Mapping the Issue Arena of Plastic Soup
Applying the Network Agenda-Setting (NAS) Model in Big-Data Research
Master’s Thesis Author: Louelle Pesurnaij Student number: 10750886
University of Amsterdam Graduate School of Communication Master’s Programme Communication Science
Supervisor: dr. I.R. (Iina) Hellsten Date of completion: 28-06-2019
Acknowledgements
Throughout the writing of this thesis I have received a great deal of assistance, support and encouragement. First of all, I would particularly like to thank my supervisor, dr. I.R. Hellsten of the Graduate School of Communication/Faculty of Social and Behavioural Sciences at the University of Amsterdam for always steering me in the right direction whenever I needed help with writing my research, while continuously allowing me to write this dissertation as being my personal work. Her guidance and expertise (especially with regard to network analysis) have been invaluable for conducting my research and further my thesis.
I would like to recognize and thank the Digital Communication Methods Lab of the Amsterdam School of Communication Research (ASCoR) at the University of Amsterdam, and especially co-director of the lab dr. Theo Aruajo, for granting me (financial) support through the Thesis Funding Grants which allowed me to spend the best of my valuable time on conducting my thesis.
Also, I would like to thank OBI4wan for all of the opportunities and for providing me with the tools that I needed to successfully conduct my research.
Finally, I must express my deep gratitude to my parents, my friends and my boyfriend. You supported and encouraged me throughout all the years of studying at the University of Amsterdam and through the process of writing my thesis. Thank you.
Abstract
This paper empirically tested an emerging agenda-setting theory—the Network Agenda-Setting (NAS) model—by analyzing large datasets on Twitter through computer assisted methods. Traditional agenda-setting research asserts that media have the ability to influence the salience of topics and topic attributes on the public agenda. The NAS model extends this view by asserting that topics and attributes are not just transferred as individual elements but transferred as interconnected bundles from the media agenda to the public agenda. The aim of this study was to expand the NAS model by examining the network attribute agendas of the media, non-governmental organizations (NGOs) and the public from an issue arena
perspective. While agenda-setting theory is media-centered, the issue arena approach asserts that issue arenas (i.e., platforms where stakeholders discuss issues) are not media- nor NGO-centered, but issue-centered. This study examined to what extent attributes of the Plastic Soup issue are transferred from the attribute networks of the media and NGOs’ agendas to the public agenda on Twitter in the Netherlands. Empirically, this study found that both media and NGOs have the ability to set the public agenda and dictate how the public links attributes of Plastic Soup. Additionally, the results suggest that NGOs have a stronger agenda setting power than news media. The findings support the results of earlier research on the NAS model that media can influence how the public links different attributes and, furthermore, this study expanded the scope of previous NAS research by examining the agenda-setting power of both media and NGOs. The NAS model provides insights into how public opinion can be influenced by linking different attributes of an issue on social media, which is relevant for both the academic literature and the practice of corporate communications, PR and
journalism. The findings of this study can help practitioners to effectively design campaign messages or news articles, by frequently linking different attributes of environmental issues such as Plastic Soup in their communications to influence public opinion on those issues.
Introduction
Social media platforms such as Twitter are becoming more important for communication between stakeholders (Hellsten, Jacobs, & Wonneberger, 2019). Especially regarding societal and environmental issues, social media platforms such as Twitter have proven useful for communication among all parties involved, because information is disseminated rapidly and Twitter allows actors (e.g., individuals, organizations) to interact and discuss the issues (Getchell & Sellnow, 2016).
One important, global issue that has gained public concern in the past decades and involves many stakeholders is the issue of plastic marine pollution (Dauvergne, 2018). In the past decades, more and more plastic floats in the oceans and seas (the so-called “Plastic Soup”) as a result of disposed plastic products, such as food packaging, fishnets, synthetic clothing, toothbrushes and plastic furniture. This non-biodegradable pollution is lethal to marine life, enters the food chain and damages human’s health (Dauvergne, 2018; Plastic Soup Foundation, n.d.).
Communication about issues such as Plastic Soup generally take place on so-called “issue arenas”, (social media) platforms where societal and environmental issues are discussed by and among stakeholders (Luoma-aho & Vos, 2010). The issue arena approach (Luoma-aho & Vos, 2010) implies that organization-centered views of stakeholders are becoming outdated and highlights the importance that with environmental issues such as Plastic Soup stakeholders are viewed from an issue-centered perspective. The issue arena perspective hence states that not the immediate stakeholders of organizations (e.g., corporations, NGOs, the media) matter but rather the stakeholders surrounding specific (environmental) issues (Luoma-aho & Vos, 2010).
The dissemination of information regarding an environmental issue such as Plastic Soup starts, according to agenda-setting theory (McCombs & Shaw, 1972), at the media who
put certain issues on the public or political agenda which are deemed worthy of discussion (Carroll & McCombs, 2003). Contradictory to the issue arena perspective, which proposes that issues arise as they are discussed by and among stakeholders (Luoma-aho & Vos, 2010), traditional agenda-setting research (McCombs & Shaw, 1972) assumes that media have the ability to raise awareness for specific issues, put these issues on the public agenda and initiate public debate. The media hence dictates what the public speaks about and how they speak about it.
Recently, the Network Agenda Setting (NAS) model was introduced (Guo, 2012), an emerging agenda-setting theory that has been researched in traditional and social media. The NAS model extends traditional agenda-setting theories by theorizing that news media also have the ability to link issues and their attributes (i.e., detail and describe the given objects) and transfer these associations simultaneously in bundles onto the public agenda (Vu, Guo, & McCombs, 2014). Previous research on the NAS model (e.g., Guo & Vargo, 2015)
additionally showed that even in online issue arenas, media retain their agenda-setting power. However, not only media have the ability to influence public opinion. Previous research on (environmental) issues (e.g., Burchell & Cook, 2013) has shown that campaigning
activities of non-governmental organizations (NGOs) have largely increased public awareness and concern on social, ethical and environmental impacts (Burchell & Cook, 2013). NGOs have proven to be powerful at setting the public agenda by urging the public towards a pro-environmental stance (Cialdini et al., 2006; Dobele, Westberg, Steel, & Flowers, 2014).
The aim of the present thesis study is to gain a more complete picture of the agenda-setting roles of the media and NGOs in the Plastic Soup issue arena on Twitter. More specifically, this study seeks to apply the NAS model in order to research how information regarding Plastic Soup is disseminated on Twitter and to what extent attributes of the Plastic
Soup issue are transferred from the media network attribute agenda and the NGO network attribute agenda to the public network attribute agenda, in the Netherlands. The research question of this thesis study is as follows: “To what extent are the attribute networks transferred from the media and NGOs agendas onto the public agenda?”.
In this thesis study, an empirical research is conducted which analyses and enriches the NAS model. The NAS model has been largely researched in areas such as political
communication, public relations, and public opinion research (e.g., Vargo, Guo, McCombs, & Shaw, 2014; Vu et al., 2014). The present study seeks to apply the NAS model in a
corporate communication context and showcase the model its potential contribution to theory in the field of sustainability/CSR and issue arenas. This study offers insights into the
interplay between the news media, NGOs and the public and clarifies the agenda-setting phenomenon. Applying the NAS model enables a more nuanced representation of online information networks between the news media, NGOs and the public, by analyzing how the media/NGOs link different attributes of Plastic Soup and in turn influence how the public links those different attributes.
From an organizational perspective, the research outcomes can help practitioners finding the relevant issues and attributes, issue arenas and the network agendas for interaction, facilitating the organization-public debate and through this managing organizational reputation (Luoma-aho & Vos, 2010, p. 315). Additionally, the network concept of the NAS model provides practitioners (both communication professionals and journalists) the right tools to effectively design campaign messages or news articles by linking different attributes, to establish the ownership of new and/or existing attributes and influence how the public links different attributes.
Theoretical Framework Agenda-setting theory
The agenda-setting theory was introduced in the late ‘90s by McCombs and Shaw (1972) and has since then been extensively researched worldwide in the fields of political
communication, PR and corporate communication (e.g., Carroll & McCombs, 2003; McCombs & Shaw, 1972;). Traditional agenda-setting theory was mainly influenced by theories about “the almighty media” and specifically focuses on the successfulness of traditional mass media in telling its publics what to think about. Traditional agenda-setting research consists of first and second level agenda-setting and the main proposition is that “the prominence of elements in the news influences the prominence of those elements among the public” (Carroll & McCombs, 2003, p.36).
First level of agenda-setting. The first level of agenda-setting is concerned with the prominence (i.e., salience) of topics that are covered by news media and the ability of the media to influence the salience of those topics on the public agenda (Carroll & McCombs, 2003; McCombs & Shaw, 1972). More specifically, the first level of agenda-setting asserts that the salience of topics is transferred from the media agenda to the public agenda (Carroll & McCombs, 2003; Guo, 2012). Topics in agenda-setting research could refer to
issues, public figures, institutions or corporations. For example, Vargo et al. (2014) examined agenda-setting effects of news media during the 2012 U.S. presidential election and identified the 16 most salient issues of the election (i.e., taxes, unemployment, economy, international relations, border issues, health care, public order, civil liberties, environment, education, domestic politics, poverty, disaster, religion, infrastructure, and media and internet).
Second level of agenda-setting. The second level of agenda-setting is concerned with the prominence (i.e., salience) of the attributes of topics. Attributes describe a given topic (Carroll & McCombs, 2003; Vargo & Guo, 2017). Every single topic on the media agenda
has an agenda of attributes and the second level of agenda-setting asserts that the news media agenda thus not only influences the salience of topics on the public agenda, but also
influences the salience of attributes of those topics on the public agenda (Carroll &
McCombs, 2003). In other words, news media not only tell us what to think about, but also how to think about it. According to Carroll and McCombs (2003) attributes can be described in terms of two dimensions: cognitive and affective. For example, in the study by Guo and McCombs (2011) the various traits of political candidates are examined. The specific traits (e.g., leadership, experience, competence and credibility) are cognitive attributes. In turn, each of these cognitive attributes has an affective dimension (e.g., positive, neutral, negative, and mixed), which recognizes that besides sharing information also feeling and tone is conveyed (Carroll & McCombs, 2003).
Third level of agenda-setting. Traditional agenda-setting studies have proven that media are effective in transferring the prominence of individual elements, both topics (i.e., first level of agenda-setting) and attributes (i.e., second level of agenda-setting), to audiences (Vargo et al., 2014). The NAS model (i.e., third level of agenda-setting) extends this view by asserting that the more often media link topics and/or attributes in news coverage, the more likely it is that the public will also link those topics and/or attributes (Vu, Guo, & McCombs, 2014).
Vargo et al. (2014) argue that individuals associate elements in their minds to make sense of social realities. The NAS model states that news media play a fundamental role in how the public links these elements and asserts that media have the power to build new associations, as well as to enhance already existing ones, and dictate the way the public links different objects and/or attributes in their minds (Guo & Vargo, 2015). Besides telling the public what to think about and how to think about it, media are also capable of telling us what and how to associate.
The issue arena perspective
Agenda-setting theory (e.g., Carroll & McCombs, 2003; McCombs & Shaw, 1972; Vargo et al., 2014) asserts that news media have control over communication and thereby have the power to influence the public. However, with the rise of new communication technologies, organizations (such as the media, non-governmental organizations and corporations) are losing grip on communication. Contradictory to agenda-setting theory, the issue arena perspective (Luoma-aho & Vos, 2010) asserts that communication with and among stakeholders are moving outside of organizations’ control. Social media platforms enable stakeholders to express their opinions to a wider public and the places where these public debates about these issues take place are referred to as issue arenas (Luoma-aho & Vos, 2010).
The issue arena perspective (Luoma-aho & Vos, 2010) was developed because traditional ways of thinking about stakeholders were becoming outdated with the rapidly changing media landscape. Traditionally, stakeholder theories (e.g., Freeman, 1984) were developed to map the organizational context and identify organizations’ constituencies but with the rise of a multiplicity of social media platforms and increased communication by and among stakeholders the idea of issue arenas was introduced (Luoma-aho & Vos, 2010). According to Luoma-aho and Vos (2010, p, 316): “nowadays, issues and topics, not
organizations that are at the center of communication”. Issue arenas are not media-centered nor organization-centered, but they are issue-centered platforms constituted by
communication networks between all actors regarding (societal) issues (Hellsten et al., 2019; Luoma-aho & Vos, 2010). The most important dynamic in issue arenas is that stakeholders and organizations express their opinions and interact/discuss ideas and issues among each other. According to Luoma-aho and Vos (2010), all actors in an issue arena have their own agendas and strategies.
The media agenda in issue arenas. In contrast to their owned (online and offline) channels, media do not have absolute power or autonomy on social media. Issue arenas offer stakeholders the possibility to put topics on the agenda, to build constituencies and reach a wider audience. To reclaim their control over communication and influence the public debate, media participate in issue arenas and use agenda-setting to call attention to specific (new) issues (Carroll & McCombs, 2003; Luoma-aho & Vos, 2010; Pallas & Ihlen, 2014). Even though issue arenas are issue-centered and not media-centered, previous agenda-setting studies that applied the NAS model (e.g., Guo & Vargo, 2015; Vargo et al., 2014) have proven that news media retain their strong agenda-setting effects on social media platforms. The more frequently news media associate two (or more) topics on social media, the more likely it is that the audience considers the two (or more) topics as interconnected (Guo, 2012).
The NGO agenda in issue arenas. In (online) issue arenas, stakeholders have the possibility to call attention to issues, to initiate societal change and to mobilize other actors (Hellsten et al., 2019; Meriläinen & Vos, 2011). One particular group of stakeholders, namely stakewatchers (i.e., environmental groups, pressure groups, activists), participate in issue arenas to watch the stake of the community or the environment with care, attention and scrutiny (Fassin, 2009). Non-governmental organizations (NGOs) are “private, non-profit, professional organizations, with a distinctive legal character, concerned with public welfare goals” (Clarke, 1998, p. 36) and often take the role of stakewatchers in issue arenas (Dobele et al., 2014; Fassin, 2009). Because NGOs are non-profit organizations, they are highly dependent on the public and the media in order to influence policy makers. Thus, the agenda of NGOs in issue arenas is mostly focused on motivating the public to participate in their campaigns and sign petitions, to “enhance their value in the eyes of decision makers” (Meriläinen & Vos, 2011, p. 306).
In traditional agenda-setting research (McCombs & Shaw, 1972) it was assumed that the media set the public agenda and journalists would primarily use NGOs as sources of information (Meriläinen & Vos, 2011) but with the rise of social media platforms NGOs became independent actors (Burchell & Cook, 2013) and were given the opportunity to directly interact with the public and reach a wider audience (Luxon & Wong, 2017; Meriläinen & Vos, 2011). Nowadays, NGOs have great power and generally dominate the issue arenas in new media environments (Luoma-aho & Vos, 2010). NGOs successfully use online communication to mobilize the public to participate in (online) debates and to take part in grassroot activities such as environmental activism (Luxon & Wong, 2017; Meriläinen & Vos, 2011).
Yang and Saffer (2018) found that especially for NGOs, well-established online relationships with stakeholders are a powerful predictor for NGOs’ prominence in social media conversations. When NGOs construct online relationships with stakeholders and engage with the public, it significantlyinfluenced both social media conversations and media coverage. The public is likely to mention NGOs or share their messages on social media, which allows the message of NGOs to “travel through these stakeholders’ social networks and reach more audiences” (Yang & Saffer, 2018, p. 435). These social media conversations in turn influence the news media agenda, who often adopt trends on social media (Yang & Saffer, 2018).
In sum, it is assumed that NGOs have the ability to directly set the public agenda by engaging with stakeholders online and through these stakeholders’ networks NGOs gain an influential position to build the media agenda who in turn also sets the public agenda.
Therefore, it is expected that not only the media has the power to influence the public agenda, but also NGOs.
In the present study, the agenda-setting role of news media and NGOs in the Dutch Plastic Soup issue arena on Twitter is examined by applying the NAS model. More
specifically, this study aims to research to what extent attributes of the Plastic Soup issue are transferred from the media and NGOs’ network attribute agendas to the public network attribute agenda on Twitter. In order to answer the research question of this study, the following hypotheses have been constructed:
Hypothesis 1 (H1): The media network attribute agenda positively correlates with the
public network attribute agenda regarding Plastic Soup
Hypothesis 2 (H2): The NGO network attribute agenda positively correlates with the
public network attribute agenda regarding Plastic Soup
Because it is assumed that NGOs have both agenda-building and -setting power on social media, it is expected that NGOs have a greater power in setting the public agenda than media do.
Hypothesis 3 (H3): The NGO network attribute agenda regarding Plastic Soup is more
strongly correlated to the public network attribute agenda, than the media network attribute agenda regarding Plastic Soup
Hypothesis 4 (H4): The NGO network attribute agenda regarding Plastic Soup better
predicts the public network attribute agenda regarding, than the media network attribute agenda regarding Plastic Soup
Description of the Plastic Soup issue
The term “Plastic Soup” is worldwide used and was introduced in the mid-1990s, when captain Charles Moore identified floating pieces of plastic in the Pacific Ocean and called this phenomenon the “Plastic Soup” (Plastic Soup Foundation, n.d.). This marine pollution was a result of the increased manufacturing of plastic products in the 1990s (Dauvergne, 2018). In the past two decades, scientists have urged manufacturers to terminate the usage of plastics in their products. Also, a great number of (anti-plastic) grassroot movements and NGOs arose and became powerful actors by fostering public concern and endorsing governmental bans on microplastics (Dauvergne, 2018). Especially since 2015, when the United Nations paid attention to reducing (plastic) waste production in their 2030 Agenda for Sustainable Development (The United Nations, n.d.; Plastic Soup Foundation, 2018), the Plastic Soup issue was acknowledged worldwide and became an important, global affair that involves and interests many different stakeholder groups (Dauvergne, 2018).
The Plastic Soup issue is a suitable case for researching the NAS model from an issue arena perspective, because this approach asserts that issue arenas consist of many
stakeholders discussing the issue and positions online (Luoma-aho & Vos, 2010). The Plastic Soup issue is such an issue that encompasses the involvement from numerous stakeholders (e.g., organizations, politics, the public) and conflicts of interests (Dauvergne, 2018).
The Plastic Soup issue arena on Twitter
In the present study, Twitter has been chosen as the platform for data collection and analyses. Grimmelikhuijsen and Meijer state that Twitter is a popular social medium in the Netherlands (as cited in Hellsten et al., 2019) and known to be used for the dissemination and discussion of information concerning public issues (Albu & Etter, 2016; Haro-de-Rosario, Saez-Martin
& del Carmen Caba-Perez, 2016; Hellsten et al., 2019; Hong, Shin & Kim, 2016; Manetti, Bellucci & Bagnoli, 2016).
Especially from an issue arena perspective, Twitter is a suitable platform for studying the discussion about Plastic Soup because NGOs and news media consider Twitter as a highly relevant platform for communication. The issue arena approach focuses on corporate communications in the era of multiple social media platforms (Luoma-aho & Vos, 2010) and in line with this approach, NGOs widely use Twitter for PR campaigns, the development of new organizational practices and interaction with both internal and external stakeholders (Albu & Etter, 2016; Lovejoy, Waters, & Saxton, 2012; Yang & Saffer, 2018). Similarly, news media use Twitter to disseminate news articles, aiming to set the agendas of the public and politicians (Vargo et al., 2014). The general public thereby often uses Twitter to stay up-to-date (Voorveld, van Noort, Muntinga, & Bronner, 2018), engage in (organization-initiated) discussions and spread news (van den Heerik, van Hooijdonk, Burgers, & Steen, 2017).
Method
The aim of the present study is to research to what extent the attribute networks are transferred from the media and NGOs’ agenda onto the public agenda. Empirically, the present study analyzes tweets about Plastic Soup posted by three groups – news media, NGOs and the public – on Twitter over the year 2018, through quantitative content- and network analyses.
Data collection and sampling
Through the OBI4wan social media monitoring tool, all Dutch tweets between January 2018 and December 2018 in which Plastic Soup, #PlasticSoup, Plastic Soep or #PlasticSoep is mentioned were collected. This one-year period was chosen because it represented a
relatively recent view of the issue. Additionally, this period was chosen because it contained important developments in the fight against Plastic Soup, such as the announcement that the European Parliament introduced a ban on single-use plastics (The European Parliament, 2018) and the launch of the Ocean Cleanup initiative its first (technological) cleanup system into the North Pacific (The Ocean Cleanup, n.d.), which ensured that the topic was often discussed on Twitter.
To decide whether next to original tweets also retweets and replies would be included in this study, a pretest was conducted in which the most frequently retweeted and replied Twitter accounts were examined. The results of the pretest showed that the public mostly replied to and retweeted news media- and NGO-accounts on Twitter. Retweets and replies were thereby considered important messages with regard to agenda-setting. When news media or NGOs tweet about Plastic Soup, replies and retweets of the public were considered an agenda-setting effect because the media/NGOs influence what the public speaks about. Therefore, it was decided to include retweets and replies in the dataset. In total, 32.887 tweets were collected which originated from 14.554 unique Twitter accounts.
This dataset was manually divided into three stakeholder groups: news media, NGOs and the public, based on the Twitter user accounts (i.e., usernames) and Twitter biographies. Regarding news media, the accounts of Dutch TV/radio (both broadcasters, channels and programs), newspapers (both national, regional and local) and online-only news websites appeared in the dataset. With regard to NGOs, both (inter)national and local environmental non-profit organizations were part of the dataset. Besides the corporate Twitter accounts of news media and NGOs, also the accounts of individual employees were allocated to the stakeholder groups: both freelance and employed journalists were assigned to the news media stakeholder group and NGO-employees such as campaign have been assigned to the NGO stakeholder group. The remaining Twitter accounts that did not belong to the news media nor
NGOs’ stakeholder group were considered the public (excluding Dutch governmental- and political accounts). In total, 95.1% of the tweets originated from the public (n = 31.290),
Ethical and privacy considerations
Some considerations with regard to ethics and privacy were taken into account when collecting Twitter data. Under the OECD Privacy Principles (ESOMAR, 2017), researchers are obliged to fairly and lawfully obtain personal data that can be linked to individual persons (e.g., names, address, date of birth or IP address; Wigan & Clarke, 2013). To respectfully and lawfully process personal information the dataset has been anonymized (i.e., identifying data has been removed of pseudonymized). In line with the ethical regulations of the University of Amsterdam (University of Amsterdam 2013, p.17), “personal data that can be traced back to specific people play no role in the analysis process and are not brought into the encryption scheme or the analysis”. Public figures (i.e., people who are recognized by a majority of people in their public role) form an exception to this when they speak on behave of their public function or role. Tweets by or about people who do not fulfil this criterion, or who are acting in a private capacity, are processed anonymously. Finally, security measures in terms of data storage have been executed and data used in the present study was stored on a secured, encrypted ICTS storage.
Identifying attributes
The present study examined the attribute networks of the media, NGO and the public agendas and solely focused on cognitive attributes (i.e., characteristics of the Plastic Soup issue).
The first step into analyzing these attributes, was to collect keywords describing and representing the attributes of the Plastic Soup issue. To get an impression of those most common keywords in tweets about Plastic Soup, lists of the most frequently used words in
those tweets were created for each stakeholder group. These frequency lists were generated through an automated content analysis (e.g., big data analytics) in Python through the open-source web application Jupyter Notebook (see Appendix A). For each group, the 250 words (at maximum) that occurred most often in tweets were selected to represent the attributes of Plastic Soup. The threshold of 250 keywords was chosen to increase the feasibility of the analyses and it was expected that these keywords covered the most prominent topics (i.e., attributes) in tweets regarding Plastic Soup. Thresholds are commonly used in topic modelling and network analyses to increase the feasibility of researching big data sets.
The second step was to examine the keywords and identify attributes. The lists of each stakeholder group, consisting of 250 keywords, were manually examined and filtered (see Appendix B). Only words that directly corresponded to specific attributes (i.e., topics that describe the Plastic Soup issue) were selected and stop words (e.g., “the”, “an”, “it”, “that”, “I”, “we”) were manually removed. Finally, the frequency lists of all three stakeholder groups were merged into one large list of keywords. This final keyword list consisted of 261
keywords in total that were commonly used by at least one of the stakeholder groups. In total fourteen attributes were identified that described either the origin of Plastic Soup, the current state-of-the-art or the fight against Plastic Soup:
A. Fighting plastic pollution (i.e., preventive, cultivating clean-up) B. VAT on plastic products
C. The effect of Plastic Soup on animal welfare
D. Using digitization and innovation to combat Plastic Soup E. Donations to organizations fighting plastic pollution F. Research and information provision about Plastic Soup G. The introduction of deposit money on plastic products
H. Specific waters (e.g., oceans, seas, rivers) polluted by plastic
I. Education about the origin, the effects and the current state of Plastic Soup J. Specific mentions of the origin of plastic pollution
K. Sustainability and environmental impact with regard to Plastic Soup L. Health (i.e., general health issues, water- and food pollution, air quality), M. The mention of specific (plastic) products (i.e., general litter, microplastics,
balloons, confetti, straws, cigarettes, cotton swabs, fishnets, plastic packages and bags)
N. The involvement of stakeholders (i.e., the Dutch government & political parties, Dutch provinces, municipalities & cities, the EU, NGOs, supermarkets, beverage- and food manufacturers, business & retail industry, fishery) in both causing and combating the Plastic Soup issue.
Automated content analysis and manual content analysis
In the previous steps, fourteen attributes of the Plastic Soup issue were identified. Through an automated content analysis (ACA or computer-assisted content analysis), the attributes were placed in attribute construct lists (i.e., search queries, see Appendix C). Using a custom Python-function in Jupyter Notebook (see Appendix D), all tweets that used one of the keywords in the attribute construct lists were flagged. The Python-function hereby
automatedly analyzed the full dataset of tweets with extended search queries and notated the presence of attributes in tweets. It is important to note that the final fourteen attributes are not mutually exclusive, some overlap (i.e., the same keywords) existed between the attributes.
Next to the ACA, a manual content analysis (MCA) was conducted and used as a method to test the validity of the search queries and the ACA. In total, 2.165 random tweets from the total dataset (n=32.887) were manually assessed using a codebook (see Appendix
E). The results of the MCA were compared to the results of the ACA and the results were found to be valid. The ACA and MCA agreed on 2.046 of the 2.157 cases. On average, 94.7% of the MCA-results matched the ACA-results, with no query scoring below 80%. The present study proceeded using ACA because the results suggest this was found to be a valid method for analyzing tweets (Vargo et al., 2014).
Network Analysis
Semantic networks focus on the relationship between words in communication (Hellsten et al., 2019) and were used in the present study for topic modelling: to map and quantify the (inter)relationship among attributes presented in tweets of the media, NGOs and the public (Guo, 2012; Guo & Vargo, 2015). Methodologically, semantic network analyses require matrices that arrange attributes (i.e., nodes) in the rows and columns and represent the (non)directional network links (i.e., ties) between those attributes in the cells (Guo, 2012). In the present study, the data of all three stakeholder groups was arranged as (unweighted) matrices as a preparation for the semantic network analyses (Guo, 2012; Vargo et al., 2014). The rows and columns represent the fourteen attributes and the cells represent the connection between two attributes. As described by Guo (2012, p. 623), the matrices for all three
network agendas are hence constructed according to the interrelationship between the attributes found in tweets. The matrices are based on frequency: the more frequently two attributes (i.e., nodes) co-occur in one tweet, the stronger their connection (i.e., tie).
Using Rstudio (i.e., a software environment for statistical computing and visualizations; The R Foundation, n.d.) the matrices and network visualizations were constructed (see
Appendix F). In the network visualization, the weight of the lines (i.e., ties) between nodes reflects the strength of the connection between two nodes; the thinker the line between two nodes, the stronger the connection (e.g., the more often the two attributes co-occur). The
position of the nodes (more centered or more decentralized) is determined by the “Degree centrality”, a centrality measurement that refers to the number of ties a node has (Wasserman & Faust, 1994). The degree centrality does not correspond to the attribute that is most
frequently covered in tweets, but it corresponds to the attribute that has the most associations with other attributes (Guo, 2012). The more centered the node is presented in the network visualization, the more central the attribute is located in the network.
In order to show to what extent the attributes on the network agendas of the media and NGOs are transferred the public network agendas, Quadratic Assignment Procedure (QAP) correlation and regression tests were conducted to compare the three network agendas and test the hypotheses of this study (Guo, 2012; Guo & Vargo, 2015; Vargo et al., 2014; Vu et al., 2014). QAP tests are permutation tests that control for the non-independence of dyads (Guo, 2012).
First QAP correlation tests are conducted to examine the correlation between (items in) two network matrices: Media–the Public and NGOs–the Public. The QAP correlation thereby “addresses the strength and specification of ties from one network to another and calculates a correlation coefficient” (Vargo et al., 2014, p.306). Using R, two QAP correlation tests were performed to answer H1, H2 and H3.
Finally, QAP regression test were conducted to examine the agenda-setting power of the media- and NGO network attribute agendas (independent matrices) and the public network attribute agenda (dependent matrix) and to answer H4. The QAP regression test thereby examines whether one matrix (or more matrices) —the independent matrix— can predict another —the dependent matrix (Vu et al., 2014, p.676).
Results
In Table 1, the total amount of tweets and unique Twitter accounts are displayed (per stakeholder group).
Table 1. Amount of tweets and unique Twitter accounts
Tweets Unique Twitter accounts
Stakeholder groups n % n %
News media 611 1.9 307 2.1
NGOs 986 3.0 61 0.4
The public 31.290 95.1 14.554 97.5
Total 32.887 100.0 14.922 100.0
The total amount of tweets (per stakeholder group and per attribute) are presented in Table 2. Stakeholder involvement (n=18.616) is the most discussed topic by both news media, NGOs and the public with 56.6% of all tweets, followed by fighting plastic pollution with 40.8% of all tweets (n=13.425) and specific (plastic) products with 37.8% of all tweets (n=12.441). When specifically focusing on stakeholder involvement, the Dutch government and political parties (n=8.720) and NGOs (n=7.024) are the most frequent mentioned
stakeholders with respectively 25.1% and 21.4% of all tweets regarding Plastic Soup. VAT is the least discussed topic and is only mentioned in three out of 32.887 tweets. Strikingly, firework remains as the origin of Plastic Soup (n=115) is the only attribute that is solely raised by the public and has not been discussed by either news media or NGOs.
Table 2. Number of tweets per attribute
Stakeholder groups
News media NGOs The public Total
Attributes n % n % n % n %
A Fighting plastic pollution 258 42,2 474 48,1 12.693 40,6 13.425 40,8
Preventive 158 25,9 347 35,2 9.568 30,6 10.073 30,6
Cultivating clean-up 124 20,3 159 16,1 4.369 14,0 4.652 14,1
B VAT 1 0,2 1 0,1 1 0,0 3 0,0
C Animal welfare 27 4,4 35 3,5 1.666 5,3 1.728 5,3
D Digitization & innovation 8 1,3 24 2,4 533 1,7 565 1,7
E Donations 1 0,2 14 1,4 151 0,5 166 0,5
F Research & information provision 16 2,6 100 10,1 963 3,1 1.079 3,3
G Deposit money 87 14,2 291 29,5 5.706 18,2 6.084 18,5
H Waters 195 31,9 293 29,7 9428 30,1 9.916 30,2
I Education 4 0,7 21 2,1 369 1,2 394 1,2
J The origin of plastic pollution - - 2 0,2 156 0,5 158 0,5
K Sustainability & environmental impact 72 11,8 166 16,8 4.624 14,8 4.862 14,8
L Health 52 8,5 27 2,7 2.292 7,3 2.371 7,2
General health issues 16 2,6 10 1,0 424 1,4 450 1,4
Drink water pollution 12 2,0 6 0,6 105 0,3 123 0,4
Food pollution 14 2,3 6 0,6 210 0,7 230 0,7
Air quality 39 6,4 22 2,2 1.723 5,5 1.784 5,4
M Specific (plastic) products 179 29,3 410 41,6 11.852 37,9 12.441 37,8
Table 2. Number of tweets per attribute (continued) Microplastics 9 1,5 42 4,3 826 2,6 877 2,7 Balloons 46 7,5 9 0,9 2.176 7,0 2.231 6,8 Confetti - - 2 0,2 179 0,6 181 0,6 Straws 19 3,1 6 0,6 631 2,0 656 2,0 Cigarettes - - 6 0,6 27 0,1 33 0,1 Cotton swabs 11 1,8 4 0,4 753 2,4 768 2,3 Fishnets 3 0,5 7 0,7 298 1,0 308 0,9
Plastic packages and bags 71 11,6 145 14,7 4.181 13,4 4.397 13,4
Firework (remains) - - - - 115 0,4 115 0,3
N Stakeholder involvement 327 53,5 702 71,2 17.587 56,2 18.616 56,6 Dutch government & political parties 126 20,6 164 16,6 7.980 25,5 8.270 25,1 Dutch provinces, municipalities & cities 50 8,2 137 13,9 2.577 8,2 2.764 8,4
The EU 37 6,1 62 6,3 2.095 6,7 2.194 6,7
NGOs 145 23,7 505 51,2 6.374 20,4 7.024 21,4
Supermarkets 28 4,6 40 4,1 1.734 5,5 1.802 5,5
Beverage- and food manufacturers 3 0,5 25 2,5 413 1,3 441 1,3 Business & retail industry 6 1,0 27 2,7 528 1,7 561 1,7
Fishery 20 3,3 15 1,5 1.309 4,2 1.344 4,1
Total 611 100 986 100 31.290 100 32.887 100
The Media network attribute agenda
The attributes found in tweets of all three stakeholder groups were arranged as (unweighted) matrices. In Table 3 the matrix of the media network attribute agenda is presented.
Table 3. Matrix of Media Network Attribute Agenda A B C D E F G H I J K L M N A - B - - C 10 - - D 8 - - - E 1 - - - - F 9 - 1 1 - - G 72 - 2 - 1 2 - H 89 - 12 2 - 9 23 - I 1 - 1 - - 3 - 1 - J 258 - - - - K 72 - 3 - - 6 9 32 2 - - L 13 - 4 - - 1 5 25 - - 11 - M 120 - 5 6 - 5 57 49 1 - 25 19 - N 172 1 21 7 1 11 65 112 2 - 51 36 116 -
Note: A = Fighting plastic pollution; B = VAT; C = Animal welfare; D = Digitization & innovation; E = Donations; F = Research & information provision; G = Deposit money; H = Waters; I = Education; J = The origin of plastic pollution; K = Sustainability &
environmental impact; L = Health; M = Specific (plastic) products; N = Stakeholder involvement
For example, the number in the cell corresponding to A and C is 10, which means that the attributes “Fighting plastic pollution” and “Animal welfare” were mentioned together in
10 out of 611 tweets by news media. Similarly, the number in the cell corresponding to A and B is 0, which means that in none of the 611 tweets by news media, the attributes “Fighting plastic pollution” and “VAT” co-occurred. The strongest connection (i.e., tie) is between the attributes (i.e., nodes) “Fighting plastic pollution” and “The origin of plastic pollution”, which are linked in 258 out of 611 tweets by news media.
The attribute “Fighting plastic pollution” is the most central node in the media network attribute agenda, which means that it is the attribute with the strongest associations with other attributes. Besides “Fighting plastic pollution”, other central nodes in the network are:
“Stakeholder involvement”, “Specific (plastic) products”, “Waters” and “The origin of plastic pollution”. These attributes are, next-to being central nodes, often interrelated to each other and therefore form a sub-structure (i.e., a group of nodes composed of more than two
attributes; Guo, 2012). These strong associations are visualized in Figure 1 by the many thick lines (i.e., ties) between “Fighting plastic pollution” and –for example– “The origin of plastic pollution” and “Stakeholder involvement”.
The centrality of the nodes and the thickness of the association lines in Figure 1 clearly demonstrates that the five attributes “Fighting plastic pollution”, “Stakeholder involvement”, “Specific (plastic) products”, “Waters” and “The origin of plastic pollution” are closely connected to one another in describing Plastic Soup. This sub-structure highlights the focus of media on the introduced ban on single-use plastics (e.g., plastic straws, cotton swabs) by the European Parliament and the introduction of a legislation regarding deposit money on plastic bottles and cans by the Dutch government.
The NGO network attribute agenda
Table 4 displays the NGO network attribute agenda matrix and Figure 2 presents the NGO network attribute agenda visualization. Contrary to the media network attribute agenda,
“Stakeholder involvement” is the most central attribute in the network that has the strongest connections with other attributes, followed by “Fighting plastic pollution”, “Specific (plastic) products”, “Deposit money” and “Waters”. In line with the media network attribute agenda, the strongest connection (i.e., tie) in the NGO agenda network is between “Stakeholder involvement” and “Fighting plastic pollution”, which co-occurred in 36.0% (n=355) of all tweets by NGOs (n=986).
Table 4. Matrix of NGO network attribute agenda
A B C D E F G H I J K L M N A - B - - C 14 - - D 24 - 1 - E 4 - - - - F 44 - 11 5 - - G 274 - 12 - 2 22 - H 138 - 10 1 4 55 88 - I 2 - 5 - - 18 1 4 - J 1 - 1 - - - 1 1 - - K 84 1 2 2 1 17 69 59 1 - - L 10 - 5 - - - 4 6 - - 1 - M 286 - 17 20 2 47 191 111 2 2 59 14 - N 355 1 24 23 14 57 236 207 5 2 132 24 289 -
Note: A = Fighting plastic pollution; B = VAT; C = Animal welfare; D = Digitization & innovation; E = Donations; F = Research & information provision; G = Deposit money; H
= Waters; I = Education; J = The origin of plastic pollution; K = Sustainability & environmental impact; L = Health; M = Specific (plastic) products; N = Stakeholder involvement
Similar to the media network attribute agenda, “Fighting plastic pollution”,
“Stakeholder involvement”, “Specific (plastic) products” and “Waters” form a sub-structure. However, “The origin of plastic pollution” –an important node in the media network attribute agenda –is replaced by “Deposit money” in the sub-structure of the NGO network attribute agenda. NGOs thus more strongly focus on (solutions for) combating the Plastic Soup issue and thereby more often focus on the involvement of stakeholders than media do.
The strong connections between all these attributes (“Fighting plastic pollution”, “Stakeholder involvement”, “Specific (plastic) products”, “Deposit money” and “Waters”) highlights the strong focus of NGOs on the possible introduction of new legislation (by the Dutch government & political parties) regarding deposit money on litter, plastic bottles and cans.
The Public network attribute agenda
Contrary to the media network attribute agenda but in line with the NGO network attribute agenda, “Stakeholder involvement” is the attribute that is most central in the public network attribute agenda (i.e., the attribute with the strongest ties to other attributes), followed by “Fighting plastic pollution”, “Specific (plastic) products”, “Waters” and “Deposit money”. As shown in Table 5 and Figure 3 and in line with the media and NGO attribute agenda, “Stakeholder involvement” and “Fighting plastic pollution” have the strongest connection (i.e., tie), which are mentioned together in 28.2% (n= 9.265) of all tweets by the public (n=32.887).
Table 5. Matrix of Public network attribute agenda A B C D E F G H I J K L M N A - B 9 - C 279 - - D 520 - 9 - E 79 - 1 - - F 366 - 124 37 1 - G 5289 12 141 36 57 173 - H 3373 25 672 67 28 404 1168 - I 83 - 109 5 1 243 29 76 - J 31 - 10 5 - 3 19 40 - - K 1953 48 156 34 51 136 870 1549 43 42 - L 734 6 423 11 - 3 164 1257 2 6 549 - M 7725 12 394 464 70 360 3988 2786 88 73 2086 1235 - N 9265 89 1194 376 147 736 4702 5500 250 71 2901 1417 7990 -
Note: A = Fighting plastic pollution; B = VAT; C = Animal welfare; D = Digitization & innovation; E = Donations; F = Research & information provision; G = Deposit money; H = Waters; I = Education; J = The origin of plastic pollution; K = Sustainability &
environmental impact; L = Health; M = Specific (plastic) products; N = Stakeholder involvement
Equivalently to the NGO network attribute agenda, “Fighting plastic pollution”, “Stakeholder involvement”, “Specific (plastic) products”, “Waters” and “Deposit money”
form a sub-structure in the network attribute agenda of the public. This might imply that, similar to NGOs, the public more strongly focused on solutions for combating the Plastic Soup issue (e.g., the introduction of a new legislation regarding deposit money) and the involvement of stakeholders (e.g., the Dutch government & political parties) than media do.
Hypothesis testing
H1 stated that the network attribute agenda of the media is positively correlated with network attribute agenda of the public. According to the QAP correlation test the media and public network attribute agendas are significantly and evidently correlated (Pearson’s r = .723, p < .001). This finding provides support for H1 that the public’s network attribute agenda is strongly aligned with the network attribute agenda of the media.
H2 stated that the network attribute agenda of NGOs is also positively correlated with network attribute agenda of the public. According to the QAP correlation test the NGO and public network attribute agendas are significantly and strongly correlated (Pearson’s r = . 975, p < .001). This result supports H2 that the network attribute agenda of NGOs is strongly aligned with the public’s attribute agenda. The QAP correlation coefficient is higher between NGOs–the public (r = . 975) than the correlation coefficient between media–the public (r = .723), thus, H3 is also supported. The NGO network attribute agenda is more strongly correlated to the public network attribute agenda, than the media network attribute agenda.
In order to test H4, which hypothesizes that the network attribute agenda of NGOs is better at predicting the public network attribute agenda than the media network attribute agenda, QAP regression tests were performed. The QAP regression model with news media—independent matrix—and the public—dependent matrix— shows that, at a
significance level of α=0.05, the network attribute agenda of the media significantly predicts the public’s network attribute agenda, F (1, 180) = 197, p < 0,001. Based on the
Goodness-of-fit (i.e., R2), the network attribute agenda of the media is able to predict 51.2% (R2 = .512) of the differences in the public network attribute agenda.
The QAP regression model with NGOs—independent matrix—and the public— dependent matrix— shows that, at a significance level of α=0.05, the network attribute agenda of NGOs also significantly predicts the public’s network attribute agenda, F (1, 180) = 3.502, p < 0,001. Based on the Goodness-of-fit (i.e., R2), the network attribute agenda of the media is able to predict 95.1% (R2 = .951) of the differences in the public network attribute agenda. Compared with the network attribute agenda of the media, the NGOs network attribute agenda has the highest explanatory power, that is, the highest R2 value.
When computing both media and NGOs as independent matrices and the public as a dependent matrix, the exploratory power of the QAP regression model slightly improves, offering an increase of .003 R2. The QAP regression model is significant F (2, 179) = 1.888, p < 0,001 at a significance level of α=0.05. Together, the media and NGOs network attribute agendas are able to significantly predict 95.4% (R2 = .954) of the differences in the public network attribute agenda. The QAP regression results are presented in Table 6.
When considering the individual predictors, the presence of a media tie significantly (p=.047) increases the probability of a public tie by 3,76 and the presence of an NGO tie significantly (p<.001) increases the probability of a public tie by 22,54.
In line with the initial expectations, the network attribute agenda of NGOs is better at predicting the public network attribute agenda than the network attribute agenda of the media. Thus, H4 is supported.
Table 6. R2 Value Comparison Across All QAP Correlation tests and Regression Models QAP correlation (Pearson’s r) QAP regression (Adjusted R2)
Media – the Public .723 .520
NGO – the Public . 975 .951
Media * NGO – the Public .954
Note. QAP = Quadratic Assignment Procedure.
All tests showed statistically significant results, p < .001.
Conclusion & Discussion
Through a series of computer assisted methods, this study empirically tested an emerging agenda-setting theory—the NAS model (e.g., Guo & Vargo, 2015; Vargo et al., 2014; Vu et al., 2014)—using large datasets on Twitter. The aim of this study was to apply the NAS model and research to what extent attributes of the Plastic Soup issue are transferred from the media network attribute agenda and the NGO network attribute agenda to the public network attribute agenda. In the next sections, the conclusions of this research are discussed
empirically and theoretically and both scientific and practical implications are offered. In this study, the tweets of media, NGOs and the public agendas were examined from a network perspective (Vargo et al., 2014), that is, how the media, NGOs and the public
associated different attributes to discuss the Plastic Soup issue on Twitter. The results provide a more detailed view of the agenda-setting role of the media and NGOs on social media, by suggesting that both the media and NGOs do not simply transfer the salience of attributes related to the Plastic Soup issue, they instead transfer the salience of the interconnections amongst these attributes (i.e., network attribute agenda).
The findings show that both news media and NGOs play a fundamental role in how the public links attributes of Plastic Soup, because both stakeholder groups have the power to set
the public agenda. The more often news media or NGOs associate two (or more) attributes in their tweets, the more likely it is that the public will also associate these two (or more) attributes in their tweets. Strikingly, the results suggest that NGOs have a stronger agenda-setting power than news media and thereby have a greater ability to build new associations, as well as enhancing already existing ones, and determine the way the public links different attributes with regard to Plastic Soup.
The media mostly link the attributes “Fighting plastic pollution”, “Stakeholder
involvement”, “Specific (plastic) products”, “Waters” and “The origin of plastic pollution” in their tweets about Plastic Soup. On the other hand, NGOs more strongly focus on
“Stakeholder involvement” than media and predominantly associate this attribute with “Fighting plastic pollution”, “Specific (plastic) products”, “Waters” and “Deposit money” in their tweets about Plastic Soup. Corresponding to the NGO agenda network, the public also links “Stakeholder involvement”, “Fighting plastic pollution”, “Specific (plastic) products”, “Waters” and “Deposit money” in their tweets about Plastic Soup. Both NGOs and the public thereby focus on the introduction of a new legislation by the Dutch government & political parties regarding deposit money on litter, plastic bottles and cans.
Empirically, this study provided a new set of empirical evidence to support the NAS model, which asserts that media cannot only influence what the public thinks about (i.e., first level of agenda-setting) and how to think about it (i.e., second level of agenda-setting), but also have the ability to influence how the public links different attributes (i.e., third level of agenda-setting). In line with previous research on the NAS model (e.g., Guo & Vargo, 2015; Vargo et al., 2014), the findings of this study suggest that even in online issue arenas, media remain powerful at setting the public agenda and can influence how the public links different attributes of the Plastic Soup issue.
Additionally, where previous NAS studies examined the third level of agenda-setting from a media-centered perspective (e.g., Guo & Vargo, 2015; Vargo et al., 2014), this study applied the NAS model from an issue arena-perspective. As such, this study expanded the scope of research on the NAS model by not only researching the agenda-setting power of the media, but also that of NGOs (an important stakeholder group in environmental issue arenas; Luoma-aho & Vos, 2010). This study found that NGOs have stronger agenda-setting power than news media in online issue arenas. This finding suggests that NGOs are very influential at putting topics on the public agenda, which is in line with findings of earlier research on NGOs’ ability to influence the public on social media (e.g., Luxon & Wong, 2017;
Meriläinen & Vos, 2011; Yang and Saffer, 2018), but counterargues traditional agenda-setting views (e.g., Carroll & McCombs, 2003; McCombs & Shaw, 1972) by showing that media do not have absolute power or autonomy in setting the public agenda.
Methodologically, automated mapping of issue arenas is important in complementing the more qualitative approaches that have been applied so far (Hellsten et al., 2019). By using big data- and semantic network analyses, visualizations of the attribute networks on Twitter are enabled. As Guo & Vargo (2015, p.573) stated: “We can see how stakeholders linked attributes together.”. These methods are important in highlighting different stakeholder relations and debates in which organizations such as the media and NGOs are involved (Hellsten et al., 2019).
This study also provides several practical implications for corporate communication and PR-practitioners and journalists. For communication professionals and journalists, the NAS model provides important insights which help professionals understand how public opinion is influenced by associating different attributes in communication. NGO’s can use these insights and the findings of this study to effectively design campaign messages by linking different attributes. Based on the findings of this study, NGOs should strongly
associate attributes that are concerned with combating Plastic Soup issue (i.e., “Fighting plastic pollution” and “Deposit money”) and the involvement of stakeholders (i.e.,
“Stakeholder involvement”) in their campaign messages to influence the public towards a pro-environmental stance and mobilize the public to enhance the value of NGOs and to pressure policy makers (Luxon & Wong, 2017; Meriläinen & Vos, 2011). On the other hand, it is important for journalists to be aware that associating attributes impacts how the public is influenced and links these attributes (Guo & Vargo, 2015). Journalists could use these insights in editorial work by influencing public opinion on (environmental) issues through frequently linking attributes of these issues.
With the present study, a first step has been taken to expand the NAS model and
examine the predictive power of the attribute agendas of different stakeholder groups in issue arenas. The affirmative findings of this study suggest that this research area remains an understudied and therefore a promising area to explore in future research projects. For
example, besides the agenda-setting power of news media and NGOs, the predictive power of other stakeholders in issue arenas (e.g., politicians and political parties, corporations,
governments and local communities) could be studied. In future studies, influencer marketing theories (e.g., Aguiar & Van Reijmersdal, 2017) and the two-step flow theory (e.g., Katz, & Lazarsfeld, 1955) could be used to differentiate these stakeholder groups into organizations on the one hand and persons on the other hand, to examine the differences in agenda-setting power between organizations and opinion/influencers, who play a fundamental role in influencing the public and setting the public agenda nowadays (Aguiar & Van Reijmersdal, 2017, Katz & Lazarsfeld, 1955; van Ruler, 2018). Finally, for further lines of research it would be of great interest to not only research agenda-setting in issue arenas, but altogether research the interrelated phenomena of agenda-building and agenda-setting to gain an even clearer image of the interplay between stakeholders.
Limitations
This study is limited in various aspects. One general pitfall of big data is traditional data quality topics in social sciences (e.g., reliability, internal and external validity, sample design) are often ignored (Malthouse & Li, 2017). This study is based on non-random sampling (i.e., a large convenience sample) by using data that is technically and legally accessible, which may not represent the full population of interest (i.e., the Dutch population). Additionally, Twitter generally does not represent the Dutch population nor can it be assumed that accounts and users are equivalent. According to boyd & Crawford (2012, p. 669), “some users have multiple accounts, while some accounts are used by multiple people, some people never establish an account and some accounts are ‘bots’ that produce automated content without directly involving a person.”
Secondly, not all relevant causal factors are measured and therefore the model
estimates of the QAP regression tests can be biased (Malthouse & Li, 2017). Even though the QAP regression test shows that the network attribute agendas of both the media and NGOs are powerful at predicting the agenda of the public, it is important to note that no potential extraneous variables (e.g., real-world cues, public opinion, press release, government websites; Vargo & Guo, 2017) that could influence the relationship are included in this research. Therefore, it might be possible that practices of apophenia are enabled. According to boyd & Crawford (2012, p. 668), the practice of apophenia means that “patterns are seen where none actually exist, simply because enormous quantities of data can offer connections that radiate in all directions”. Future researchers could include multiple data sources (e.g., public opinion surveys, political debates) to identify causal factors.
One specific limitation is the inclusion of retweets in the Twitter dataset. Big data research tends to only show what users and not why they do it. The number of times a message gets retweeted on Twitter, may show a certain degree of interest by users, but
without looking at the content and/or style of a tweet, it might imply popularity, outrage or simply the “thoughtless routines of Twitter usage behavior” (Mahrt & Scharkow, 2013, p. 29). Additionally, including retweets might also bias the results. In this study a maximum of 250 keywords were chosen to identify attributes for the news media, NGOs and the public. An unintentional effect of including retweets is that for all three stakeholder groups mostly the same keywords have been identified: the maximum of 250 keywords of all three
stakeholder groups resulted in a total of 261 unique keywords in the attribute construct lists. As an effect, the three network attribute agendas might have been too much alike and the findings of this study (i.e., the high positive correlations and very significant results of the QAP tests) could be biased and be simply attributed to the methodological choices of including retweets and choosing a threshold of a maximum of 250 keywords, which both limited the number of different attributes that were examined in this study. Future research could expand the threshold and include more keywords, to expand the number of different attributes and avoid potential threshold bias.
Finally, generalizability of the results might be problematic. Environmental issues and CSR must be considered as being socially and culturally dependent. According to Golob, Turkel, Kronegger and Uzunoglu (2018), environmental- and CSR communication are generally “locally translated” due to national circumstances and preconditions, when
introduced in a specific country (such as the Netherlands). Therefore, generalizability of the results remains difficult because the meaning of environmental- and CSR communication differs according to society, countries or cultural backgrounds (Golob et al., 2018). Because of country-specific results and Twitter as a platform of data collection does not represent the Dutch population, information from many different platforms and persons (Malthouse & Li, 2017; Mahrt & Scharkow, 2013) and different countries and cultural backgrounds (Golob et
al., 2018) should be included in future research in order to expand and test the generalizability of the results.
References
Aguiar, D., & Van Reijmersdal, E. A. (2017). What is influencer marketing? In Influencer Marketing (pp. 13–27). SWOCC: Amsterdam.
Albu, O. B., & Etter, M. (2016). Hypertextuality and social media: A study of the constitutive and paradoxical implications of organizational Twitter use. Management Communication Quarterly, 30(1), 5-31.
boyd, D. & Crawford, K. (2012). Critical Questions for Big Data. Information,
Communication & Society, 15(5), 662–679. DOI: 10.1080/1369118X.2012.678878 Burchell, J., & Cook, J. (2008). Stakeholder dialogue and organisational learning: changing
relationships between companies and NGOs. Business Ethics: A European Review, 17(1), 35-46. DOI: 10.1111/j.1467-8608.2008.00518.x
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. DOI:10.1057/palgrave.crr.1540188
Cialdini, R. B., Demaine, L. J., Sagarin, B. J., Barrett, D. W., Rhoads, K., & Winter, P. L. (2006). Managing social norms for persuasive impact. Social influence, 1(1), 3-15. DOI: 10.1080/15534510500181459
Clarke, G. (1998). Non-Governmental Organizations (NGOs) and politics in the developing world (pp. 36–52). XLVI: Political Studies.
Dauvergne, P. (2018). The power of environmental norms: marine plastic pollution and the politics of microbeads. Environmental Politics, 27(4), 579-597. DOI:
Dobele, A. R., Westberg, K., Steel, M., & Flowers, K. (2014). An examination of corporate social responsibility implementation and stakeholder engagement: A case study in the Australian mining industry. Business Strategy and the Environment, 23(3), 145-159. DOI: 10.1002/bse.1775
ESOMAR. (2017, September). Esomar Data Protection Checklist. Retrieved on May 10, 2019 from: https://www.esomar.org/uploads/public/knowledge-and-standards/codes-and-guidelines/ESOMAR-Data-Protection-Checklist_September-2017.pdf
Freeman, R.E. (1984), Strategic Management: A Stakeholder Approach, Pitman, Boston, MA.
Getchell, M. C., & Sellnow, T. L. (2016). A network analysis of official Twitter accounts during the West Virginia water crisis. Computers in Human Behavior, 54, 597-606. DOI:10.1016/j.chb.2015.06.044
Golob, U., Turkel, S., Kronegger, L., & Uzunoglu, E. (2018). Uncovering CSR meaning networks: A cross-national comparison of Turkey and Slovenia. Public Relations Review, 44(4), 433-443. DOI: 10.1016/j.pubrev.2018.05.003
Guo, L. (2012). The Application of Social Network Analysis in Agenda Setting Research: A Methodological Exploration. Journal of Broadcasting & Electronic Media, 56(4), 616-631. DOI: http://dx.doi.org/10.1080/08838151.2012.732148
Guo, L. & McCombs, M. (2011). Network agenda setting: A third level of media effects. Paper presented at the ICA, Boston.
Guo, L. & Vargo, C. (2015). The power of message networks: A big-data analysis of the Network Agenda Setting Model and issue ownership. Mass Communication &
Haro-de-Rosario, A., Sáez-Martín, A., & del Carmen Caba-Pérez, M. (2018). Using social media to enhance citizen engagement with local government: Twitter or Facebook?. New Media & Society, 20(1), 29-49. DOI: 10.1177/1461444816645652
Hellsten, I., Jacobs, S., & Wonneberger, A. (2019). Active and passive stakeholders in issue arenas: A communication network approach to the bird flu debate on Twitter. Public Relations Review, 45(1), 35-48. DOI: 10.1016/j.pubrev.2018.12.009
Hong, Y. J., Shin, D., & Kim, J. H. (2016). High/low reputation companies' dialogic communication activities and semantic networks on Facebook: A comparative study. Technological Forecasting and Social Change, 110, 78-92. DOI:
10.1016/j.techfore.2016.05.003
Katz, E., & Paul, F. (1955). Lazarsfeld (1955), Personal Influence. The Part Played by People in the Flow of Mass Communication. New York.
Lovejoy, K., Waters, R. D., & Saxton, G. D. (2012). Engaging stakeholders through Twitter: How nonprofit organizations are getting more out of 140 characters or less. Public
Relations Review, 38, 313-318.
Luoma-aho, V., & Vos, M. (2010). Towards a more dynamic stakeholder model: acknowledging multiple issue arenas. Corporate Communications: An International Journal, 15(3), 315-331. DOI: 10.1108/13563281011068159
Luxon, E. M., & Wong, W. H. (2017). Agenda-setting in Greenpeace and amnesty: the limits of centralisation in international NGOs. Global Society, 31(4), 479-509.
Mahrt, M., & Scharkow, M. (2013). The Value of Big Data in Digital Media Research. Journal of Broadcasting & Electronic Media, 57(1), 20–33. DOI:
