It is not always the same, but there is always some: A quantitative assessment of Twitter and Instagram activity in Dutch politics

(1)

It is not always the same, but there is always some

A quantitative assessment of Twitter and Instagram activity in Dutch politics

Antonio Deda (10801715)

University of Amsterdam

Graduate School of Communication Master Political Communication Master Thesis

Supervisor: dr. mr. J.H.P. van Spanje

June 26, 2020

(2)

1

Abstract

Social networking sites (SNS), demanding a little amount of resources to be used effectively, are perfectly suited for use in a permanent campaign. In an effort to explore what such permanence entails in the online realm, two different conceptualizations of permanent campaigning are tested. The intensity of actors’ online activity was tested for similarity over time, and data was examined to see whether a minimum level of activity could be found at all times. An extensive data analysis, focusing on Twitter and Instagram activity of Dutch

politicians and political parties, shows that activity is not similar across periods: actors show higher levels of activity during campaigning periods compared to non-campaigning periods. Interactivity on Twitter – measured by examining @mentions and retweets – follows a similar trend. However, none of the periods is characterized by a total absence of activity. An

additional exploratory effort shows that a second-order campaigning period can be distinguished, which is found to contain significantly higher levels of activity than routine periods. Campaigning thus is not found to be permanent in terms of similar intensity. Instead, results show that political actors have established a permanent online presence.

(3)

2

Introduction

Following the 1989 Dutch parliamentary elections, the Dutch Socialist Party (SP; Dutch:

Socialistische Partij), confronted with an unfavorable outcome, decided to switch their

approach to political campaigning. In an effort to build a more unitary national image, the party started employing a more continuous mode of campaigning. Regularly, the party leader published an open letter – attacking other parties’ policies – in a national newspaper, and opinion sections of newspapers were taken up by the party as well. In addition, the party appeared at policy-aimed protests, connecting its national faction to local issues (Van der Steen, 1995). Such a mix of campaigning efforts produced one of the first notable examples of a permanent campaign being employed within Dutch politics. Since then, the advent of online media is thought to have expanded the possibilities for employing such campaigning, which asks for a thorough examination (Blumler & Kavanagh, 1999; Lilleker & Jackson, 2013; Norris, 2000).

Assessing the permanence of online political campaigning

Having been coined by Sidney Blumenthal in the early 1980s, the notion of permanent

campaigning does seem to apply well to campaigning efforts that are made by political actors

– politicians and political parties – in the present age. Characteristic of this campaigning style is the blurring of a distinction between electoral campaigning and the act of governing

(Blumenthal, 1982), or, as Hugh Heclo put it nearly two decades later: “a nonstop process seeking to manipulate sources of public approval to engage in the act of governing itself” (Heclo, 2000, p. 17). While political campaigning and governing are interdependent – one cannot exist without the other – the two are thought to be fundamentally different from one another, making such blurring problematic (Heclo, 2000). Political campaigning, on the one hand, is necessarily focused on the actors’ own electoral successes (Heclo, 2000).

(5)

4

to deliberation (Ferree, Gamson, Gerhards, & Rucht, 2002; Strömbäck, 2005). Although a loss of a clear delineation between the two – as permanent campaigning supposes – does not cause any of the practices to vanish completely, it does hold severe implications for the extent to which both acts are carried out successfully. If political actors exhibit oppositional attitudes, which are typical of campaigning periods, outside of these periods as well, this would be detrimental to the collaborative tradition of governing. The introduction of online social networking sites (SNS) in the field of political communication are believed to be perfect enablers of such incessant campaigning behavior, offering political actors to stay in direct contact with their electorate – without any third-party intervention – at a low cost of resources (Van Aelst et al., 2017). In an effort to examine whether political actors are actually using online media in a way that blurs the line between campaigning and governing, the following research question is posed:

RQ: To what extent can the campaigning efforts of political actors said to be in line with the

notion of a permanent online campaign?

The current study will regard political actors – individual politicians and political parties – in the Dutch national arena. While research regarding the online activity of Dutch political actors is not at all sparse (cf. Graham, Jackson, & Broersma, 2014; Jacobs & Spierings, 2019; Kruikemeier, 2014; Vergeer, 2015), analyses of such activity in the light of examining the permanence of political campaigns to date have mostly regarded actors in the United States (Vasko & Trilling, 2019), in some Scandinavian countries (Enli & Skogerbø, 2013; Larsson, 2015, 2016), in more depth, and throughout a multitude of countries, providing an overview (Ceccobelli, 2018). Because an assessment of Dutch national politics in this regard has not been conducted yet, the current study aims to narrow the gap that exists here. With two thirds of Dutch adults active on SNS – far above the EU average – political actors in the country are expected to attach higher value to online campaigning than political actors in

(6)

5

countries with lower SNS adoption rates, rendering a suitable locale for testing online campaigning permanence (Eurostat, 2020a).

Specifically, the present study focuses on actors’ activity on two major SNS: Twitter and Instagram. Looking at their use, examination of those platforms is highly relevant: Twitter and Instagram cater 18 and 35 percent of adult Dutch internet users (15+), respectively. While Twitter activity is often central in examining online political communication (cf. Kruikemeier, 2014; Larsson, 2015, 2016; Vergeer, 2015), the number of active Dutch user accounts on the platform is on the decline, showing a ten percent drop between 2018 and 2019 (Newcom, 2019). On the contrary, Instagram, having entered the online realm a few years later than the popular microblogging service, has gained popularity among the Dutch public, showing a twenty percent increase in the number of users in the same period (Newcom, 2019).

It should be noted that significant differences exist between the focus that the two platforms hold: while Twitter’s microblogging platform is built around sharing posts in the form of short texts, Instagram pushes its users toward publishing visual content, which explains the association of the latter of the two with a more personalized style of political campaigning (Ekman & Widholm, 2017). This diversity in content focus might explain differences in the way political actors interact with them, rendering another valid reason to include both of them in the present analyses.

Studying the activity of political actors over a certain time period, an extensive dataset was constructed based on over 800,000 posts, fetched from political actors’ Instagram and Twitter accounts. This file showed a daily average amount of posts that were published by political actors, which could then be used to compare activity in campaigning periods to that outside of these periods.

First, an overview of relevant literature on the topic of permanent campaigning will be provided, which is followed by an outline of the procedures that were employed. Then, the results from the analyses are presented, and will be further reflected upon in the concluding

(7)

6

section. Finally, a number of limitations of the study are discussed, and avenues for future research are provided.

Theory

A brief history of political communication

In the decades after WWII, political communication has constantly been evolving. Classifying these developments, Blumler and Kavanagh (1999) distinguish between three different ages of political communication.

In the first age, motivations for voting for a certain party were mainly based on voters’ identification with the party (Blumler & Kavanagh, 1999). Media were driven by a partisan

logic, providing political actors with an easy-to-use and effective instrument which they could

use to disseminate their political messages among the public (Blumler & Kavanagh, 1999; Brants & Van Praag, 2006). Additionally, political campaigning in this period involved a strong role for non-mediated communication: going out to meet the electorate formed a vital part of any political campaign (Norris, 2000).

The use of television quickly became central to most political communication

strategies in the second age (Blumler & Kavanagh, 1999; Norris, 2000). In terms of potential reach, the new mass medium offered the possibility to deliver messages to far larger groups of people. Media were governed by a public logic, serving as a platform which provides citizens with the information that they need in order for a democracy to function properly (Brants & Van Praag, 2006).

The introduction of the internet in the third age increased the pressure on media organizations to attract the attention of the audience, leading to heightened

commercialization of media (Blumler & Kavanagh, 1999; Brants & Van Praag, 2006; Entman, 1990). A fragmented audience in an environment governed by media logic meant an

important shift in power at the expense of political actors: instead of underlining the relevance of certain issues or policy stands themselves, media organizations seized the

(8)

7

power in determining what is in the spotlight (Brants & Van Praag, 2006). This left political actors confronted with the challenge of showing their relevance to the public, constantly trying to capture its attention (Blumler & Kavanagh, 1999).

A fourth age in political communication

Approaching the present day, the novelty of online media has faded. With the introduction of the internet well into the past, penetration rates in most western democracies are now well over 80 percent (Eurostat, 2020b). The emergence of platforms on which online communities could be established – in the form of interactive social networking sites (SNS) (Lilleker & Jackson, 2013; O’Reilly, 2005) – gradually enhanced the need for bringing online activities towards the center of political campaigns (Lilleker & Jackson, 2013). Scholars have argued such developments to mark the arrival of a fourth age in political communication (e.g. Blumler, 2016; Strömbäck, 2008). Instead of submitting themselves to a passive role on the receiving end, SNS provided citizens with the possibility to actively engage in communication – a contemporary translation of the direct means of campaigning, typical of the first age (Vergeer, Hermans, & Sams, 2013).

Usage of online media is often favored over traditional media by individuals that hold (the aim of getting into) a public function, for which they hold good reasons (Larsson, 2016). By using SNS, political actors gain a great amount of independence from traditional media, weakening media’s position as gatekeepers to the information that the public consumes (Brants & Van Praag, 2017; Van Aelst et al., 2017; Vergeer et al., 2013). Additionally, compared to their traditional counterparts, online media are far more efficient in terms of expending resources: usage of platforms is free of charge, platforms require a minimal level of skill to operate them, and their capacity is virtually limitless (Van Aelst et al., 2017). When used actively, SNS hold the potential to boost actors’ electoral successes (Kruikemeier, 2014). Contrasting with traditional media, online media thus are perfectly suited for carrying out a continuous or permanent campaign (Vergeer et al., 2013).

(9)

8

The permanent campaign

In his 1982 book, Sidney Blumenthal was the first to propose a permanent campaign to be employed in US politics. Central to this idea is that after traditional campaigning periods – typically the weeks or months preceding election day – have ended, political actors still continue trying to appeal to the public, leading to an amalgamation of governing and campaigning (Blumenthal, 1982). However, both practices differ significantly from one another. When campaigning, political actors hold the focus of yielding successful electoral results, which is achieved by putting an accent on persuasion (Heclo, 2000). Governing, by contrast, does not work towards a specific event, but takes place at a much lower pace, allowing time for extensive deliberation to take place (Heclo, 2000). Evaporation of a clear distinction between the two is deemed problematic, since hostility, which is inherent to campaigning, is wildly inconsistent with the collaborative nature of governing.

At the same time, there is a growing necessity for political actors to communicate to citizens outside of campaigning periods (Ceccobelli, 2018; Vergeer et al., 2013). The demise of voters’ identification with political parties means that the audience has to be targeted continuously, since unconditional, long-lasting support cannot be assumed anymore (Blumler & Kavanagh, 1999; Norris, 2000; Vergeer et al., 2013). Such practice is vital to political actors in order to attract the attention from the public.

Defining permanent campaigning

As Larsson (2016) rightfully mentions, “it is difficult to provide precise metrics regarding what such permanence would entail in an empirical setting” (Larsson, 2016, p. 276). Following Blumenthal (1982), one would expect permanency to point into the direction of equality in the degree of campaigning activity across campaigning and non-campaigning periods. Applying this definition to political actors’ online activity, results fail to provide a uniform image. Looking at political actors’ activity on SNS, while several scholars find an increase in the level of activity as election day approaches (Graham et al., 2014; Kruikemeier, 2014;

(10)

9

their analysis. However, such a dichotomous approach to assessing the permanence of campaigning does leave very little opportunity for campaigns that show slight variation in their levels of activity across periods to be classified as a permanent campaign.

An alternative conceptualization is therefore provided by Ceccobelli (2018), who pointed out that the act of permanent campaigning is not necessarily a matter of making extensive use of a single instrument all the time: it is when elements of the total set of

campaigning activities are carried out in both campaigning and non-campaigning periods that a permanent campaign is observed (Ceccobelli, 2018). This renders an argument for

examining more than one medium, which will be further elaborated upon in a next section. Most importantly, Ceccobelli (2018) notes that permanent campaigning “is a variable that can change over time and in differing contexts” (Ceccobelli, 2018, p. 123), rather than being constant across periods. Similarly, Van Aelst and De Swert (2009) argue such differences to be the base of a distinction between campaigning and non-campaigning periods. Instead of only detecting perfect permanent campaigning, as set out by Blumenthal (1982), this second approach recognizes that campaigning periods contain more activity than non-campaigning periods. Ceccobelli (2018), in his examination of Facebook activity of political leaders, only found then US president Obama to be engaging in perfect permanent campaigning, showing an equal average of daily posts during the campaigning period (1.5 posts) and the non-campaigning period (1.6 posts). Ceccobelli (2018) defined an actor to be engaging in less perfect, but still permanent campaigning when they published an average of at least 0.5 posts per day. Using this conceptualization, 34 out of 42 studied politicians were considered to be campaigning permanently, which adds to the ambiguity in determining whether or not political actors engage in permanent campaigning (Ceccobelli, 2018). Taking into

consideration the increasing adoption of SNS over the last decade (Eurostat, 2020a), the amount of activity is expected to be higher than at the time of Ceccobelli's (2018) analysis as well. Therefore, in the present study, a minimum average of 1 post per day is taken as an alternative indicator of permanent campaigning.

(11)

10

Because neither the original definition put forward by Blumenthal (1982), nor the innovative proposition by Ceccobelli (2018) seem to provide a clear view of the permanence of online campaigns on themselves, the current analysis incorporates both approaches. If the level of activity indeed increases as election day approaches, as some authors found

(Graham et al., 2014; Kruikemeier, 2014; Larsson, 2016; Vergeer et al., 2013), the campaigning period should thus show a significantly higher amount of activity than other periods, which would fail to meet Blumenthal's (1982) definition of permanent campaigning. The study then also aims to test whether the activity of actors exceeds a certain baseline level, as compliant with Ceccobelli's (2018) definition. Since second-order elections will be considered as well – for which further support will be provided in the last subsection of the literature section – campaigning periods will be referred to as first-order campaigning

periods. Examining user accounts on both Twitter and Instagram, on which details will be

provided in the next section, the following sets of hypotheses are formulated:

H1a: The average number of Twitter posts per day, published by any actor, is significantly

higher in the first-order campaigning period than in other periods.

H1b: The average number of Twitter posts per day, published by any actor, exceeds 1 in

every period.

H2a: The average number of Instagram posts per day, published by any actor, is significantly

higher in the first-order campaigning period than in other periods.

H2b: The average number of Instagram posts per day, published by any actor, exceeds 1 in

every period.

Political campaigning on Twitter

When conducting research on online political campaigning, scholars often select actors’ Twitter accounts as the subject of their examination (cf. Bruns & Highfield, 2013; Enli & Skogerbø, 2013; Graham et al., 2014; Gross & Johnson, 2016; Kelm, 2020; Kruikemeier,

(12)

11

2014; Larsson, 2015; Lev-On & Haleva-Amir, 2018; Van Aelst et al., 2017; Vasko & Trilling, 2019; Vergeer, 2015; Vergeer et al., 2013). It is easy to see why such a choice is often made: while providing a free forum for anyone to have a discussion, the userbase on the platform is far from representative of the population as a whole, with political actors at the heart of most interactions (Jungherr, 2016). Generally, political actors are also shown to be among the most active actors on the platform, providing heightened relevance for studying political actors’ behavior on the platform (Grant, Moon, & Grant, 2010).

Because of its public nature, Twitter is different from most other SNS: with the exception of explicitly blocked users, any user on the platform can view, share and interact with messages from any other user, without any need for reciprocal confirmation of a

connection, which is typical of most other SNS (boyd & Ellison, 2007; Vasko & Trilling, 2019). From the perspective of political actors, this openness of the platform renders it more

suitable for political marketing instead of posting personalized messages (Enli & Skogerbø, 2013). Such use is reflected in the style of political actors’ communication through the

platform, which seems to be rather unidirectional: from political actors to the audience (Enli & Skogerbø, 2013; Graham et al., 2014). However, Twitter does provide tools for increased interactivity. Besides posting a comment or being able to express liking a post (boyd & Ellison, 2007), it is possible to seek direct contact with another user on the platform through the use of @mentions (Bruns & Moe, 2014). While actors might not engage in such activity as an explicit part of a political campaign, this interactive way of engaging with others shows highly beneficial to political actors: those that explicitly mention other users more frequently, are found to be the ones that enjoy the greatest amounts of preferential votes in elections (Kruikemeier, 2014). Less direct communication is reflected through retweeting: bringing a post from another user to the attention of one’s own cortege (Bruns & Moe, 2014). Again, users do not need any approval to do so, which enables messages to be picked up and spread by numerous other users at a fast rate (Bruns & Moe, 2014).

Applying Blumenthal's (1982) conceptualization of permanent campaigning – that campaigning is constant throughout periods – to the contents of political actors’ activity, the

(13)

12

degree to which political actors engage in interactive behavior would then be expected to be similar across periods as well. Similar to the formulation of H1a and H2a, the current study then aims to falsify the contention that political actors show higher levels of interactivity during the campaigning period than during other periods. Using @mentions and retweets as indicators of actors’ interactivity, the following set of hypotheses is posed:

H3a: The share of @mentions in the political actors’ total number of posts during the

first-order campaigning period is significantly higher than the share of @mentions in other periods.

H3b: The share of retweets in the political actors’ total number of posts during the first-order

campaigning period is significantly higher than the share of retweets in other periods.

Political campaigning on Instagram

Having entered the online realm in 2010, only four years after the launch of Twitter, Instagram still has been studied far less than the microblogging service. The younger platform, which is less established within political communication, places an emphasis on images rather than on text, contrasting with the Twitter infrastructure (Ekman & Widholm, 2017). While such an image-based focus may seem to render the platform rather unique, to a certain extent, Instagram fits with the concept of ‘classic’ SNS. Users’ timelines are public by default, but account holders may restrict access to their profiles. In doing so, all new follower requests must first be approved by the user before posts are shown, creating a much more private sphere. In terms of interactivity, Instagram offers its users a smaller number of tools to engage in such behavior: other than the common functions of liking other users’ posts and posting a comment, no tools are provided (Highfield & Leaver, 2015), which is why no additional hypotheses will be included in this regard.

(14)

13

Breaking through the campaigning dichotomy

Studies holding the aim of providing insights into the online activity of political actors have typically distinguished between two types of periods. The period in which a government is active – and governing – has been referred to as the routine period (e.g. Van Aelst & De Swert, 2009; Vasko & Trilling, 2019). Its position opposing campaigning or election periods suggests that routine periods are expected to be relatively exempt of campaigning activities. During routine periods, political actors put in less effort to appeal to the public than in

campaigning periods, which leaves little surprise that fewer people are interested in politics during routine periods (Van Aelst & De Swert, 2009).

Looking at activity figures, Vergeer et al. (2013) and Kruikemeier (2014) both find a sudden drop in activity after elections have taken place. In most democracies, elections are not instantly followed by the installation of a new government. Although such a lame duck phase is typically associated with US politics (Jenkins & Nokken, 2008; Vasko & Trilling, 2019), the Dutch system shows a similar void between election day and the installation of a new government. Executives typically refrain from introducing new legislation during this period, but they are still actively rounding off their unfinished business, rendering them less of a lame duck than in the US context (Andeweg, Van der Tak, & Dittrich, 1980). During this lame duck period, Vasko and Trilling (2019) found political actors to be less active compared to routine periods. However, taking into consideration the contents of posts revealed that the lame duck period resembles the campaigning period (Vasko & Trilling, 2019). Thus, both in terms of activity and contents, lame duck periods are found to be different from routine

periods, which is why they will be taken as a separate period in the present analyses as well. By definition, routine periods are free of elections – that is, in the same political arena. When examining election periods in a broader sense, elections that take place in another arena may be of interest as well. From the side of journalism, a distinction between different levels of government does hardly seem to matter: regardless of the arena in which an election is organized, journalists seem eager to cover the news surrounding these events in the weeks leading up to election day (Van Aelst & De Swert, 2009). This is at least of some

(15)

14

importance to actors in the main political arena, which is, those that are active on the national level, because their appearance in the main arena forms one of the factors that influence voter behavior in second-order arenas (Lefevere, 2018; Reif & Schmitt, 1980). The activity (and interactivity) of political actors during these so-called second-order campaigning periods is therefore thought to be significantly different from that in routine periods. Specifically, because of the mentioned importance of activity during this period to political actors, actors are expected to show higher levels of activity during these second-order campaigning periods. Conversely, arguing from the perspective of Blumenthal (1982), no significant differences in activity between the periods should exist. In an effort to examine whether the mentioned trichotomy of periods is expandable by a fourth, second-order campaigning period, the following hypotheses are posed:

H4a: The average number of Twitter posts per day, published by any actor, is significantly

higher in second-order campaigning periods than in the routine period.

H4b: The average number of Instagram posts per day, published by any actor, is significantly

higher in second-order campaigning periods than in the routine period.

Methods

Research setting Locale

While several efforts have been made in studying permanent online campaigning in different locales, politics in the Netherlands have been left out of such examinations (Ceccobelli, 2018; Enli & Skogerbø, 2013; Larsson, 2015, 2016; Vasko & Trilling, 2019). Thus, in the interest of contributing to the body of literature on the subject, the present research necessarily focuses on actors that are active in Dutch national politics.

With its multi-party system, Dutch politics, similar to the already studied Scandinavian countries, is best described as moderately pluralist (Sartori, 1976; Ten Napel, 1999). In its

(16)

15

bicameral parliamentary system, only the lower house (House of Representatives; Dutch:

Tweede Kamer) is elected through direct elections. Consequently, actors from the upper

house (Senate; Dutch: Eerste Kamer) have far less to gain from engaging in campaigning activities, which might explain the centrality of lower house actors in research regarding Dutch politics (cf. Graham et al., 2014; Jacobs & Spierings, 2019; Kruikemeier, 2014;

Vergeer, 2015). In line with these previous research efforts, the current study also focuses on the activity of actors in this political arena.

Timeframe

Enabling a valid assessment of actors’ activity in a first-order campaigning period, one national election was included in the timeframe. Because other studies examining political campaigning seem to be inconclusive about the length of the period before election day that should be regarded to account for a pre-election non-campaigning period, a start of the timeframe was determined to be at roughly one year before election day. Specifically, with parliamentary elections having been held on March 15, 2017, January 1, 2016 was selected as the first day on which posts from political actors are fetched. Enabling an assessment of the activity in second-order campaigning periods as well, the timeframe also includes second-order elections. Of such elections, those for the European Parliament, held on May 23, 2019, was the most recent. Including a period that is of substantial length after these elections, the timeframe ends on March 31, 2020.

Data collection

Before all posts from political actors could be fetched, first, all relevant actors had to be identified. Unfortunately, because no clear overview of political actors was available, such a file had to be constructed first. To this end, a Python program was written that used the Application Programming Interface (API) offered by OpenKamer.org1_{to register the personal}

(17)

16

details of all individuals that entered government and parliament after the installation of the First Rutte cabinet (October 14, 2010). This point in time was chosen as this was expected to ensure that the details of maximum number of individuals that were seated from January 1, 2016 onwards could be accessed through the API. All changes in parliamentary seats were then verified against lists of mutations, published by the House of Representatives, which were found through a Google search (Tweede Kamer der Staten-Generaal, 2016, 2020). Subsequently, all positions of individual politicians, including those in government positions, were verified through a manual search on Parlement.com2_{. For every individual, every} position they were active in was recorded, including the date of installation and, if applicable, the date of resignation. The positions of parties (government or opposition), including the dates on which they were installed on these positions and, if applicable, the dates of stepping down from these positions, were recorded manually through a search on Parlement.com. A total of 335 actors that were active after October 14, 2010 could be identified, occupying 450 positions. An overview of all actors and their corresponding positions can be found in the

Supplementary Materials. Each line in this sheet represents a distinct position that the actor

held.

For each of the 335 actors, a manual search was conducted to determine whether actors held an account on Twitter and Instagram. If so, the username of the account was recorded. Authenticity of each account was checked by looking at whether the account showed a ‘verified badge’, or whether the account bio contained a description of the actor’s occupation or any links to the party. If no such indications were present, the contents of posts were checked for visual or textual cues that might reveal the identity of the account holder. When any doubts about the identity of the account holder still remained, no username was registered. The list of usernames was used as input for a second and third Python program, that used the Twitter and Instagram API’s to fetch all posts for all users on the platforms. This resulted in the creation of two separate data files, containing the details of 767,702 Twitter

2_{This website shows data on the Dutch House of Representatives, collected by the independent Parliamentary}

(18)

17

posts and 45,143 Instagram posts, respectively. For all posts, published on either Twitter or Instagram, the username, user ID number, date of publication and the text in the post were recorded. Specific characteristics of both media resulted in different sets of additional variables that were included: for Twitter posts, binary registrations of whether the post was

mentioning another user and whether the user retweeted the post were also made;

registrations of Instagram posts included a variable indicating the type of post (still image, video, set of images) and a link to the post.

Describing the datasets

The datasets containing the SNS post details were combined with the data on the positions of each actor to construct a set of aggregated datasets, using a fourth Python script. A set of eight groups was defined, to which actors could be attributed – a full list can be found in Appendix A, Table A1. For every day in the timeframe, the positions file was scanned to identify which actors belonged to which of these groups on the given day. Then, for each group, the mean number of daily posts was calculated. This was done by dividing the number of posts by all actors that were in that group on a particular day by the number of actors in the group that published those posts. Such datasets were thus created for both Twitter and Instagram activity.

An additional dataset had to be constructed to enable an assessment of actors’ interactivity on Twitter. For every day in the sample, and for each of the groups, the

aggregated files contained a total number of Twitter posts that mentioned another user, and a total number of posts that were retweets. Dividing these counts by the total number of posts that were published by each group on a given day yielded a dataset that showed the proportions of Twitter posts that contained an @mention or were a retweet.

In sum, three datasets were used for the statistical analyses: one containing the aggregates of all Twitter counts, a second for the aggregates of all Instagram counts, and a

(19)

18

third that held the proportions of @mentions and retweets for every day in the timeframe3_.

Dataset manipulations

In all three datasets, a sum of all the actors’ activity was computed in order to be able to draw conclusions about the political arena as a whole. Additionally, sums of individuals’

activity and parties’ activity were calculated.

Then, a period variable was created to split the total timeframe into four periods. Previous efforts in studying political campaigning do not seem to be in accordance about the length of a typical campaigning period, ranging from a few weeks to several months (cf. Ceccobelli, 2018; Kruikemeier, 2014; Van Aelst & De Swert, 2009; Vergeer, 2015; Vergeer & Hermans, 2013; Vergeer et al., 2013). Taking into consideration a short period of time before the official launch of the national campaign, typically at one month before election day

(Vergeer & Hermans, 2013), a period of three months is selected as the classic first-order

campaigning period. Because second-order elections are expected to be of less value to

political actors in the national arena, a period of a “short campaign” (Larsson, 2016, p. 281) seems more appropriate in studying these elections. Regarding its length, scholars seem to gravitate towards a period of one month (Larsson, 2016; Van Aelst et al., 2017; Vergeer & Hermans, 2013). Additionally, Van Aelst and De Swert (2009), who studied elections in different arenas, also took this length of one month before election day. Therefore, a period of one month prior to all second-order elections was taken to account for a second-order campaigning period. All exact dates are listed in Appendix A, table A2.

Analyses

Shapiro-Wilk tests for all groups in all three datasets showed to be significant, violating the assumption of normal data distribution that parametric tests require (Hecke, 2012).

Therefore, Kruskal-Wallis tests combined with follow-up Mann-Whitney U tests were run to

(20)

19

test H1a, H2a, H3a and H3b. H1b and H2b were tested using Wilcoxon signed rank tests, and separate Mann-Whitney U tests were run to test H4a and H4b.

Results

Before any analyses were run, the data was checked to examine the adoption of SNS among political actors, further justifying the relevance of the present study. From the 335 political actors that were identified, 293 actors (87.5%) already had an active Twitter account at the start of the examined period. Over the course of four years, this number grew to 315 actors (94%), clearly indicating Twitter to be a preferred medium among political actors. The

adoption of Instagram took a sharper rise, showing an increase from 47 (14%) to 170 political actors (50.1%) that held a user account.

Figure 1

Share of politicians that hold an account on Twitter and Instagram, 2016-2020.

All three aggregated user activity datasets contained activity coefficients for every day in the timeframe (N = 1,552). Throughout the whole timeframe, a higher average of daily posts is found on Twitter (Mdn = 3.17) than on Instagram (Mdn = 1.27). A notable part of actors’ activity shows to be interactive, with the proportion of retweets being slightly higher (Mdn =

(21)

20

.44) than the proportion of @mentions (Mdn = .29). The lowest average of Twitter posts published on a single day, by all actors, was 1.3. On Instagram, a minimum across all actors meant no activity at all. Conversely, the highest amount of daily activity Twitter saw was an average of 13.8 posts; on Instagram, this maximum was at 4.9 posts. Variance in interactivity also was found. Retweets formed a minimum of 19% of actors’ daily posts, with a maximum of 77% of posts. Concerning @mentions, at least 8% of actors’ posts was found to contain a reference to another user, with a maximum of 63% of actors’ daily posts on Twitter.

To test whether the first-order campaigning period contained significantly higher numbers of Twitter posts than other periods (H1a), a Kruskal-Wallis test was run. Differences between the four periods were significant, H(3) = 169.31, p < .001. A series of follow-up Mann-Whitney U tests showed the activity in the first-order campaigning period (Mdn = 3.78) to be significantly higher than in the routine period (Mdn = 3.21, U = 38153, z = -4.28, p < .001) and the lame duck period (Mdn = 2.60, U = 4365, z = -7.99, p < .001). However, activity in the first-order campaigning period did not differ significantly from that in the second-order campaigning period (Mdn = 3.99, U = 3606, z = -1.27, p = .204). Therefore, H1a is rejected.

Examining whether a baseline level of activity could be established, exceeding an average of one post per day on Twitter (H1b), a Wilcoxon signed rank test was run. The median values in the average number of posts per day showed significantly higher than 1, in all periods. The median was highest in second-order campaigning periods (Mdn = 3.99, Z = 8.19, p < .001), which was followed closely by the first-order campaigning period (Mdn = 3.78, Z = 8.28, p < .001). Also, both the median daily activity in the routine period (Mdn = 3.21, Z = 29.34, p < .001) and in the lame duck period (Mdn = 2.60, Z = 13.01, p < .001) were significantly higher than 1. These results therefore show support for H1b.

Instagram activity, which was hypothesized to be higher in the first-order campaigning period compared to other periods (H2a), was examined in a similar fashion. A Kruskal-Wallis test again showed activity throughout periods to be significantly different, H(3) = 97.15, p < .001. Follow-up Mann-Whitney U tests showed that, across all actors, the first-order campaigning period contained significantly higher levels of activity (Mdn = 1.56) than the

(22)

21

routine period (Mdn = 1.25, U = 23921, z = -8.62, p < .001), the second-order campaigning period (Mdn = 1.35, U = 2576, z = -4.22, p < .001), and the lame duck period (Mdn = 1.31, U = 6111, z = -5.61, p < .001). H2a was therefore fully supported.

Testing for a baseline in actors’ Instagram activity of at least one post per day (H2b), again, a Wilcoxon signed rank test was conducted. On Instagram as well, all median values in the average number of daily posts showed significantly higher than 1. The highest median was found for the first-order campaigning period (Mdn = 1.56, Z = 8.19, p < .001), followed by the second-order campaigning periods (Mdn = 1.35, Z = 8.19, p < .001). The median in the lame duck period was slightly higher (Mdn = 1.31, Z = 12.77, p < .001) than in the routine period (Mdn = 1.25, Z = 28.49, p < .001), yielding support for H2b as well.

Looking at Twitter use, actors were expected to show more interactive behavior during the first-order campaigning period, of which a heightened share of @mentions (H3a) and retweets (H3b) would be an indicator. A Kruskal-Wallis test on the shares of @mentions once again showed significant difference between the groups, H(3) = 84.56, p < .001.

Contrary to expectations, the share of @mentions in the first-order campaigning period (Mdn = .24) was significantly lower than in the routine period (Mdn = .29, U = 26000, z = -7.98, p < .001), the second-order campaigning period (Mdn = .30, U = 1753, z = -6.57, p < .001), and the lame duck period (Mdn = .27, U = 7814, z = -3.30, p = .001), rejecting H3a.

As for the share of retweets, another Kruskal-Wallis test showed significant

differences between the four periods as well, H(3) = 156.97, p < .001. Here, however, results are in line with the formulated expectation: the share of retweets in the first-order

campaigning period is indeed significantly higher (Mdn = .51) than in the routine period (Mdn = .43, U = 22614, z = -9.00, p < .001), the second-order campaigning period (Mdn = .44, U = 1883, z = -6.20, p < .001), and the lame duck period (Mdn = .48, U = 8535, z = -2.32, p = .021). Hypothesis 3b thus was supported.

Lastly, the average number of posts in the second-order campaigning periods was argued to be higher than in the routine period on both Twitter (H4a) and Instagram (H4b). Running Mann-Whitney U tests demonstrated the periods to show significantly different

(23)

22

levels of Twitter activity, U = 27800, z = -7.17, p < .001. Second-order campaigning periods indeed showed more activity (Mdn = 3.99) than routine periods (Mdn = 3.21), supporting H4a. Instagram activity was significantly different across periods as well, U = 36431, z = -4.51, p < .001. Again, the daily average of posts in second-order campaigning periods was higher (Mdn = 1.35) than in the routine period (Mdn = 1.25). H4b was therefore also

supported.

Conclusion

The findings presented in the current study once again demonstrate the difficulty in providing a clear conceptualization of permanent campaigning. By employing a two-fold approach of permanent campaigning, a more complete overview of the online campaigning activities of political actors was rendered. Actors’ online activity is found to be unevenly distributed throughout periods: on Instagram, actors publish significantly more during first-order campaigning periods compared to other periods. Surprisingly, on Twitter, activity in second-order campaigning periods was found to be higher than in the first-second-order campaigning period. Blumenthal's (1982) classic conception of permanent campaigning – that campaigning activities are always employed at a similar intensity – could therefore not be applied to actors’ online activity. Acknowledging this heightened level of activity during national elections, Ceccobelli (2018) suggested looking at whether actors use their social media accounts actively outside of campaigning periods as well. Results from the present analyses show that a baseline average of one post per day could be established across all periods, on both Twitter and Instagram. While political actors thus do not seem to continuously use the virtually unlimited capacity that the online realm offers to its fullest extent, they do establish a permanent online presence on both platforms. More than ever before, political actors hold the power in determining when to go on stage, having their audience at their fingertips – quite literally.

(24)

23

While activity figures thus vary across periods, the degree of interactive behavior – measured in the proportion of Twitter posts that contained a @mention or that was a retweet – was examined as an alternative indicator of permanent campaigning taking place. Again, both the shares of @mentions and retweets are unevenly distributed throughout periods. While actor did post more retweets during the first-order campaigning period compared to other periods, this did not hold for @mentions. By contrast, the share of @mentions was lowest during this period. Because differences across all periods showed significantly different, Blumenthal's (1982) conception again fails to apply here.

An additional objective of the current study was the exploration of the possibility of extending the trichotomy of periods in political communication, which was proposed by Vasko and Trilling (2019). While these fourth, second-order campaigning periods, are usually amalgamated with the routine period, results from the present study show the need for discerning between the two. Examining the levels of actors’ activity on Twitter and Instagram across both periods shows these second-order campaigning periods to contain significantly higher levels of activity compared to the routine period.

Discussion

While online social platforms offer political actors the opportunity to be campaigning at the same intensity and in a similar style all year round, the present study shows that they are not doing so. Fortunately, actors’ online activity does not suggest a blurring of boundaries

between campaigning and governing to occur – for actors on the national level, heightened activity, and some indicators of interactivity, usually remain reserved for campaigning periods. However, actors never fully cease publishing content, pointing to some degree of permanence. In conclusion, political actors do not seem to be campaigning permanently at the same intensity, but they have established a permanent online presence.

Studying the degree of interactive behavior on Twitter showed surprising results. While actors engaged in more retweeting during first-order campaigning periods, actual

(25)

24

conversation with other users – indicated by actors’ use of @mentions – was lowest during these periods. As more extensive use of this tool is found to be produce favorable results in terms of preferential votes (Kruikemeier, 2014), results suggest actors engage in such behavior more often during first-order campaigning periods.

Limitations

Unfortunately, the current study was subject to several limitations. First, due to constraints in the scope of research, the present examination did not consider the contents of actors’ posts. The extent to which activity could be considered campaigning could therefore not be verified.

Secondly, due to limitations of the Twitter API, only the 3,200 most recent Twitter posts were examined in the analyses. Since less data was available for dates further in the past, the reliability of observations decreases when looking at dates that are closer to January 1, 2016. This also limited the possibility of examining a full election cycle – from elections in 2012 to those in 2017.

Thirdly, because Instagram did not offer functions that are similar to Twitter’s

@mentions and retweets, such measures of interactivity were not taken into account

regarding the picture-sharing platform.

Future research

Most importantly, while datasets that were constructed in the current study were suitable for studying how activity differs across actors, this could not be explored due to constraints in scope. Such analyses could show whether these more open media platforms either reflect and strengthen existing power structures, or that they form an “opening for the outsiders” (Strandberg, 2009, p. 835), offering newcomers to the political arena more possibilities to reach the public than more traditional media outlets.

Secondly, future efforts in studying permanent campaigning activities should focus on messages’ contents. A more accurate conclusion about whether campaigning messages are actually posted outside of campaigning periods as well could then be drawn, evaluating

(26)

25

Blumenthal's (1982) notion of a permanent campaign once again. Such analyses should, as results from the present study suggest, consider distinguishing between routine and second-order campaigning periods. Also, studying online activity in a full election cycle would provide a more complete image of activity in the routine period. Research focusing on the contents of messages could employ a distinction between routine and second-order campaigning

periods to examine whether these are dissimilar across periods as well. If contents show similar to that of first-order campaigning messages instead, campaigns may then not be limited to the classic first-order campaigning periods, but rather show traces of permanent campaigning, adding to findings from the present analyses.

Lastly, future data analyses that focus on Twitter data should try to circumvent the limitations the platform imposes by scraping new posts over an extensive period of time and storing these locally. This way, a more reliable sample of posts is constructed.

(27)

26

References

Andeweg, R. B., Van der Tak, Th., & Dittrich, K. (1980). Government Formation in the Netherlands. In R. T. Griffiths (Ed.), The Economy and Politics of the Netherlands

Since 1945 (pp. 223–249). doi: 10.1007/978-94-017-1382-5_9

Blumenthal, S. (1982). The permanent campaign. Simon & Schuster.

Blumler, J. G. (2016). The fourth age of political communication. Politiques de

Communication, No 6(1), 19–30.

Blumler, J. G., & Kavanagh, D. (1999). The Third Age of Political Communication: Influences and Features. Political Communication, 16(3), 209–230. doi:

10.1080/105846099198596

boyd, danah m, & Ellison, N. B. (2007). Social Network Sites: Definition, History, and Scholarship. Journal of Computer-Mediated Communication, 13(1), 210–230. doi: 10.1111/j.1083-6101.2007.00393.x

Brants, K., & Van Praag, P. (2006). Signs of Media Logic Half a Century of Political Communication in the Netherlands. Javnost - The Public, 13(1), 25–40. doi: 10.1080/13183222.2006.11008905

Brants, K., & van Praag, P. (2017). Beyond Media Logic. Journalism Studies, 18(4), 395– 408. doi: 10.1080/1461670X.2015.1065200

Bruns, A., & Highfield, T. (2013). Political Networks on Twitter. Information, Communication

& Society, 16(5), 667–691. doi: 10.1080/1369118X.2013.782328

Bruns, A., & Moe, H. (2014). Structural layers of communication on Twitter. In A. Bruns, M. Mahrt, K. Weller, J. Burgess, & C. Puschmann (Eds.), Twitter and society [Digital

Formations, Volume 89] (pp. 15–28). Retrieved from https://eprints.qut.edu.au/66324/

Ceccobelli, D. (2018). Not every day is Election Day: A comparative analysis of eighteen election campaigns on Facebook. Journal of Information Technology & Politics, 15(2), 122–141. doi: 10.1080/19331681.2018.1449701

Ekman, M., & Widholm, A. (2017). Political communication in an age of visual connectivity: Exploring Instagram practices among Swedish politicians. Northern Lights: Film &

Media Studies Yearbook, 15(1), 15–32. doi: 10.1386/nl.15.1.15_1

Enli, G. S., & Skogerbø, E. (2013). Personalized Campaigns in Party-Centred Politics.

Information, Communication & Society, 16(5), 757–774. doi:

10.1080/1369118X.2013.782330

Entman, R. M. (1990). Democracy without Citizens: Media and the Decay of American

Politics. Oxford University Press.

Eurostat. (2020a, April 15). Individuals using the internet for participating in social networks. Retrieved June 16, 2020, from

https://ec.europa.eu/eurostat/databrowser/view/tin00127/default/table?lang=en

Eurostat. (2020b, April 15). Level of internet access—Households. Retrieved May 31, 2020, from https://ec.europa.eu/eurostat/databrowser/view/tin00134/default/table?lang=en

(28)

27

Ferree, M. M., Gamson, W. A., Gerhards, J., & Rucht, D. (2002). Four Models of the Public Sphere in Modern Democracies. Theory and Society, 31(3), 289–324. Retrieved from JSTOR.

Graham, T., Jackson, D., & Broersma, M. (2014). New platform, old habits? Candidates’ use of Twitter during the 2010 British and Dutch general election campaigns: New Media &

Society. doi: 10.1177/1461444814546728

Grant, W. J., Moon, B., & Grant, J. B. (2010). Digital Dialogue? Australian Politicians’ use of the Social Network Tool Twitter. Australian Journal of Political Science, 45(4), 579– 604. doi: 10.1080/10361146.2010.517176

Gross, J. H., & Johnson, K. T. (2016). Twitter Taunts and Tirades: Negative Campaigning in the Age of Trump. PS: Political Science & Politics, 49(4), 748–754. doi:

10.1017/S1049096516001700

Hecke, T. V. (2012). Power study of anova versus Kruskal-Wallis test. Journal of Statistics

and Management Systems, 15(2–3), 241–247. doi: 10.1080/09720510.2012.10701623

Heclo, H. (2000). Campaigning and governing: A conspectus. The Permanent Campaign and

Its Future, 1–37.

Highfield, T., & Leaver, T. (2015). A methodology for mapping Instagram hashtags. First

Monday, 20(1), 1–11.

Jacobs, K., & Spierings, N. (2019). A populist paradise? Examining populists’ Twitter adoption and use. Information, Communication & Society, 22(12), 1681–1696. doi: 10.1080/1369118X.2018.1449883

Jenkins, J. A., & Nokken, T. P. (2008). Partisanship, the Electoral Connection, and Lame-Duck Sessions of Congress, 1877–2006. The Journal of Politics, 70(2), 450–465. doi: 10.1017/S0022381608080420

Jungherr, A. (2016). Twitter use in election campaigns: A systematic literature review.

Journal of Information Technology & Politics, 13(1), 72–91. doi:

10.1080/19331681.2015.1132401

Kelm, O. (2020). Why do politicians use Facebook and Twitter the way they do? The influence of perceived audience expectations. Studies in Communication and Media,

9(1), 8–34. doi: 10.5771/2192-4007-2020-1-8

Kruikemeier, S. (2014). How political candidates use Twitter and the impact on votes.

Computers in Human Behavior, 34, 131–139. doi: 10.1016/j.chb.2014.01.025

Larsson, A. O. (2015). The EU Parliament on Twitter—Assessing the Permanent Online Practices of Parliamentarians. Journal of Information Technology & Politics, 12(2), 149–166. doi: 10.1080/19331681.2014.994158

Larsson, A. O. (2016). Online, all the time? A quantitative assessment of the permanent campaign on Facebook. New Media & Society, 18(2), 274–292. doi:

10.1177/1461444814538798

Lefevere, J. (2018). The Impact of Election Campaigns on the Nationalization of Voting Behavior in Local Elections: A Case Study of the Antwerp Local Election Campaign.

(29)

28

Lev-On, A., & Haleva-Amir, S. (2018). Normalizing or equalizing? Characterizing Facebook campaigning. New Media & Society, 20(2), 720–739. doi: 10.1177/1461444816669160 Lilleker, D. G., & Jackson, N. (2013). Political Campaigning, Elections and the Internet:

Comparing the US, UK, France and Germany. Routledge.

Newcom. (2019, January 26). Nationale Social Media Onderzoek 2019. Retrieved March 27, 2020, from

https://www.newcom.nl/downloads/Newcom_Nationaal_Social-Media_Onderzoek_2019.pdf

Norris, P. (2000). A Virtuous Circle: Political Communications in Postindustrial Societies. Cambridge University Press.

O’Reilly, T. (2005, September 30). What Is Web 2.0: Design Patterns and Business Models for the Next Generation of Software. Retrieved May 31, 2020, from

https://www.oreilly.com/pub/a/web2/archive/what-is-web-20.html

Reif, K., & Schmitt, H. (1980). Nine Second-Order National Elections – a Conceptual

Framework for the Analysis of European Election Results. European Journal of Political

Research, 8(1), 3–44. doi: 10.1111/j.1475-6765.1980.tb00737.x

Sartori, G. (1976). Parties and Party Systems: A Framework for Analysis. Cambridge University Press.

Strandberg, K. (2009). Online campaigning: An opening for the outsiders? An analysis of Finnish parliamentary candidates’ websites in the 2003 election campaign: New Media

& Society. doi: 10.1177/1461444809105355

Strömbäck, J. (2005). In Search of a Standard: Four models of democracy and their normative implications for journalism. Journalism Studies, 6(3), 331–345. doi: 10.1080/14616700500131950

Strömbäck, J. (2008). Four Phases of Mediatization: An Analysis of the Mediatization of Politics. The International Journal of Press/Politics, 13(3), 228–246. doi:

10.1177/1940161208319097

Ten Napel, H.-M. (1999). The Netherlands: Resilience amidst change. Changing Party

Systems in Western Europe. London: Pinter, 163–182.

Tweede Kamer der Staten-Generaal. (2016, January 12). Mutaties leden Tweede Kamer

sinds laatste verkiezingen. Retrieved from

https://www.tweedekamer.nl/sites/default/files/atoms/files/mutaties_sinds_laatste_verki ezingen_4.pdf

Tweede Kamer der Staten-Generaal. (2020, January 22). Mutaties leden Tweede Kamer

sinds laatste verkiezingen. Retrieved from

https://www.tweedekamer.nl/sites/default/files/atoms/files/mutaties_sinds_laatste_verki ezingen_33.pdf

Van Aelst, P., & De Swert, K. (2009). Politics in the news: Do campaigns matter? A comparison of political news during election periods and routine periods in Flanders (Belgium). Communications, 34(2), 149–168. doi: 10.1515/COMM.2009.011

Van Aelst, P., Strömbäck, J., Aalberg, T., Esser, F., De Vreese, C. H., Matthes, J., … Stanyer, J. (2017). Political communication in a high-choice media environment: A challenge for democracy? Annals of the International Communication Association,

(30)

29

Van der Steen, P. (1995). De doorbraak van de ’gewone mensen’-partij: De SP en de Tweede-Kamerverkiezingen van 1994. In Jaarboek DNPP 1994. Groningen. Vasko, V., & Trilling, D. (2019). A permanent campaign? Tweeting differences among

members of Congress between campaign and routine periods. Journal of Information

Technology & Politics, 16(4), 342–359. doi: 10.1080/19331681.2019.1657046

Vergeer, M. (2015). Twitter and Political Campaigning. Sociology Compass, 9(9), 745–760. doi: 10.1111/soc4.12294

Vergeer, M., & Hermans, L. (2013). Campaigning on Twitter: Microblogging and Online Social Networking as Campaign Tools in the 2010 General Elections in the

Netherlands. Journal of Computer-Mediated Communication, 18(4), 399–419. doi: 10.1111/jcc4.12023

Vergeer, M., Hermans, L., & Sams, S. (2013). Online social networks and micro-blogging in political campaigning: The exploration of a new campaign tool and a new campaign style. Party Politics, 19(3), 477–501. doi: 10.1177/1354068811407580

(31)

30

Appendix A

Tables

Table A1

Groups of actors

Parties Individual politicians Government party Member of Parliament Oppositional party Minister without Portfolio

Secretary of State Minister

Deputy Prime Minister Prime Minister

Table A2

Definition of timeframes

Timeframe From date Through date

Complete timeframe January 1, 2016 March 30, 2020 First-order campaigning period December 15, 2016 March 15, 2017 Second-order campaigning periods

Municipal election period February 21, 2018 March 21, 2018

Provincial election period February 20, 2019 March 20, 2019

European Parliamentary election period April 23, 2019 May 23, 2019

It is not always the same, but there is always some: A quantitative assessment of Twitter and Instagram activity in Dutch politics