What’s happening?
Analysing the impact of Twitter
on the successes of public relations efforts
K.R. van der Maas 10595694
Master’s Thesis
Graduate School of Communication
Corporate communication, Communication Science Supervisor: Prof. dr. R. Vliegenthart
Friday June 28, 2019
ABSTRACT
Within the field of public relations, efforts are made to gain agenda building capacities. To do so, press releases are used to affect news coverage. The context in which these press releases are spread are of main interest in this study.
Focussing on the context of social media, this study tries to answer to what extent social media influence the agenda building capacities of organizations. Efforts made by both government ministries and corporate organizations on Twitter and their impact on journalistic churnalism are studied. A quantitative content analyses in conducted on 200 Dutch press releases, its following attention on Twitter (13472 tweets) and its following Dutch news coverage (403 news
articles). Results show the effort of government ministries and corporations to be of marginal significance on journalists using information from press releases. Overall attention and intensity of sentiment on Twitter to shows to be a solid pre-indicator for news coverage, arguing the context of social media to be a factor impacting news coverage, regardless of direction of sentiment.
INTRODUCTION
The reputation of an organization is dependent on the type of media attention it receives. Therefor, it is beneficial for an organisation to know what factors affect gaining and
influencing media attention. Media outlets are influenced by a wide variety of actors trying to determine which issues are discussed, also referred to as agenda setting, as well as how they are discussed (i.e. framing). This influences the way information sources are used, for
instance information subsidies like press releases (Verhoeven, 2009). Media determine how to utilize information based these actors or their own agenda, known in literature as agenda setting theory (McCombs & Shaw, 1972; McCombs, 2009). This theory states that covering certain (elements of) objects by organizations influences the prominence of those (elements of) objects on the agenda of news media (Carrol & McCombs, 2003), further to be discussed in the theoretical framework. Due to their differential treatment of press releases, media differ in both the use of press releases for news coverage, as well as in how they utilize these press releases in news production. For Dutch national newspapers, one in ten articles is initiated by a press release (Boumans, 2018). This can be explained by certain news factors, as Boukes and Vliegenthart (2017) have identified. These news factors were identified (mostly portrayed in economic news coverage); personification, negativity, controversy, influence and
relevance, eliteness, geographical proximity. Other news factors are surprise and impact-negative consequences (Schafraad, Van Zoonen & Verhoeven, 2015). Factors as such can be considered pre-indicators for media attention, meaning these factors increase the likelihood of information from press releases to be used by Dutch news media. By taking these factors in consideration, organizations are able to influence the effectiveness of press releases. While these factors are considered of influence in regards to their effectiveness in gaining news media attention, the role of social media attention is not taken into regard as a potential pre-indicator increasing the likelihood of information from a press release to be used by Dutch
news media. As social media create a context to spread content of press releases, they have different characteristics, but this doesn’t mean they can not be considered a factor for news media to be influenced by.
This study will focus on the role of the social media platform Twitter and whether this social media platform can be considered a pre-indicating factor in regards to the effectiveness of press releases of organisations. It studies the context of Twitter and its influence on
journalistic news production. As Waters, Tindall & Morton (2010) state in their research on changes in journalism “public relations, including media relations, cannot fall into a routine
of producing static programming; instead, practitioners should seek out new communication channels and possibilities for engaging all of our stakeholders.” They argue journalists
increasingly use social media as platform to find experts and content on subjects (Waters, Tindall & Morton, 2010). Regarding US news media, journalists working in mainstream media embrace Twitter as a tool for reporting while maintaining routines in selecting specific sources (Moon & Hadley, 2014). Therefor, social media attention to press releases may be a considerable factor to influence news coverage about the concerned press releases. By gaining more extensive insight into the question if this is a solid case and how publication of press releases can be more effective, this research adds value for the strategic decision-making of organizations to gain agenda building capacities and for academic literature into the field of corporate communication. To gain insight, the type of organization studied is of major
importance. Because of their public role to inform the public, government organizations must be able to gain media attention. Government organizations tend to use social media, because social media can be informational, interactional and collaborative in nature (Brainard & McNutt, 2010; Sandoval-Almazan & Gil-Garcia, 2012). Therefor, the public sector is taken into account in this study. The commercial sector with corporations also uses social media like Twitter to gain attention for their organizational goals. Both corporate and government
organizations are taken into account in this study as they both use social media to gain news media attention. As they differ, for example in their organizational goals, they could differ in their effectiveness of using Twitter. Overall, do social media platforms (in regard of this research Twitter) pre-indicate the relevance of a press release for journalists? And if so, are there differences between government and corporate organizations? To research this, the following research questions are proposed:
RQ1: To what extent do social media influence the agenda building
capacities of press releases by organizations on publications of news articles by news media?
RQ2: To what extent do government ministries and corporations differ in their agenda building capacities through Twitter?
To answer these research questions, current literature on agenda building capacities of organizations, journalistic churnalism and factors impacting success of press releases are discussed. However, existing literature on effects of social media on the success of press releases is not abundant. This study addresses this gap by conducting quantitative content analysis on press releases, attention on Twitter for these press releases and the following news coverage on information from press releases.
THEORETICAL FRAMEWORK Agenda building capacities
Agenda setting theory is a well-known framework within the field of communication science, coming from the political communication domain (McCombs & Shaw, 1972). This theory states that organizations are to a certain extent able to set an agenda by putting forth subjects
(first level) and by emphasizing certain attributes (second level). The theory states in debated subjects covered in traditional news media not to be directly chosen by organizations. This gatekeeping role is a role maintained by journalists (Strömbäck, 2008). The discipline of public relations and/or press offices in turn make efforts to create attention and/or salience for these subjects. This salience on objects or attributes of an object, which in turn influences the prominence of those (attributes of) objects, is at the core of agenda setting theory (Carrol & McCombs, 2003). As agenda setting theory states, attention for certain objects in media can be categorized as ‘first level agenda setting. For example, an organization wants to gain attention for their health week and publishes information on health subjects. The salience for attributes of these objects can be categorized as ‘second level agenda setting’. By certain salience in communications, an organization can set the agenda on the first level (by putting forth a certain topic) and the second level (by putting forth salience on attributes in their communication efforts). This salience in communications by an organization can therefore have influence on the way stakeholders approach a certain object, according to agenda setting theory. From this theory, an agenda building perspective emerged. Agenda building is often defined by scholars as a dynamic exchange of priorities among stakeholders (Kroon & Van der Meer, 2018). It is a process shaped by the supply of information to journalists by organizations (Schafraad, Van Zoonen & Verhoeven, 2015). As journalists are in control of the news agenda, this information supply is of importance for both the organization and the journalist. Organizations have a interest in the attention they receive from media for
organizational and reputational goals (Kiousis et al., 2007) and journalists have a interest in the information they receive from organization for their journalistic goals. Both sources and gatekeepers can benefit from this mutual relationship (Kroon & Van der Meer, 2018). Current studies have revealed significant relations between news media agendas and organizational agendas (Moon & Hyun, 2014; Kim, Kiousis & Xiang, 2015). Although a mutual relation can
be considered beneficial for both parties, time constraints and work pressure increases
journalist’s dependence on information distributed by organizations (Davis, 2008). Journalists increasingly consume and process public relations material, which do not always benefit public interest (Jackson & Moloney, 2016). As for Dutch news articles, there are several findings which state one third of all Dutch domestic news is partly or fully copied from press releases (Hijmans, Schafraad & Buijs, 2009; Hijmans, Schafraad, Buijs & d’Haenens, 2011; Scholten & Ruigrok, 2009; Kroon & Schafraad, 2013). Other findings state one in every ten Dutch news articles is initiated by a press release; for press agencies this is slightly higher (Boumans, 2018).
These findings show an intertwined relation between corporate information subsidies and news agenda on the first level of agenda setting, in which they either influence each other or in which corporations lead (Ragas et al., 2011; Kroon & Van der Meer, 2018). Information subsidies from organizations can be distributed through press conferences, spokespersons, or press releases. Press releases are regarded to have a primary role in agenda building processes (Turk & Franklin, 1987; Kiousis et al., 2007). With these processes, agenda building capacity is regarded as the success of the topic in provided information subsidies achieving news coverage. On the one hand succeeding in creating news coverage for a decided topic or not, i.e. first level agenda setting (Schultz, Kleinnijenhuis, Oegema, Utz, & Van Atteveldt, 2012; Kroon & Van der Meer, 2018). On the other hand, when content of press releases explicitly resonates as intended in news coverage and with the intended salience on attributes. Other variables used to measure agenda building capacities are the amount of news articles
regarding the press release and the one on one reproduction of information, also referred to as churnalism (Davis, 2008).
Churnalism
Churnalism refers to the process of copying information subsidies from organizations (Jackson & Moloney, 2016). These information subsidies distributed through press releases end up at news media through press agencies like the Dutch press agency ANP (currently the only Dutch press agency) and through the organization’s own publications. Where public relations practitioners often refer to success of a press release as one on one exact formulation of text, a balanced article is eventually valued more (Jackson & Moloney, 2016). This balance is needed according to public relations practitioners to stay credible. Higher degrees of
journalistic churnalism erodes trust between news media and its reader, which in turn erodes the credibility of a news medium (Jackson & Maloney, 2016).
When not distributing press releases through press agencies, other channels like corporate websites and social media can be used. As for U.S. news media, over 90 % of journalists state their reliance on social media has increased (Middleberg/SNCR, 2011), because social media increasingly serve as a reliable tool for news stories. Although this is the case, the majority of both print media and online journalists are sceptical about the reliability of information on social network sites (Cision & George Washington University, 2009). While this scepticism is present, a quarter of newspaper journalists used Twitter as a primary source and newspaper journalists use the internet more compared to national and regional broadcasters (Hermans, Vergeer & d’Haenens, 2009; Moon & Hadley, 2014). Reporters increasingly rely on internet for newsgathering, fact checking and other activities to provide news stories (Arketi Group, 2011; Pavlik, 2000). These findings would imply that a social media platform like Twitter would serve as a broadcast medium and hereby affecting
mainstream media coverage (Ahmad, 2010). By accessing Twitter, news organizations value the function of Twitter as an awareness system to find information (Gleason, 2010). Sourcing of news media in a democratic society is a major premise of scholars (Moon & Hadley, 2014)
and technological developments are major variables at play affecting the news selection process (Livingston & Bennett, 2003). Providing information subsidies through a social network source like Twitter is one aspect which is not studied in depth in current literature. While current studies paint a picture of organizational information subsidies influencing news media attention and Twitter functioning as a perceived reliable tool by journalists to gain information, this would suggest Twitter as one of the mechanisms for organizations through which information subsidies attract news media attention and thus being an important mediating variable in agenda setting processes. Although thorough research into the role of Twitter regarding the effectiveness of press releases is not abundant, findings show
differences between type of organizations in creating attention.
Non profit and governmental organizations versus commercial organizations show differences in drawing attention from journalists, in favour of NGO’s and governmental organizations (Berkowitz & Adams, 1990; Fenton, 2010; Van Leuven & Joye, 2014). As McLuhan’s famous line goes, ‘the medium is the message’, the sender of a press release could therefor differ the outcome of effectiveness of a press release. While previous studies have showed differences between commercial corporations and government organizations, this mediating influence of Twitter as described above is not pointed out for the use of Twitter specifically (Berkowitz & Adams, 1990; Fenton, 2010; Van Leuven & Joye, 2014). As governmental and corporate organizations differ in their organizational goals and position in society, this differences is taken into regard in this study. When taking into regard that Twitter is a platform which could serve as a channel for communication of organizations and by journalists as a source for information (Pavlik, 2000; Gleason, 2010; Ahmad, 2010; Arketi Group, 2011) and to test the agenda building capacities of organizations by using Twitter as a medium, the following hypotheses are proposed:
H1: The use of Twitter by an organization to spread the publication of press releases increases publications of news articles about the press releases by news media.
H2: The use of Twitter by an organization to spread the publication of press releases increases churnalism of information from press releases in news media.
From prior research can be derived that success of press releases increases with the intensity of it’s news value (Eilders, 2006). This intensity increases with the presence of several factors, as news factor theory states.
News factors
Press releases and other publications can include certain characteristics which can determine the newsworthiness for news media (Schafraad, Van Zoonen & Verhoeven, 2015).
Newsworthiness is a journalistic judgement which is increased by the relevance of news factors (Eilders, 2006). It is assumed there is an observable and stable relationship between these characteristics and the news value assessed by journalists (Kroon & Schafraad, 2013). This implies that the relevance of certain characteristics is of importance for the effectiveness of a press release to influence the news media agenda. In current literature, these
characteristics are news factors as identified by Boukes and Vliegenthart (2017) and other underlying factors as identified by Seletzky and Lehman-Wilzig (2010). These news factors within the content of press releases include respectively personification, negativity, eliteness, influence and relevance, controversy, geographical proximity and continuity (Boukes & Vliegenthart, 2017). Additional underlying factors are the news substance (novelty and practical usefulness) of an press release, text composition, timely transfer to the journalist and original source of the item (Seletzky & Lehman-Wilzig, 2010). According to Kroon &
Schafraad (2013) as the intensity of news factors increases, the amount of news media publications increase. The joint intensity of news factors has a stronger explanatory power compared to individual news factors, concluding that the consistency of news factor intensity has an affect on the type of journalistic judgement of newsworthiness. With a higher news factor intensity, the likelihood of news media directly copying information from press releases increases (Kroon & Schafraad, 2013). The additional underlying factors (i.e. news substance, text composition, timely transfer to the journalist and originality of source) studied have found to play a part in the likelihood of copying information by journalists (Seletzky & Lehman-Wilzig, 2010). With these news factors and underlying factors, what has not been studied thoroughly in current research is the context in which press releases are distributed. Contexts such as social media to gain attention for these press releases.
Because of the uprising of social network sites (SNS) receiving news via SNS is increasing. For instance, the amount of Facebook users accessing news via SNS is two thirds (Pew Research Center, 2017; Newman et al., 2017). With these results comes the observation of users commenting on news articles on SNS at least once a week (Newman et al., 2017). Consequently, reading comments is more widespread; half of American SNS users read comments on news articles, mostly on social media platforms (Ksjazek, 2016; Stroud et al., 2016). When algorithms of a social media platform like Twitter boost social media post with higher overall engagement (amounts of comments, likes and favourites), the reach of social media posts increase. While the reach of news articles and information subsidies via social media increases, it is arguably more likely for a journalist’s news value assessment of an item such as a press release to get higher. More and more journalists are using twitter as a source to know the actualities of the moment (Waters, Tindall & Morton, 2010). This would imply social media like Twitter to play a crucial role in the effectiveness of press releases being picked up by journalists, as it creates a context in which the attention around a press release
could increase its news value. To study this proposition and to study if social media attention can be considered a pre-indicator for news media attention, the following hypotheses are proposed:
H3: The extent to which a press release’s information received attention on Twitter increases churnalism of information from press releases in news media.
Some of the news factors (Boukes & Vliegenthart, 2017) can be found in tweets themselves. Personification and eliteness are arguably to be found in the sender of a tweet. A message tweeted by the CEO of an organization in comparison to a message tweeted by the
organization itself can make the tweet more personified. As news factor theory would suggest, this would increase its newsworthiness. Therefor the following hypothesis is proposed:
H4: The presence of tweets by CEO of an organization about a press release increase churnalism of information from press releases in news media compared to tweets by the concerned organization.
The news factors negativity and controversy are arguably to be found in the attention a tweet receives on Twitter. As negativity is found to be a news factor which increases the
newsworthiness of a press release (Boukes & Vliegenthart, 2017), it is arguably logical to expect the same for the attention a tweet receives. By receiving more negative sentiments in the attention to tweets, it would be arguable that the newsworthiness increases. Based on studies from the field of psychology, a negativity bias found in reaction of humans to news coverage. People tend to use more cognitive energy to think about negative things than to think about positive things (Abele, 1985; Fiske 1980). Negative news elicits stronger and more sustained reactions in comparison with positive news (Soroka & McAdams, 2015). This stronger reaction arguably increases the likelihood to elicit more reaction on Twitter as well.
With more reactions and a stronger sentiment on a press release, the reach and visibility of a press release on Twitter could increase, resulting in a increased news value assessment by a journalist. With this, its arguably more likely for a journalist to use more information from the press release. To test this assumption, the following hypothesis is proposed:
H5: The more negative sentiment in Twitter attention, the more churnalism of information from press releases in news media.
METHOD
A quantitative content analysis is conducted on press releases and the Twitter attention and news coverage followed by it. The press releases of three Dutch government ministries and one large Dutch corporation are included. The Dutch ministries included in this research are the Ministry of General Affairs, the Ministry of Economic Affairs and Climate Policy and the Ministry of Education, Culture and Science. These ministries are selected randomly and cover a quarter of the Dutch government (from a total of twelve Dutch ministries). To compare possible differences between government ministries and corporations as a form of
organization, Royal Dutch Shell (Shell) is included in this research. Shell is Dutch’s largest multinational corporation, with over 82.000 employees globally and a revenue of over 380 billion dollars in 2018 (Shell Annual Report and Form 20-F 2018). Both types of
organizations are of a large scale and have impact with their actions. To research whether the sector, i.e. government and corporate, differ in their use of Twitter and attention they receive on Twitter, these organizations are included in the sample. Overall, only Dutch press releases are included as Dutch news media coverage is studied in this research. For Shell, this study includes press releases from Shell Nederland.
Sample selection
The data are collected across three types of data; press releases (Npress releases=200), twitter
(Ntweets=13472) and news articles, both online and print (Nnews articles=403). The units of
analysis are individual press releases, individual tweets based on topic of press releases and individual news articles about information from press releases (print and online). For the collection of tweets and news articles, a timeframe of three days after the publication of a press release is applied to keep data-collection feasible. With this timeframe, tweets posted and news articles published three days after publication of the press release are included in this sample. During the process of method design, two interviews have taken place with a PR practitioner of a Dutch minister and with a political editor of a public national TV talk show. Both these interviews indicated a timeframe of three days to be sufficient and up to date with the pace of current news production.
The 50 most recent press releases of the included organizations are manually retrieved through online publications of the organizations on their corporate website (rijksoverheid.nl and shell.nl), resulting in 200 press releases (50 per government ministry and corporate). The collection of press releases is not at random, but based on being as recent as possible. The collection of press releases has resulted in a collection of press releases from April 2019 (start of data collection) back to October 2017. In this timeframe all published press releases of the ministries and Shell Nederland are included. The scope of this sample selection (50 press releases per organization) is set for feasibility within the timeframe of this research. The title and text of the press releases are retrieved separately for two reasons: 1) to use the title for collecting tweets and news articles on topic of the news articles and 2) to analyse text similarity between the text of the press release and the text of the news articles at an individual level.
To retrieve tweets based on topic of press releases, a combination of words from a press release’s title are used as search query. Meaning tweets with any combination of words from the press release’s title are retrieved. For example, a title like “Minister-president Rutte
in gesprek met bewoners Schilderswijk” would the following search query:
(“minister-president” OR rutte)(“in gesprek” OR “met bewoners” OR schilderswijk)
This query includes all tweets including terms ‘minister-president’ or ‘rutte’ and ‘in gesprek’,
‘met bewoners’ or ‘Schilderwijk’. For example, a tweet with the following text is included; “#zuhm rutte komt naar Zeeland grijp je kans tol of is het alleen bonbarie
door de telefoon”
(example taken from id_press_release = 1)
The results of every search query are manually checked if the combination of words in fact refer to the press release’s topic. In cases of a collection with irrelevant tweets, search queries are adjusted to lose the irrelevant tweets. This resulted in a collection of 13,472 tweets. The selected tweets are retrieved using Coosto, a social media management tool which includes mostly Dutch social media data. When studying Dutch ministries and a Dutch multinational Coosto is a good fit, as it focusses mainly on Dutch social media. By using Coosto’s database, unique tweet-urls were gathered. Because Coosto’s deliverance of data loses a lot of metadata of tweets, the unique tweet-urls are used to scrape available data from Twitter using a Twitter API. By doing so, a more extensive dataset is created.
To retrieve news articles, the same method for retrieving tweets is used; a combination of words from the press release’s title are used as search query. The database used for
collection of news articles is Nexis Uni®. This database includes publications from over 10,000 sources, including news media. From the results of the search query, only news
articles (print and online) from Dutch national news sources are selected. This results in the following news media (print and online): Algemeen Dagblad, ANP, AT5, BNR.nl, De
Telegraaf, De Volkskrant, FD.nl, Het Financieele Dagblad, EenVandaag, Het Parool, Parool.nl, Metro, Metro Nieuws, Nederlands Dagblad, NOS.nl, NRC, NRC.nl, NU.nl, Reformatorisch Dagblad, RTL Nieuws and Trouw. These print news media reach 40 percent
of Dutch citizens above 13 years of age (NOM 2019-I) and represent all of Dutch national newspaper media in print. The online news media reach 47,1 percent of Dutch citizens above 6 years (NOBO, 2019-I) and together represent over 50 percent of Dutch online news media. This selection resulted in 224 print news articles, 128 online news articles and 51 press agency articles from the Dutch main press agency Algemeen Nederlands Persbureau (ANP) (Ntotal news articles=403).
Measurement: reliability and variables
The content analysis is conducted by one individual, collecting data through online public archives. Using only existing data, only to be retrieved, this minimizes room for
(subconscious) manipulation of data extensively and increases reliability of the collected data. For this analysis the following independent variables are measured: 1) use of Twitter by an organization to spread publication of press releases, 2) the extent to which a press release’s information received attention on Twitter.
The use of twitter by an organization to spread publication of press releases is
operationalized by the total number of tweets an organization and its CEO post about a press release. In the analysis tweets by the CEO are analysed separately. The extent to which a press release’s information received attention on Twitter is operationalized as all original tweets, retweets and favourites and average sentiment of tweets about a press release. The average sentiment is calculated using SentiStrength. SentiStrength is an automatic sentiment
analysis algorithm based on multiple lexicon with multiple libraries of sentiment words, which combines two scales for positive and negative sentiment. The algorithm checks every word against the libraries of sentiment words, calculating a negative and positive sentiment score. By running this tool on every tweet included in the sample, an average sentiment score is calculated for every press release. Limiting the scope of this research by using the title of a press release for collecting news articles and retrieving tweets, it is possible some tweets regarding press releases are not included in the sample. As based on agenda setting theory, this method bases its data collection of tweets on the assumption of first level agenda setting (i.e. topic on the agenda). By putting forth a subject, it either succeeds in reaching the agenda or not. Thereby being included in this sample or not. This assumption reinforces the reliability of the collected data, creating a more strict sample and thus a conservative assessment of number of tweets bases on press releases. Only tweets and news articles which contain terms used in the title of the press release are included in the sample.
Dependent variables are 1) publications of news articles by news media and 2)
churnalism of information from press releases in news media. Publication of news articles by news media is operationalized as the number of all print, online and press agency news articles published by Dutch national news media within a timeframe of three days after the publication date of a press release. Churnalism is operationalized as the level of text similarity between a press release and news article. As churnalism is also referred to as copy and paste’ behaviour with little to no journalistic efforts (Davis, 2008), this operationalization is based on the assumption that the more similar texts are, the more text is copied and used by news media. With this assumption, conclusions do not depend on qualitative content assessment, but on the assessement of the degree of text overlap. Therefor, conclusions can be made to a certain extent on level of copying editorial subsidies (Jackson & Moloney, 2016). To calculate this level of churnalism, cosine text similarity is measured.
Cosine text similarity
To measure churnalism in text, similarity between texts is measured. To measure text
similarity, cosine similarity is calculated. Cosine similarity is a metric used for text analytics with several advantages; it calculates the similarity between texts irrespective of text size, it copes well with high dimensionality (i.e. amount of words) and its output is a score between zero and one (zero meaning no text similarity and one being total text similarity). The measured score is the calculated cosine of the angle between vectors projected in a
multidimensional space, with every dimension being a word from text. The calculated cosine is the output, giving a score between zero and one. The used mathematical equation is the following: 𝑐𝑜𝑠∅ = 𝐴 ∙ 𝐵 | 𝐴 |+| 𝐵 |+ = 𝐴,𝐵, -,./ 𝐴,+ -,./ -,./𝐵,+
In this formula, A stands for vector A (i.e. text of a press release) and B stands for vector B (i.e. text of a news article), a vector being all words from the text put into dimensions. The dot product is a multiplication of each word from the both vectors added together. For example, the text A “dog dog dog cat cat” and text B “dog cat cat cat cat”, would result in the following dot product in a two dimensional vector space (as two words are present);
Text A = (3, 2) and text B = (1, 4) 0∗/ 2(+∗4) (06)2(/6) (+6)2(46)
=
027 82/ +2/9=
// /: /7 = 0,81989159Analysis
As both dependent variables (number of publications of news articles and churnalism of
information from press releases in news media) and independent variables (use of Twitter by an organization, the attention on Twitter regarding a press releases information, tweets by CEO of an organization regarding a press release, attention on twitter regarding a press release with a negative sentiment) are measured at interval level, analysis is conducted
through regression models and ANOVA.
RESULTS
On average, press releases are followed by two news articles (M=2,01, SD=4,03). With this, the average number of articles in print (M=1,13, SD=2,34) exceed that of online articles (M=0,64, SD=1,80) and articles by press agencies (M=0,25, SD=0,56). A press release is followed by an average of 67 tweets (M=67,35, SD=126,05), a total of 89 favourites
(M=88,99, SD=309,71) and an average cosine text similarity score of cosθ = 0,33 (M=0,33, SD=0,35). For all news articles, the average cosine text similarity resulted in cosθ = 0,67 (M=0,67, SD=0,14). For government ministries, a press release is followed by an average of 65 tweets (M=65,11, SD=98,88), a total of 93 favourites (M=93,54, SD=342,62) and an average cosine text similarity score of cosθ = 0,33 (M=0,33, SD=0,35). For Shell (i.e. corporate), a press release is followed by an average of 74 tweets (M=74,06, SD=186,37), a total of 75 favourites (M=75,38, SD=179,90) and an average cosine text similarity score of
cosθ = 0,32 (M=0,32, SD=0,36).
Overall, results show the organizations do not use Twitter extensively in regards to creating attention for press releases. The average number of tweets by both the organization (CEO included) about a press release is low (M=0,54 per press release). For the number of tweets by a government ministry or corporation or the CEO about publication of press
releases, marginal significant results are found of affecting the amount of publications of news articles about the press releases by news media. The total amount of news articles decreases by the number of tweets by the organisation about the press release, on account of a marginally significance (β = -,126, t(194) = -1,685, p = 0,09, R2 = 0,071). With an
unstandardized b = -,342, for every tweet by the organization the amount of news articles decrease with 0,342. It shows the effort of organizations to create attention by tweeting about their press releases to have a counterproductive result. This result show as a organization, tweeting more about a press release does not increase the amount of news articles about it. For tweets by the CEO no significant results are found (p = 0,98). When separating the total of news articles by print (ptweets by organisation = 0,12, ptweets by CEO = 0,69), online (ptweets by organisation
= 0,34, ptweets by CEO = 0,53) and press agencies (ptweets by organisation = 0,19, ptweets by CEO = 0,58),
no significant results are found. With these results, marginal support for hypothesis 1 is found. While not being hypothesized, a few other results are worth mentioning. Significant results are found for the total amount of tweets increasing the total number news articles (β = ,306, p
= 0,01, R2 = 0,071). When excluding corporations, the significance and beta increase (β =
,786, p < 0,01, R2 = 0,071). This would indicate the amount of tweets about government
ministries’ press releases to be more impactful on the amount of articles published compared to the amount of tweets about corporation’s press releases. With an unstandardized b = ,010, for every tweet the total number of news articles increases with 0,010. Where the tweets of organizations show to decrease the amount news articles about the information from a press release, the amount of tweets of twitter users shows the opposite. The more tweets about a press release’s information, the more news articles about the information from a press release are published. Based on unstandardized b (b = ,010), for a hundred tweets the amount news articles would increase by one. Furthermore for government ministries, the total amount of favourites decreases the amount of news articles about press releases (β = -,540, p < 0,01, R2
= ,168). With an unstandardized b = -,006, for every favourite on a tweet about the press release the number of news articles decreases with 0,006.
For hypothesis 2, while organizations do not tweet extensively regarding their press releases, overall results indicate a significant influence of Twitter’s attention regarding a press release’s information on churnalism of information from press releases in news media. For the use of Twitter by a government ministry or corporation regarding press releases, the number of tweets by government ministries (p = ,61) or corporations (p = ,18) do not significantly increase text similarity between press releases and news articles. With this result, no support for hypothesis 2 is found. Despite this insignificance, a weak correlation is found between the
Table 1
Effects on total number of news articles
F df β t Sig. R2 Npress releases Overall model 2,974 5, 194 1,560 4,698 .013 ,071 200 Tweets by CEO (government ministries) ,001 ,087 ,015 ,0864 ,988 ,389 Tweets by organization (government ministries) -,126 -,168 -1,685 -1,659 ,094** ,099** Total tweets (government ministries) ,306 ,786 2,576 4,479 ,011* ,000* Favourites (government ministries) -,075 -,540 -,650 -3,055 ,517 ,003*
Average tweet sentiment (government ministries) ,108 ,123 1,546 1,560 ,124 ,121
Note: Dependent variable is the total number of news articles * significant at p < 0,05.
number of tweets by an organisation and text similarity of press releases and news articles (r = ,13, p = ,06).
For hypothesis 3, for the attention on Twitter regarding a press release’s information significant results are found. Overall, the total number of tweets about a press release’s information (β = ,350; t(193) = 1,83, p = ,07) and the average sentiment of tweets about a press release’s information ((β = ,120; t(193) = 1,71, p = ,09) are marginally significant and to that end tentative predictors of increasing average text similarity between press releases and news articles (R2 = ,072). For government ministries only, the significance and beta for the effect of total number of tweet about a press release’s information on average text similarity increases (β = ,831; t(193) = 2,430, p = ,016), indicating this effect to be stronger for government ministries. The average cosine text similarity score increases by 0,001 for every tweet (b = ,001) and 0,118 for every sentiment score increased by one (b =,118). This results on increasing average text similarity do indicate a thousand or more tweets about a press release to result in full text similarity. This logic does not stand, as the text similarity score cannot exceed 1 (full text similarity). However, it is intriguing to know if there would be a tipping point would be for the total amount of tweets to stop increasing the text
similarities. From the dataset, no valid tipping point could be determined as the maximum of tweets about a press release did not exceed over 931 tweets for a press release and therefor no valid tipping point emerges (see Figure 1).
When excluding corporations, the total amount of favourites of tweets about a press release’s information is also a significant predictor of decreasing text similarity (β = -,441;
t(143) = -1,99, p = ,05). This would imply the more favourites a tweet about a press release’s
information from a government ministry receives, the more it decreases the text similarity between a press release and a news article about it. A separate linear regression on the effect
Figure 1
Average cosine text similarity by total number of tweets
of the amount of favourites on the average text similarity score shows no significance (p = ,13).
When excluding corporations, the total number of tweets about a press release’s information (β = ,831; t(143) = 2,43, p = ,02) and the average sentiment of tweets about a press release’s information ((β = ,169; t(143) = 2,087, p = ,04) are significant predictors of increasing average text similarity between press releases and news articles (R2 = ,121). Differences between original tweets and retweets are found. The number of retweets do significantly increase text similarity (β = ,201; t(197) = 2,33, p = ,02, R2 = ,05), whereas the number of original tweets do not (p = ,76). This would suggest retweets to be more likely to reach a journalist compared to original tweets on Twitter. With the total number of tweets, average sentiment scores and amount of retweets significantly increasing text similarity scores, these results show support for hypothesis 3. Within these results, no significant
effect on churnalism (p = ,47).
Analysis for hypothesis 4 through linear regression (p = ,39) and a t-test (p = ,41) show no significant results. Possible differences between the influence of tweets by the organization and tweets by the CEO of the organization on the average cosine similarity are not present in this study. For all 200 press releases, only nine tweets were posted by CEO’s. The assumption of a tweet being more personified with a CEO as the sender remains to be justified. For hypothesis 4 no support is found.
Table 2
Effects on average cosine text similarity
F df β t Sig. R2 Npress releases Overall model 2,488 6, 193 ,276 9,272 ,000* ,072 200 Tweets by CEO (government ministries) ,054 ,115 ,760 1,103 ,448 ,272 Tweets by organization (government ministries) ,078 -,053 1,045 -,504 ,297 ,615 Original tweets (government ministries) -,152 -,221 -1,075 -1,097 ,284 ,274 Total tweets (government ministries) ,350 ,831 1,831 2,430 ,069** ,016* Favourites (government ministries) -,038 -,441 -,314 -1,992 ,754 ,048*
Average tweet sentiment
(government ministries) ,120 ,169 1,708 2,087 ,089** ,039*
Note: Dependent variable is the average cosine text similarity *significant at p < 0,05.
For hypothesis 5, ANOVA analysis indicates that the average sentiment scores per press release show significant differences in their text similarity scores (F (2, 197) = 10,243, p < ,01). From the post-hoc tests the following results emerge; the average text similarity score differs significantly between both press releases with neutral and positive average sentiment scores (p < ,01) and press releases with neutral and negative average sentiment scores (p < ,01). The average text similarity does not differ significantly between press releases with positive and negative sentiment scores (p = ,92). As there are no differences between positive and negative average sentiment scores in their effect on text similarity, no support for
hypothesis 5 is found. Although hypothesis 5 finds no support, the result of both negative and positive sentiment to be influence average text similarity scores is of interest. As former research finds humans to react more strongly to negative compared to positive news content (Soroka & McAdams, 2015), this result indicates this negativity bias not to be present specifically for tweets about press release’s information. It shows the more extreme a
sentiment score of a tweet, both negative and positive, to be more impactful on the way text of press releases are used in news articles. This is a intriguing result, for former research
indicates only negative news to elicit more reactions of people (Soroka &McAdams, 2015).
DISCUSSION & CONCLUSION
This study tries to gain more in depth knowledge firstly to what degree the use of Twitter can be considered a factor for news value and secondly to what degree the use of Twitter
influences churnalism of text from press releases in news articles. By conducting a quantitative content analysis, results find that several elements from the use of Twitter do have a significant effect on text similarity and the number of news articles regarding press releases.
Firstly, the total number of news articles regarding a press release is positively affected by the total number of tweets regarding a press release preceding the publication of news articles. This shows a significant role of Twitter in gaining attention by news media, which aligns with the assumption that (Dutch) journalists are influenced by Twitter in their news production. This also aligns with former research (Pavlik, 2000; Gleason, 2010; Ahmad, 2010; Arketi Group, 2011). From the total of tweets regarding a press release, the number of tweets from the organization itself shows only to have a marginal negative effect on the total number of news articles regarding the press release. Differences between affecting the amount news articles in print, online or articles from press agencies are not to be found in this
research. These findings suggest attention on Twitter (i.e. total of tweets and its sentiment) to be a solid pre-indicator of news media attention for information from press releases. In future research on measuring public relations efforts, this role of social media attention as a factor impacting the news value of press releases should be taken into account. Despite this result, the use of Twitter (i.e. number of tweets) by organizations publishing the press releases shows to not increase and only marginally decrease the attention of news media for information from press releases. Arguably, this indicates the included Dutch news media to be rather
uninfluenced by the use of Twitter by the organizations themselves, but influenced by overall attention on Twitter regarding a press releases information. The role and influence of public relations and/or PR practitioners can therefor be argued to be minimal in regards to efforts via Twitter.
Secondly, text similarity between press releases and news articles is positively affected by the total number of tweets and the average sentiments scores. The assumption of only negative sentiment having an impact on text similarity seems to be false. No differences are to be found between positive and negative sentiments in their effect on text similarity, but both do affect text similarity significantly compared to neutral sentiment, which does not. From
this can be derived the more extreme the sentiment (both negative and positive) in attention on Twitter is, the more text similarity between texts of press releases and news articles follows. For public relation efforts, this would imply the kind of sentiment in attention received on Twitter not to be of interest for effectiveness of press releases. The intensity of sentiment shows to be of interest, regardless of the direction of sentiment (i.e. positive of negative). The positive effect of total number of tweets, i.e. the attention on Twitter regarding a press release’s information, on text similarity corresponds to the effect it has on total of news articles. These findings show the attention on Twitter to information from press releases to result in higher text similarity between a press release and news articles. It shows this attention to be a significant factor for news media in their news production. While higher attention on Twitter results in increased text similarity, organizations themselves do not have a significant impact with their efforts by using Twitter regarding press releases.
This research focusses mainly on Dutch government ministries. By including only one Dutch corporation (Royal Dutch Shell), this research limits its explanatory power for
corporations and their efforts of Twitter. Future research can be extended by including multiple corporations to extend insight into the impact of their public relation efforts via Twitter. The method of combining words from the title of a press release has its implications. Cases of combining prepositions, verbs and/or other noun could indirectly show implications of exclude tweets and news articles not including the specific combination from the search query. However this method does result in a strict sample, reinforcing reliability of the data. By only taking into account the use of Twitter, this research limits itself to one social media platform. Expectations for other social media platforms would include overall attention to posts to increase the impact of the content of the post (e.g. a press release). Nevertheless, as social media platforms differ in form, the explanatory power of this research only serves as indication for these other social media platforms. Future research into other platforms with
more imagery, such as Instagram, or specific networks, such as Facebook or LinkedIn, could explain more on the role of social media in relation to its impact on the effectiveness of press releases.
This research paints a picture in which news media are more likely to copy
information from press releases when the attention of Twitter users regarding this information is high. It shows the degree of journalistic churnalism among other to be dependent on
attention of Twitter users. While PR practitioners seem to be concerned regarding journalistic churnalism, wanting to see ‘balanced journalism’ (Jackson & Malony, 2016), this research shows the efforts of organizational use of Twitter only to have marginal impact on the amount of news articles published by news media. The use of Twitter by organizations does not have significant impact on text similarity in this painted picture. This leads to the agenda building capacities of organizations to be of marginally increased by using Twitter as a tool to push press releases. Government ministries and corporations do not have significant differences in their agenda building capacities through Twitter, according to this study. Although
differences between government ministries and corporations are not to be significant in this research, results however do give an interesting indication. Government ministries tend to show higher significance in the impact of its received attention on Twitter (i.e. amount of tweets about a press releases information) on both news media coverage (i.e. amount of news articles published) and churnalism (i.e. average text similarity). For future research, this indication should be taken into account.
In order to increase impact of a press release, this research shows for organizations using Twitter for public relations efforts to put their efforts into eliciting responses with sentiment, regardless of the direction of sentiment, to their press releases instead of posting more tweets themselves. For this research shows Dutch journalism to be only marginally impacted by tweets of organizations themselves and to be influenced more by the focus of
Twitter user’s attention. Showing the attention of Twitter users to be of influence on news coverage, this study argues a social media platform like Twitter to be a pre-indicating factor for news media coverage, regardless the direction of sentiment thereof.
APPENDIX List of references
Abele, A. (1985). Thinking about thinking: Causal, evaluative, and finalistic cognitions about
social situations. European Journal of Social Psychology, 15, 315–332.
Ahmad, A. N. (2010). Is Twitter a useful tool for journalists? Journal of Media Practice, 11, 145–155.
Arketi Group. (2011). How do BtoB journalists use the Web? Retrieved from http://www.arketi.com/surveys.html
Berkowitz, D. & Adams, D.B. (1990). Information subsidy and agenda-building in local television news. Journalism & Mass Communication, 67(4), 723-731.
Boukes, M., & Vliegenthart, R. (2017). A general pattern in the construction of economic newsworthiness? Analyzing news factors in popular, quality, regional, and financial
newspapers. Journalism, 1464884917725989.
Boumans, J. (2018). Subsidizing The News? Journalism Studies, 19:15, 2264-2282, DOI: 10.1080/1461670X.2017.1338154
Brainard, L.A., & Mcnutt, J.G., (2010). Virtual Government–Citizen Relations:
Informational, Transactional, or Collaborative? Administration & Society 42(7):
836–858. DOI: 10.1177/0095399710386308
Carroll, C. E., & McCombs, M. (2003). Agenda-setting effects of business news on the public's images and opinions about major corporations. Corporate reputation
review, 6(1), 36-46.
Cision & George Washington University. (2009). Social media & online usage study. Retrieved from http://www.gwu.edu/ newsctr/10/pdfs/gw_cision_sm_study_09. Davis, A. (2008). “Public relations in the news”. In Pulling newspapers apart, edited by Bob Franklin, 256-264. Abingdon: Routledge.
Eilders, C. (2006). News factors and news decisions. Theoretical and methodological advances in Germany. Communications, 31(1), 5-24.
New Media, Old News, Journalism and Democracy in the Digital Age, edited by
Natalie Fenton, 153–169. London: Sage Publications.
Fiske, S. T. (1980). Attention and weight in person perception: The impact of negative and
extreme behavior. Journal of Personality and Social Psychology, 38, 889–906.
Gleason, S. (2010). Harnessing social media: News outlets are assigning staffers to focus on
networking. American Journalism Review, 32(1), 6-8.
Hermans, L., Vergeer, M., & d’Haenens, L. (2009). Internet in the daily life of journalists:
Explaining the use of the Internet by work-related characteristics and professional opinions. Journal of Computer-Mediated Communication, 15, 138–157.
Hijmans, E., Schafraad, P. & Buijs, K. (2009). Nieuwsbronnen en de kwaliteit van de
journalistiek. Een verkennende analyse van binnenlandse nieuwsonderwerpen in vier
Nederlandse dagbladen. In B. Ummelen (ed.), Journalistiek in diskrediet. (pp. 41-66).
Diemen: Uitgeverij AMB.
Hijmans, E., Schafraad, P., Buijs, K. & d’Haenens, L. (2011). Wie schrijft ons nieuws? Een
verkennende analyse van binnenlandse nieuwsonderwerpen in Vlaamse en
Nederlandse dagbladen. Tijdschrift voor Communicatiewetenschap 39(2), 77-91.
Jackson, D. & Moloney, K. (2016). Inside Churnalism, Journalism Studies, 17(6), 763-780, DOI: 10.1080/1461670X.2015.1017597
Kim, J. Y., Kiousis, S., & Xiang, Z. (2015). Agenda building and agenda setting in business:
Corporate reputation attributes. Corporate Reputation Review, 18, 25-36.
DOI:10.1057/ crr.2014.18
Kiousis, S., Popescu, C., & Mitrook, M. (2007). Understanding influence on corporate
reputation: An examination of public relations efforts, media coverage, public opinion,
and financial performance from an agenda-building and agenda-setting perspective. Journal of Public Relations Research, 19(2), 147-165.
Kroon, A. C., & van der Meer, T. G. (2018). Who Takes the Lead? Investigating the
Reciprocal Relationship Between Organizational and News Agendas. Communication Research, DOI:0093650217751733.
Kroon, A., & Schafraad, P. (2013). Copy-paste of journalistieke verdieping?. Tijdschrift voor
Communicatiewetenschap, 41(3), 283.
Ksiazek, T. B. (2016). Commenting on the news. Journalism Studies, 19, 650-673. DOI:10.1080/1461670X.2016.1209977
Livingston, S., & Bennett, W. L. (2003). Gatekeeping, indexing, and live-event news: Is
technology altering the construction of news? Political Communication, 20, 363–
380.
McCombs, M. and Shaw, D. (1972). The agenda- setting function of mass media. Public Opinion Quarterly, 36, 176–187.
Middleberg/Society for New Communications Research. (2011, November 4). How are media and journalism evolving? Retrieved from http://sncr.org/sites/dfault/files/files/SNCR Middleberg_SNCRsymposium_11_4_11.pdf
Moon, S.J. & Hadley, P. (2014). Routinizing a New Technology in the Newsroom: Twitter
as a News Source in Mainstream Media, Journal of Broadcasting & Electronic Media, 58(2), 289-305, DOI: 10.1080/08838151.2014.906435
Moon, S. J., & Hyun, K. D. (2014). Online media relations as an information subsidy: Quality
of Fortune 500 companies’ websites and relationships to media salience. Mass Communication and Society, 17, 258-273. DOI:10.1080/15205436.2013.779716
Newman, N., Fletcher, R., Kalogeropoulos, A., Levy, D. A. L., & Nielsen, R. K. (2017). Reuters Institute digital news report 2017. Retrieved from
https://reutersinstitute.politics.ox.ac.uk
NOBO, 2019-I. Retrieved from https://vinex.nl/nobo-top-50-jan-2019
Pavlik, J. (2000). The impact of technology on journalism. Journalism Studies, 1, 229–237. Pew Research Center. (2017). News use across social media platforms 2017. Available from
http://www.journalism.org
Ragas, M. W., Kim, J., & Kiousis, S. (2011). Agenda-building in the corporate sphere:
Analyzing influence in the 2008 Yahoo!–Icahn proxy contest. Public Relations Review, 37, 257-265. doi:10.1016/j.pubrev.2011.03.012
Schafraad, P., Van Zoonen, W. & Verhoeven, P. (2015) The news value of Dutch corporate press releases as a predictor of corporate agenda building power. Public Relations Review 42(3).
Scholten, O. & Ruigrok, N. (2009). Bronnen in het nieuws. Een onderzoek naar brongebruik
in Nederlandse dagbladen 2006-2008. Amsterdam: Nederlandse Nieuwsmonitor.
Schultz, F., Kleinnijenhuis, J., Oegema, D., Utz, S., & Van Atteveldt, W. (2012). Strategic framing in the BP crisis: A semantic network analysis of associative frames. Public
Relations Review, 38, 97-107. DOI:10.1016/j.pubrev.2011.08.003
Sandoval-Almazan, R. & Gil-Garcia, Jr. (2012). Are Government Internet Portals Evolving Towards More Interaction, Participation, and Collaboration? Revisiting the Rhetoric of
e-Government Among Municipalities. Government Information Quarterly 29(1): S72–
S81. DOI:10.1016/j.giq.2011.09.004
Seletzky, M. &Lehman-Wilzig, S. (2010). Factors Underlying Organizations' Successful Press Release Publication in Newspapers: Additional PR Elements for the Evolving “Press
Agentry” and “Public Information” Models, International Journal of Strategic
Communication, 4(4), 244-266, DOI: 10.1080/1553118X.2010.515538
Shell Annual Report and Form 20-F 2018. Retrieved from https://www.shell.com/investors/financial- reporting/annual-publications.html
Soroka, S. & McAdams, S. (2015). News, Politics, and Negativity, Political Communication,
32(1), 1-22, DOI: 10.1080/10584609.2014.881942
Strömbäck, J. (2008). Four Phases of Mediatization: An Analysis of the Mediatization of
Politics. The International Journal of Press/Politics, 13, DOI:
10.1177/1940161208319097
Stroud, N. J., van Duyn, E., & Peacock, C. (2016). News commenters and news comment readers. Retrieved from
https://mediaengagement.org/wp-content/uploads/2016/03/ENP-News-Commenters-and-Comment-Readers1.pdf Turk, J. V., & Franklin, B. (1987). Information subsidies: Agenda-setting traditions. Public
Relations Review, 13(4), 29-41.
Van Leuven, S., & Joye, S., (2014). Civil Society Organizations at the Gates? A Gatekeeping
Study of News Making Efforts by NGOs and Government Institutions. The
International Journal of Press/Politics 19 (2): 160–180.
Verhoeven, P. (2008). Mediatization of organizations theory. In W. Donsbach (Ed.), The international encyclopedia of communication: Vol. VII (pp. 3044-3047). Malden, MA: Blackwell.
Waters, R.D., Tindall, N.T.J. & Morton, ST., (2010). Media Catching and the Journalist– Public Relations Practitioner Relationship: How Social Media are Changing the
