. Second, we will show that the communication needs to be understood as going well beyond simple expression of a point of view. There is a great deal of interac- tion in this stream unlike television/mainstream media which is broadcast with little or no interaction. Third, those two sections are preparation for the section on how the stream is different from mainstream media but also how the two streams interact. In this case, the social media stream of communication is largely driven through sharing, and the sharing is carried by references to external sources. The sources that are referenced will be shown to be very different from mainstream media. References to NSA and Snowden in the mainstream media will be tracked with special focus on the extent to which they refer to Twitter communication. The point of the section is to demonstrate the semi- independence of the social media stream and the interaction between this stream of com- munication and the communication of the mainstream media. That leads to the conclusion that agenda setting can no longer be understood as a monopoly of the mainstream media indexed to the actions of political leaders. Giving voice to the people introduces an entirely new dimension of agenda setting.

Agenda setting

By the early 20th century, the rise of journalism and mass communication invited the inference that the national agenda was importantly a function of the decisions made by those in the news business (Lasswell, 1927; Lippmann, 1922; Park, 1922). While the notion of the “hypodermic” injection of political information into the public stream was contested (Klapper, 1960), scholarly embrace of the agenda-setting hypothesis repre- sents one of the broader threads of social science research (see, for example, Bacharach and Baratz, 1962, 1970; Behr and Iyengar, 1985; Cobb and Elder, 1972; Cohen, 1963; Cook et al., 1983; Ebring et al., 1980; McCombs and Shaw, 1972; see also Dearing and Rogers, 1996 for a useful summary). Indeed, Klapper’s “minimal effects” thesis has been described as “one of the most noticeable embarrassments of modern social science” (Bartels, 1993).

The agenda-setting process can be seen as unfolding in three components: the media agenda, the public agenda, and the policy agenda, each with its own distinct body of research. Two additional concepts, priming and framing, also lie at the nexus of mass communication and the public mind. Priming denotes the role of accessible information in shaping cognition. Media attention to particular issues serves to prime the capacity of such issues to rise in public awareness (Iyengar et al., 1984; Iyengar and Kinder, 1987; but see also Jenkins, 2002; Johnston et al., 2004; Lenz, 2005; Miller and Krosnick, 2000). Framing is the contextual perspective in which news is embedded (Callaghan and Schnell, 2005; Iyengar and Kinder, 1987; Lasswell et al., 1952; MacKuen, 1981; McCombs, 1981; Schattschneider, 1960; but see also Maher, 2001). Frames suggest the

Boynton and Richardson 1919

interpretative meaning ascribed to events and ideas. Absent counterframing and critical and independent analysis of dominant frames become highly unlikely if not impossible (Entman, 2004). Within the media, certain news organizations are recognized for their significant influence on the national discourse, particularly the New York Times (Boydston, 2013; Dearing and Rogers, 1996; Grau, 1976; Mazur and Lee, 1993; Weiss, 1974; Winter and Eyal, 1981).

Researchers have also addressed the actors capable of influencing the media in its agenda-setting role. Protess, et al. (1991) emphasize the interaction between journalists and policy makers (see also Cook, 1998). Glynn (2005) analyzes a range of events (see also Birkland, 1998; DeFleur, 1987; Deutschmann and Danielson, 1960), activities, and organizations (see also Danielian and Page, 1994; Manheim and Albritton, 1984) that help shape the media’s agenda. Students of political communication have noted in par- ticular the ability of top government officials to shape the agenda (Bennett, 1983, 1991). For Bennett et al. (2007), “journalists index the range of viewpoints in the news to the divisions of power they perceive within various decision making circles of government” (p. 217 n. 2; see also Tannenbaum, 1955). Entman (2004) describes the processes of cascading activation through which the thoughts and feelings that support specific news frames flow back and forth among the different hierarchical strata of top White House officials, other political elites, media, and the public. For Entman, the president and key advisors possess the greatest capacity to influence the agenda. Bennett et al. (2007) argue that rather than critically examining official pronouncements, the media essentially records and reports them. Patterson (1998) questions whether the media, which is “delib- erately short-sighted,” (p. 55) and “rooted in novelty rather than precision,” can ever serve as a suitable counterweight to elite hegemony.

Just as earlier technological innovations shaped the news environment and the pro- cess of agenda setting, so too do current developments carry the potential to restructure the process (see, for example, Bennett and Entman, 2001). Callaghan (2005: 179–180) suggests that the rise of nontraditional forms of communication pose particular chal- lenges to extant approaches to framing (see also Baum, 2002; Entman and Herbst, 2001). Woodly (2008) suggests that blogs, closely followed by elite journalists, have the poten- tial to “change the way political communication works.” Blogs, Woodly argues, offer a conduit for the bottom-up percolation of ideas, in contrast to the top-down cascade domi- nated by elites (see also Drier and Martin, 2010; Wallsten, 2007, 2013). Others have suggested, however, that blogs are largely dependent on the mainstream media (Lee, 2007; Sweetser et al., 2008). Wallsten (2014) found that mainstream media references to Twitter users were dominated by nationally recognizable political leaders (see also Boersma and Graham, 2013). Parmalee (2014) specifically explored political leaders’ use of Twitter and its impact on US journalists (see also Jang and Pasek (2015) on the interaction of Twitter and mainstream media).

Research methods

Twitter is the new media communication stream used in this research. It is by default public communication. It is used very widely. The messages about politics number in the millions per day. And in reading the messages mentioning politics, it is clear that public

1920 new media & society 18(9)

communication is not only the default, it is also the intention. The writers are addressing others. They do it in a number of ways, but there can be little doubt that public commu- nication is what is intended unlike in a service like Facebook where the intention is generally to limit communication to an established group.

The Twitter collection was acquired using Desktop Archivist, which is a program run- ning on a computer using the Windows operating system. The program queries the Twitter Search application programming interface (API) every 5 minutes and Twitter responds with up to 1600 tweets. The program acquires the complete metadata for each Twitter message. It can obtain up to 432,000 tweets per day. However, if the stream of messages is flowing faster than 1600 in 5 minutes, only a time-based sample is obtained. In addition, about 1% of the messages were missed due to computer or network malfunc- tion. The number of messages was adjusted for 28 June 2013 through 7 July 2013 because of machine failure. The full counts were obtained from Twitter for the week and the sur- rounding weeks, and the ratio of the count for the week to the surrounding weeks was used to interpolate the counts in the collection. The search term was NSA OR Snowden. The collection began on 24 June 2013, which was 19 days after the first report on NSA practices and continued through April of 2014. Fourteen million tweets were obtained. That is a substantial percentage of the 20.7 million messages posted to Twitter.

All URLs in messages posted to Twitter are shortened by Twitter. One cannot identify the originating server by inspecting the shortened URLs. To analyze the URLs, a sample of 10,000 tweets was drawn for each week. A computer program used the shortened URL to activate and obtain the original URL. A small fraction could not be found either because the file no longer existed or because it had been moved or because the server did not respond within 3 seconds. The resulting list of URLs is used to analyze the sources referenced in the Twitter messages.

A search of the “source type” “newspapers” included in the Lexis-Nexis database was used to compile our measure of the stream of “print” coverage; the “broadcast tran- scripts” database in Lexis-Nexis was used to compile our measure of the stream of “broadcast” coverage. These databases are international in focus. Searches were per- formed for stories including either Snowden or NSA, and separately, either Snowden and Twitter or NSA and Twitter during June of 2013 through April of 2014.

Reach

The advantage of television was reach. By the 1960s, it reached across the entire coun- try and became available to everyone. Not everyone was interested enough to watch the news, and in the 1990s, cable TV split the audiences as the number of available news broadcasts increased. Even so, it is possible to plan to reach the nation through the population’s use of TV. If social media is an important mode of communication, it must also be widely available. In this section, reach of the messages posted to Twitter is examined.

Figure 1 traces the number of messages per week that mentioned either NSA or Snowden. The pattern is straightforward. There is an initial surge of messages of more than a million a week. During the days before our collection started, there were even more tweets than recorded here.

Boynton and Richardson 1921

After the first month, the numbers fell to between 200,000 and 300,000 per week. The spike in October was a response to the discovery that Chancellor Merkel’s phone had been tapped by the NSA. Then it returned to between 200,000 and 300,000 a week. During the spring of 2014, the number of tweets per week fell to approximately 200,000 per week. These are very large numbers for tweets per week for a given subject. More impressive is the continuation of messaging for 11 months. Interest in and communica- tion about the surveillance activities of NSA and GCHQ continued at a high level.

As impressive as these numbers are, they are only the beginning of the reach of mes- sages about NSA and Snowden. They were the messages posted. They do not tell how many people had access to them. There are two ways a person might find these messages. One is search. It is possible to search and find messages about a given search term using Twitter search or other software systems. However, there is no effective way for us to know how much search was being conducted during these 11 months. The second way to encounter messages is as a follower of someone who posted the message. Followers automatically receive the messages posted by those who they follow. Information about the number of followers of the persons posting a message is available as part of the meta- data of the tweet. Thus, we can count the potential reach of the messages.

Two adjustments need to be made in the counting. It is well known that fake followers are attached to individuals who have large followings. In earlier research, we determined that roughly 30% of followers of individuals who had more than 1000 followers were fakes (Boynton et al., 2013). In creating the counts used in Table 1, the number of fol- lowers for users with more than 1000 followers was multiplied by .7. The second consid- eration is the density of the networks of users posting a message to Twitter. If a person is a follower of two persons who post a message, that person will receive tweets from both. But counting that person as a follower of two persons exaggerates the total number of followers in the system; one person is counted twice. Computing the density for net- works for this number of users would be a monumental task. In research that is not yet published, we computed the density of the networks of potential candidates for the

Figure 1. NSA/Snowden Tweets per Week.

1922 new media & society 18(9)

Republican nomination for the presidency. On average, individuals followed four candi- dates. If you added all of the followers of the candidates, you would then have to divide that number by four to get to the unique number of individuals present in the networks. That does not mean that the users following the candidates only followed an average of four other Twitter users. It does mean within this specific network configuration that the users were following an average of four candidates. The density of networks such as constructed by posting about NSA or Snowden is unknown, but it seems safe to assume that the average would be at least two or three.

Table 1 gives the results for a week late in June (2013), a week late in September (2013), a week late in November (2013), and a week late in April (2014). Given the amount of information, the numbers for all 45 weeks cannot readily be presented. However, the results show considerable stability.

The number of individuals posting a message to Twitter decreases from 233,905 to 120,972 over the 11 months. However, the average number of messages posted stays relatively constant at approximately two per person posting. The distribution of the num- ber of followers is important for two reasons. First, it is a sharp difference for these mes- sage streams compared with the total Twitter flow of messages. The best estimate of the number of followers for all Twitter users is that 85% have fewer than 50 followers. For this population, the percentage with 50 or fewer followers is just over 20%. Approximately, 77% have more than 50 followers and more than 30% have between 100 and 500 follow- ers. This difference is not unique to messages about NSA and Snowden. More generally, distributions similar to this are found for a wide range of collections of communications about politics (Boynton et al., 2014). Second, it helps make plausible the high average number of total followers once that has been adjusted for network overlap. The average number of followers is 795–875. That seems reasonable given the distribution in which the mode is 101 to 500. If 233,905 users have an average of 797 followers, then 186 mil- lion followers do not seem outlandish. These are the number of individuals receiving a message during a week; not on a single day. They are very large numbers however. Concern about the surveillance activities of NSA and GCHQ was not limited to US citi- zens. Messages were being posted and read around the world. Even so, the numbers appear to be very close to the total number of active users of Twitter.

Table 1. Assessing audience size.

20/06/2013 29/07/2013 24/11/2013 23/03/2014

Users posting 233,905 182,551 125,417 120,972 Average posts 2.1 2.1 2.2 1.8 % followers 50 or fewer 22.4 23.3 21.0 16.8 % followers 51–100 11.7 11.1 11.1 10.2 % followers 101–500 37.4 33.5 34.2 33.0 % followers 501–1000 13.6 16.7 17.5 19.2 % followers 1001–5000 10.9 11.3 11.9 14.7 % followers 5000+ 3.8 4.1 4.2 4.0 Average followers 797 875 795 819 Total followers divided by 3 186,428,062 159,795,379 99,812,212 99,093,476

Boynton and Richardson 1923

What can we know? These were large streams of messages posted by large numbers of individuals. The follower numbers are in the millions. The top 100 user accounts, in terms of followers, have between 1 million and 12 million followers. It seems clear that the reach is equal to or exceeds the reach of television. It is also clear that research needs to be done to get a better understanding of the density of networks constituted in political communication.

The section began by noting that the advantage of television as a technology for the public discussion of politics is its reach. Everyone in the United States owns a TV set or in one way or another is able to access television broadcasts. If social media are going to be an equally important technology for public discussion of politics, then the reach needs to also be great.

In 2014, Twitter had 60 million active users in the United States (The Statistics Portal 2014b) and Facebook had 151.8 million users at the same time (The Statistics Portal 2014a). These message streams begin to have the same reach as television news.

Interaction

Speaking to an audience, such as on a whistle stop or the modern equivalents, is broad- cast to a people who can understand themselves as an audience. If the audience is small enough, there can be interaction between speaker and audience. It is a form of interper- sonal relations that is given meaning, at least in part, because of its unmediated per- sonal character. Television is broadcast, but it is an imagined audience that has difficulty understanding itself as anything other than isolated individuals or very small groups. There is no sense of a broader audience or crowd. There is no interaction in this broadcast medium. Twitter is broadcast. Tweets are broadcast to the world, and one interpretation of Twitter messages is that they are simply expression of rather isolated individuals. Rather like the expression that does not require an other, for example, when you pound your finger with a hammer, the messages posted to Twitter are expres- sive. They do not require an other and for the most part they are ignored by others. That is a rather standard interpretation of Twitter communication. There is no direct sense of audience in this communication, but unlike the expressive interpretation, there is interaction. The interaction is with other individual Twitter users rather than a distant figure as in TV or even the distance of a large audience. In political communication via Twitter, there are relatively high levels of interaction which is distinctive for a broad- cast form of communication.

Interaction in Twitter communication is possible both in a search mode and in relation to followers. When we studied the use of Twitter in connection with the State of the Union addresses, we found this pattern (Boynton and Richardson, 2014). Before the address, there was both reference to other sources using URLs and search using #sotu or other relevant search terms. During the hour of the address both decreased, but the use of URLs decreased more than search terms. After the speech tweets, using the search terms increased but the use of URLs did not increase. This is interaction in search. They were using standard search terms to find what others were writing and to share their own thoughts. The most interesting feature of that pattern was the recovery of search terms after the speech. Twitter users wanted to share their views with others using the same

1924 new media & society 18(9)

search terms. Hashtags have this character. They are agreed ways to find others inter- ested in a topic in which you are interested.

The week in November illustrates this form of communication. Thirty-three percent, 91,910, of the tweets contained one or more hashtags. There is no necessary limit on hashtags per tweet and the total number of hashtags was 137,257 for an average of 1.5 hashtags per tweet containing a hashtag. Since each tweet contained either NSA or Snowden, it is not a surprise to find that these were the dominant hashtags. NSA was used as a hashtag 43,798 times and Snowden was a hashtag 16,911 times. Together, they were 44% of the number of hashtags appearing in the tweets. It is a bigger leap to con- clude that these were interactions through search to a greater extent than was the case for the messages surrounding the State of the Union address. But there are hashtags appear- ing much less frequently that seem to suggest interaction through search. One is Anonymous, which is the name of a group of computer activists committed to making the world a better place using their computers. Anonymous was used as a hashtag 1884 times, and there seems little point in using it as a hashtag other than inviting communica- tion by individuals with a shared interest. The same is true for Benghazi. The deaths at Benghazi were a symbol of the utter incompetence of the Obama administration for very alienated Americans. Benghazi was used as a hashtag 1475 times, and like Anonymous seems to be used as a call for communication about NSA and Snowden among those who identified their alienation with the perceived failure in Benghazi.

It is clear that hashtags are used to call for communication among those interested in a subject. But the extent to which that call is effective is supposition. Supposition is not necessary with retweeting. A retweet is a form of interaction between the person who originally wrote the message, the persons who read it and wanted to pass it along to fol- lowers, and the followers who read the retweet. Retweeting is interaction. Figure 2 shows the percentage of messages containing a retweet for the entire 11 months.

The number of tweets per week varies greatly. The number was very large at the early stages, the number fell, there was a spike in October, and the number fell again. But the percentage of these messages containing a retweet is practically a constant. It reaches 50%

Figure 2. Percentage NSA/Snowden Tweets Containing RT @.

Boynton and Richardson 1925

once and fell below 40% several times. But a linear approximation declining slightly over time seems a very good fit to the data. Forty plus percent of the messages are retweets.

The reach of the Twitter messages is great. For a broadcast medium, the extent of interaction in the stream of messages is substantial. Both use of hashtags and retweeting show relatively high levels of interaction. Thus, the stream of Twitter messages mention- ing NSA and Snowden can be understood as both broad and interactive, making it a distinctive mode of large-scale communication.

Two streams

This section compares the two streams of communication. The questions are as follows: (1) how are they similar and how are they different? and (2) what is the pattern of interac- tion between them?

First, the news media coverage of the revelations and controversies is presented. Lexis-Nexis was used to count newspaper stories and TV/radio broadcast transcripts for mention of NSA or Snowden—the same search terms as used in searching for messages posted to Twitter. The Lexis-Nexis “broadcast” database is comprises a sample of English language TV and radio sources from across the globe. The Lexis-Nexis “newspaper” database includes a very large and geographically diverse number of English language periodicals and news service reports. The counts are plotted in Figure 3.

The count for newspaper stories is plotted by the blue line, and the count for TV/radio is plotted with the red line. The first thing to note about the two is the difference in scale. There are many newspapers in the United States relative to the number of TV/radio broadcasters. The range for newspaper stories is given on the left vertical axis and ranges from 0 to 3000. For TV/radio, the range is from 0 to 700 and is located on the right verti- cal axis.

For newspapers, there is an early surge in stories, followed by a drop, a spike in the middle of the period, and two more very small spikes in December and January and then it trails off to about 100 stories per week. That pattern is similar to the pattern of Twitter messages with two notable exceptions. First, the newspapers got off to a slow start on this story. It was not until July that the first spike occurred. The surge on Twitter was much earlier, and the first spike for TV/radio was also earlier. The spike in the middle of the time period for newspaper reporting was at the end of November. The surge in the middle of the time period for messages posted to Twitter was at the end of October. At the end of October, documents were released indicating that the NSA had tapped Chancellor Merkel’s telephone. That was big news for the people communicating via Twitter. That did not produce a surge in newspaper stories. What did produce a surge in stories a month later was the “heat” in Washington D.C. The argument among the politi- cal leadership had gotten so heated that the head of NSA volunteered to resign. The offer was rejected by President Obama, but the controversy in D.C. was enough to spike stories in newspapers. Apparently, that did not seem such big news to individuals using Twitter.

The pattern for TV/radio reporting was similar. It started earlier than newspaper reporting. It was the most volatile of the streams. After the initial surge, there was decline, then three spikes in October, December, and January and then it trailed off to 100 or less.

1926 new media & society 18(9)

The spike in October is similar to the spike in Twitter messages. The spikes in December and January were about interviews with Snowden. The spike in November occurred at the time of the first major interview with Snowden in Russia. There are small spikes for Twitter messages and newspaper stories at the same time. The spike in January reflects the Obama administration saying there would be no pardon for Snowden, Snowden admitting he could not return, and Republicans up in arms over the actions of the NSA demanding that it be disbanded. Again, there are small spikes in the other streams at the same time. One might invoke indexing to account for the pattern of newspaper reporting. That may not be a correct interpretation, but what is clear is that the overall patterns are relatively similar with some important differences.

Interaction between the streams of communication is the second question to be addressed. One form of interaction is mentioning the other stream. The percentage of TV/radio news stories about NSA or Snowden that mentioned Twitter was 28.3%. That is an impressive reach across the “divide” between the two streams. The num- ber for newspaper stories was 8.5%. Less reach, and that is probably related directly to the organization of the two mainstream media. Television and newspapers, whether on paper or on websites, operate with different capabilities and different constraints. The differences are easy to spot, but those differences are not the sub- ject of this article. What is clear is that Twitter communication is not completely ignored by the mainstream media. There is reaching across to this quite different stream of communication.

What about Twitter? Assume for this point that the following can be taken as a good sample of what most would mean by mainstream media: ABC, Al Jazeera, CBC, CBS, Der Spiegel, Fox, The Guardian, Le Monde, MSNBC, NBC, The Intercept, Wall Street Journal, Washington Post, and the New York Times. The percentage of tweets mention- ing one or more of those news broadcasters per week was 1.1%. That very small number

Figure 3. (Lexis-Nexis) Newspaper and TV/Radio reports about NSA-Snowden.

Boynton and Richardson 1927

of mentions may also be a result of the different organization of Twitter messages. Rather than mention sources explicitly, the practice is to mention sources, such as these, with a URL reference. The practice is click on this URL, and you will be taken to the reference of this Tweet. So the URLs have to be examined to determine the practice of referring to mainstream media.

The number of messages posted to Twitter per week varied quite widely over the 11 months. Figure 4 plots the percentage of those messages that contained a URL.

The figure shows that this stream of communication is a sharing of information. Almost 70 % of the messages contained a URL, which is a reference to an independent source. Not only is the extent of sharing high, it is also impressively constant. One week, it was below 60% and two weeks, it was over 70%. For the remainder of the weeks, it was between 60% and 70%. There is a substantial emphasis on sharing what one has seen elsewhere. Via this mechanism, the “elsewhere” is being brought into the communica- tion that is the Twitter stream.

Twitter shortens all URLs. As a result, one has to unshorten them to find the refer- ence. A sample of 10,000 tweets was drawn from each week. Approximately 70% of the messages contained URLs, and the 7000 URLs per week were unshortened for this analysis.

There are three ways to assess the results. Since a source could be mentioned in many tweets, one important count is how many different sources were mentioned in the 7000 URLs. A second way to look at the concentration of mentions is to look at the 10 sources mentioned most repeatedly and determine what percentage of the 7000 URLs referred to the 10 most frequently mentioned. Third, we look at the media referenced among the top 10 to determine how many weeks they are among the top 10. A source might be men- tioned frequently enough to be one of the top 10, one week or every week. This proce- dure tells which sources are the most frequently mentioned over the 11 months.

Figure 4. Percentage NSA/Snowden Tweets with URLS per Week.

1928 new media & society 18(9)

The sample produced approximately 7000 URLs per week during the entire 11 months. Since URLs could be mentioned more than once, what was the number of independent URLs posted in the Twitter messages of the week? The average per week for the entire 11 months was 1975 with a standard deviation of 175. Each week, almost 2000 different sources were mentioned in the sample. Two thousand is a large number of sources per week. This is not concentration on a few sources.

How concentrated was this posting? Were a few URLs mentioned so frequently that most of the rest were mentioned only once or twice? Figure 5 plots the number of times the 10 most referenced sites were mentioned.

There were 7000 URLs included in the sample of tweets. For the 10 URLs, most fre- quently mentioned in a week, the average number of times the 10 were mentioned was 2041 and the standard deviation was 308.The figure shows that the number was as low as 1500 one week and just over 3000 another week. But most fell in the range of 1700 and 2400. It is not quite a constant, but it is a very stable pattern. If the top 10 were men- tioned approximately 2000 times, then the other URLs were mentioned 5000 times. Five thousand is a lot of not concentration at the top.

Seventy-four news organizations were among the top 10 most frequently mentioned organizations at least one week. Thirty-five were in the top 10 only one of the 44 weeks. That included well-known news organizations such as: Le Monde, The Atlantic, Time, and The Wall Street Journal. Many more were either not ordinarily thought of as news media, for example, the American Civil Liberties Union (ACLU), or were not ordinarily thought of as important news organizations, for example, The Jakarta Post.

Three organizations were in the top 10 all 44 weeks: The Guardian, YouTube, and Russian Television. The Guardian was there because it was the primary source of reports based on the documents made available by Snowden. YouTube is a favorite of Internet users. Then there is Russian Television which was at the top each week.

Figure 5. Total References for Ten Most Mentioned News Sources in NSA/Snowdend Tweets.

Boynton and Richardson 1929

The top third, 23 organizations, are listed in Table 2. The table is divided between leading news organizations and organizations that would not be thought of as in that category. Most of the organizations in the left column are well known, though the URL identifier may not be well known. Mashable, Wired, Techcrunch, Ars Technica, and Cnet are well known in the Internet world and reflect the interests and reading habits of the population most interested in the revelations about surveillance practices. Huffington Post and Politico are well-known political blogs. And the Electronic Freedom Foundation is an organization that takes very active interest in freedom of expression on the Internet.

The most striking feature of the list on the right is how short it is. While the news organizations included are distinguished news media outlets, only eight news organiza- tions of note were in the top 10 mentioned as many as three of the 44 weeks. Television in the United States is not included aside from CNN. Politically oriented magazines are missing.

Reference to media external to Twitter is very high in the messages posted to Twitter. The authors are sharing what they are reading, and they are reading widely with 2000 different URLs in the weekly samples. And the top is not dominated by news organiza- tions that would be construed as mainstream media. The mainstream media are there, but they are more than equaled by organizations that would not fit that category. The other organizations at the top are well known, but they are just not mainstream media in the ordinary use of that term.

Conclusion

Between the campaign of Harry Truman and those of Jeb Bush/Ted Cruz/Hillary Clinton, television was the domain in which public communication about politics was practiced. Television had the reach necessary to surpass any other communication technology. The

Table 2. Organizations most frequently mentioned.

1 rt.com 44 theguardian.com 44 2 youtube.com 44 washingtonpost.com 29 3 huffingtonpost.com 19 bbc.co.uk 19 4 mashable.com 16 reuters.com 15 5 wired.com 11 cnn.com 13 6 firstlook.org 7 spiegel.de 12 7 techcrunch.com 6 nytimes.com 7 8 theverge.com 6 forbes.com 3 9 arstechnica.com 4 10 facebook.com 4 11 download.cnet.com 3 12 eff.org 3 13 freesnowden.is 3 14 politico.com 3 15 techdirt.com 3 Totals 176 142

1930 new media & society 18(9)

extensive research on agenda setting in those decades specified how this technology was organized to configure public discussion of politics.

Our data present a case study where television could no longer set the agenda. Here, documents were made available by an individual, publication was in the Guardian, (which was not previously seen as an agenda setting publication), and there were 20 mil- lion tweets–demonstrating a very different agenda setting process for a story about changes in the practice of the politics of mass surveillance that have continued from the summer of 2013 through the spring of 2014. Decisions about governmental surveillance are now public. The USA Patriot Act needs to be reauthorized early in 2015, and the scope of government surveillance is now part of the argument about how to write the law. The secret court decision using that legislation to justify bulk collection of information about telephone calls was secret. It is now public. The law can no longer be considered only by committees of congress meeting in secret. It can no longer be handled without opposition from individuals concerned about government overreach and companies in the Internet business (Hosenball, 2015). In Europe, there is a case in court about the abil- ity of the NSA to obtain information from US Internet companies that is not available by law in Europe. (Moody, 2015) Prism was its name, and no one knew about the ability of NSA to do this because the companies were legally required to keep that information from the public. It is now public, and a European court will decide how to deal with the difference between European Union law and US law. If all of this had remained secret, if it had not found a publication home, and if the expression of concern reflected in 20 mil- lion tweets had not happened, none of this would be occurring. That is agenda setting.

Other instances of new media communication that have had an impact on the practice of politics are readily available. Occupy Wall Street is an obvious example. It was begun by individuals no one had heard of, and it became a worldwide movement because of the ability to use new media to organize and articulate their concerns. Skewed income distri- bution had not been a part of political discussion for decades before Occupy Wall Street acted. It was not on the agenda, but protests and millions of tweets put it onto the agenda. Any protest movement can become an agenda setter through some mechanism other than important political leaders taking a lead and news media following because protest is, de facto, protest against those establishments. New media opens a new stream of communi- cation that is not as easily limited.

Breaking secrecy is one way of agenda setting. Protests are another agenda-setting tool. But it is certainly not limited to those. The day Senator Cruz announced his inten- tion to seek the nomination of the Republican Party, a hashtag was “invented”— #TedCruzCampaignSlogans. Twitter users employed that as hashtag and then filled in what the campaign slogans would be. During day 1 through day 3, there were more than 80,000 tweets posted using the hashtag. Those 80,000 tweets potentially reached 49 mil- lion followers. This is a form of communication that was impossible in the age of televi- sion. It is fun, as is obvious from reading them, and it is informing about a candidate who was not very well known. Political discussion is no longer limited to the agenda setting structures of 1948 through 2012. It will also flow independently, and our task is to spec- ify how it will work as carefully as did the scholars for the television age.

Agenda setting is the structure of public communication about politics and the conse- quences that flow from that structure. A new stream of communication has been added.

Boynton and Richardson 1931

Theorizing when and how the new stream becomes important in the public discussion of politics is only the beginning. Each agenda-setting case in which new media play a sig- nificant role will be different in specifics. The focus of this article has been on the char- acteristics they will share: reach, interaction, and broadening the ideas being brought into the discussion. This single case of agenda setting illustrates all three.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Bacharach P and Baratz M (1962) The two faces of power. American Political Science Review 56(4): 947–952.

Bacharach P and Baratz M (1970) Power and Poverty. New York: Oxford University Press. Bartels LM (1993) Messages received: the political impact of media exposure. American Political

Science Review 87: 267–285. Baum M (2002) Sex, lies, and war: how soft news brings foreign policy to the inattentive public.

American Political Science Review 96: 91–109. Behr RL and Iyengar S (1985) Television news, real-world cues, and changes in the public agenda.

Public Opinion Quarterly 49: 38–57. Bennett WL (1983) News: The Politics of Illusion. New York: Longman. Bennett WL (1991) Toward a theory of press-state relations. Journal of Communication 40(2):

103–125. Bennett WL and Entman RM (2001) Mediated Politics: Communication in the Future of

Democracy. Cambridge: Cambridge University Press. Bennett WL, Lawrence RG and Livingstone S (2007) When the Press Fails: Political Power and

the News Media from Iraq to Katrina. Chicago, IL: University of Chicago Press. Birkland TA (1998) Events, mobilization and agenda setting. Journal of Public Policy 18(1):

53–74. Boersma M and Graham T (2013) Twitter as a news source: how Dutch and British newspapers

use Twitter in their news coverage, 2007–2012. Journalism Practice 7(4): 446–464. Boydston AE (2013) Making the News: Politics, the Media, and Agenda Setting. Chicago, IL:

University of Chicago Press. Boynton GR and Richardson GW Jr (2014) Reframing audience: co-motion at #SOTU. In: Solo

AMG (ed.) Political Campaigning in the Information Age. Hershey, PA: IGI Global, pp. 148–164.

Boynton GR, Bates A, Bettis E, et al. (2013) The reach of politics via Twitter—can that be real? Open Journal of Political Science 3(3): 91–97.

Boynton GR, Cook J, Daniels K, et al. (2014) The political domain goes to Twitter: hashtags, retweets and URLs. Open Journal of Political Science 4(1): 8–15.

Callaghan K (2005) Controversies and new directions in framing research. In: Callaghan K and Schnell F (eds) Framing American Politics. Pittsburgh, PA: University of Pittsburgh Press, pp. 179–189.

1932 new media & society 18(9)

Callaghan K and Schnell F (2005) Framing American Politics. Pittsburgh, PA: University of Pittsburgh Press.

Cobb R and Elder C (1972) Participation in American Politics: The Dynamics of Agenda-Building. Boston, MA: Allyn & Bacon.

Cohen BC (1963) The Press and Foreign Policy. Princeton, NJ: Princeton University Press. Cook FL, Tyler TR, Goetz EG, et al. (1983) Media and agenda-setting: effects on the public,

interest group leaders, policy makers and policy. Public Opinion Quarterly 47(Spring): 16–35.

Cook TE (1998) Governing with the News. Chicago, IL: University of Chicago Press. Danielian LH and Page BI (1994) The heavenly chorus: interest group voices on TV news.

American Journal of Political Science 38(4): 1056–1078. Dearing JW and Rogers EM (1996) Agenda-Setting. Thousand Oaks, CA: SAGE. DeFleur ML (1987) The growth and decline of research on the diffusion of the News 1945–1985.

Communication Research 14(1): 109–130. Deutschmann PJ and Danielson W (1960) Diffusion of the major news story. Journalism Quarterly

37: 345–355. Drier P and Martin CR (2010) How ACORN was framed: political controversy and media agenda

setting. Perspectives on Politics 8(3): 761–792. Ebring L, Goldenberg EN and Miller AH (1980) Front-page news and real-world cues: a new look

at agenda setting by the media. American Journal of Political Science 24(1): 16–49. Entman RM (2004) Projections of Power: Framing News, Public Opinion, and U.S. Foreign

Policy. Chicago, IL: University of Chicago Press. Entman RM and Herbst S (2001) Reforming public opinion as we have known it. In: Bennett LW

and Entman RM (eds) Mediated Politics: Communication in the Future of Democracy. New York: Cambridge University Press, pp. 203–225.

Glynn CJ (2005) Public opinion as a social process. In: Dunwoody SL, Becker LD, McLeod DM, et al. (eds) The Evolution of Key Mass Communication Concepts. New York: Hampton Press, pp. 139–164.

Grau CH (1976) What publications are most frequently quoted in the congressional record. Journalism Quarterly 53: 716–719.

Hosenball M (2015) U.S. to stop collecting bulk phone data if spying law expires. Reuters, 23 March. Available at: http://www.reuters.com/article/2015/03/23/us-usa-telecommunications- intelligence-idUSKBN0MJ1XE20150323

Iyengar S and Kinder D (1987) News That Matters: Television and American Opinion. Chicago, IL: University of Chicago Press.

Iyengar S, Kinder DR, Peters MD, et al. (1984) The evening news and presidential evaluations. Journal of Personality and Social Psychology 46(4): 778–787.

Jang SM and Pasek J (2015) Assessing the carrying capacity of Twitter and online news. Mass Communication and Society 18(5): 577–598.

Jenkins RW (2002) How campaigns matter in Canada: priming and learning as explanations for the reform party’s 1993 Campaign Success. Canadian Journal of Political Science-Revue Canadienne de Science Politique 35(2): 383–408.

Johnston R, Haden MG and Jamieson KH (2004) The 2000 Presidential Election and the Foundation of Party Politics. New York: Cambridge University Press.

Klapper JT (1960) The Effects of Mass Communications. Glencoe, IL: Free Press. Lasswell HD (1927) Propaganda Techniques in the World War. New York: Alfred A. Knopf,

Inc. Lasswell HD, Lerner D and De Sola Pool I (1952) The Comparative Study of Symbols. Stanford,

CA: Stanford University Press.

Boynton and Richardson 1933

Lee JK (2007) The effect of the Internet on the homogeneity of the media agenda: a test of the fragmentation thesis. Journalism and Mass Communication Quarterly 84(4): 745–760.

Lenz G (2005) Learning and opinion change, not priming: reconsidering the priming hypothesis. American Journal of Political Science 53(4): 821–837.

Lippmann W (1922) Public Opinion. New York: Simon & Schuster. McCombs ME (1981) The agenda setting approach. In: Nimmo D and Sanders KR (eds) Handbook

of Political Communication. Beverly Hills, CA: SAGE, pp. 121–140. McCombs ME and Shaw DL (1972) The agenda-setting function of the mass media. Public

Opinion Quarterly 36: 176–187. MacKuen MB (1981) Research on agenda setting. In: MacKuen MB and Coombs SL (eds) More

than News: Media Power in Public Affairs. Thousand Oaks, CA: SAGE, pp. 19–39. Maher TM (2001) Framing: an emerging paradigm or a phase of agenda setting? In: Reese SD,

Gandy OH and Grant AE (eds) Framing Public Life. Mahwah, NJ: Lawrence Erlbaum, pp. 83–94.

Manheim JB and Albritton RB (1984) Changing national images: international public relations and media agenda setting. American Political Science Review 78(3): 641–657.

Mazur A and Lee J (1993) Sounding the global alarm: environmental issues in the US national news. Social Studies of Science 23(4): 681–720.

Miller JM and Krosnick JA (2000) News media impact on the ingredients of presidential evalua- tions: politically knowledgeable citizens are guided by a trusted source. American Journal of Political Science 44(2): 301–315.

Moody G (2015) EU: don’t use Facebook if you want to keep the NSA away from your data. Ars Technica, 25 March. Available at: http://arstechnica.com/tech-policy/2015/03/eu-dont-use- facebook-if-you-want-to-keep-the-nsa-away-from-your-data/ (accessed 14 May 2015).

Park RE (1922) The Immigrant Press and Its Control. New York: Harper & Brothers. Parmalee JH (2014) The agenda building function of political tweets. New Media & Society 16(3):

434–450. Patterson TE (1998) Time and news: the media’s limitations as an instrument of democracy.

International Political Science Review 19(1): 55–67. Protess DL, Fay Lomax Cook JC, Dopplet JS, et al. (1991) The Journalism of Outrage: Investigative

Reporting and Agenda Building in America. New York: Guilford Press. Schattschneider EE (1960) The Semi-Sovereign People: A Realist’s View of Democracy in

America. New York: Holt, Rinehart, and Winston. Sweetser Kaye D, Guy J, Golan, et al. (2008) Intermedia agenda setting in television, advertising

and blogs during the 2004 election. Mass Communication & Society 11: 197–216. Tannenbaum P (1955) The indexing process in communication. Public Opinion Quarterly 19(3):

292–302. The Statistics Portal (2014a) Leading countries based on number of Facebook users as of May

2014 (in millions). Available at: http://www.statista.com/statistics/268136/top-15-countries- based-on-number-of-facebook-users/

The Statistics Portal (2014b) Number of monthly active Twitter users in the United States from 1st quarter 2010 to 2nd quarter 2015. Available at: http://www.statista.com/statistics/274564/ monthly-active-twitter-users-in-the-united-states/

Wallsten KJ (2007) Agenda setting and the blogosphere: an analysis of the relationship between mainstream media and political blogs. Review of Policy Research 24(6): 567–587.

Wallsten KJ (2013) Old media, new media sources: the blogosphere’s influence on print media news coverage. International Journal of E-Politics 4(2): 1–20.

1934 new media & society 18(9)

Wallsten KJ (2014) Microblogging and the news: political elites and the ultimate retweet. In: Solo AMG (ed.) Political Campaigning in the Information Age. Hershey, PA: IGI Global, pp. 128–147.

Weiss CH (1974) What America’s leaders read. Public Opinion Quarterly 38: 1–22. Winter JP and Eyal CH (1981) Agenda setting for the civil rights issue. Public Opinion Quarterly

45: 376–383. Woodly D (2008) New competencies in democratic communication? Blogs, agenda setting and

political participation. Public Choice 134(1–2): 109–123.

Author biographies

GR Boynton is a professor of new media and politics in the department of political science at the University of Iowa. The focus of his research is on public communication about politics.

Glenn W Richardson Jr. is a professor in the Department of Political Science and Public Administration at Kutztown University of Pennsylvania. His research is focused on political communication.