Social Media Indicator and Local Elections in the Netherlands: Towards a Framework for Evaluating the Influence of Twitter, YouTube, and Facebook

  • Robin Effing
  • Jos van Hillegersberg
  • Theo Huibers
Part of the Public Administration and Information Technology book series (PAIT, volume 15)

Abstract

Social media has become a popular tool in the political landscape. As a result, it is of increased importance to evaluate social media campaigns of politicians. However, there is currently little knowledge how to measure and evaluate the influence of social media in political campaigns, especially at the local scale. This chapter is a step further towards the development of a theoretical framework and an algorithm that contributes to more reliable impact measurement of social media campaigns by politicians. The Social Media Indicator-2 framework and a related scoring algorithm are introduced to evaluate the influence of individual political candidates via social media on their social environment. The framework is tested by applying it in an empirical pilot study based on the local 2014 municipal elections in the Netherlands. We collected data for the political candidates and their parties in a pre-defined period and were able to relate scores to voting outcome. Positive correlations were revealed between social media contribution scores of politicians and their preference votes within the province of Overijssel in the Netherlands.

15.1 Introduction

In recent years, social media has become a key part of the web landscape. In 2014, Facebook has more than one billion active users. The group of active users on the micro-blogging service Twitter has increased to more than 250 million people. More than three-quarters of those who had access to the internet were using social media (Nielsen 2012; Van der Veer and Boekkee 2014). In most areas of the world, we have seen people adopting social media (Goode 2013). Social media channels such as Facebook, YouTube, Blogs, Instagram, LinkedIn and Twitter are, at the time of writing, dominating the web landscape (Lipsman et al. 2014). The use of social media is now predominantly taking place via mobile platforms such as smartphones and tablets according to Comscore (Lipsman et al. 2014). People and politicians can use these social media channels to produce and share their own so-called User Generated Content (Kaplan and Haenlein 2010). This underpins the idea that the web has continued to grow and evolve over the years towards a platform that fulfils the need of people to be socially connected with each other. This provides politicians with a unique public sphere to reach out to citizens. While there is a plethora of social media channels available, there is currently little knowledge how to measure the influence of social media within political campaign and elections, especially at the local scale. The role and impact of social media channels such as Facebook, Twitter, and YouTube at the local government are not well understood. In cases of Barack Obama and Ségolène Royal, it seemed that social media was a key success factor for electoral success (Montero 2009; Ren and Meister 2010), while in other political cases the electoral value was questioned (Jungherr 2012; Ren and Meister 2010). Furthermore, politicians easily lose interest in using social media after the Election Day. For example in Italy, there was a drastic decrease in social media use by politicians after the elections. Mario Monti, the ex prime minister, tweeted 273 times in campaign time and only three times in the following period (Di Fraia and Missaglia 2014). Others even completely disappeared from social media after the day of elections.

Yet online campaigning is not a complete replacement for the offline activities: “Linking the online community from myobama.com with offline actions such as making phone calls and organizing events, the campaign harnessed the full potential and user initiative of the online community. It made volunteer participation in the campaign easy—people could participate from their homes; they could participate with nothing but their mobile phones.” (Ren and Meister 2010:18). Some argue that social media does not lead to a revolution of political systems (D’heer and Verdegem 2014; Deželan et al. 2014). Not all political social media examples as described in literature are equally successful. For example, the YouTube attempt by Prime Minister Gordon Brown in the United Kingdom in April 2009 to address the expenses controversy was disastrous. It resulted in negative sentiments and it “served as a warning to other leaders that these tools must be handled with care” (Ren and Meister 2010:13). In Turkish elections, for example, it became clear that “social media practices of Turkish political parties and leaders are unilateral and do not support interaction.” (Bayraktutan et al. 2014:198). Consequently, local politicians have to be careful spending too much time on social media because generally they have little time for their political tasks and traditionally rely on personal forms of communication (Effing et al. 2013; Rustad and Sæbø 2013). In the end it seems that “the road from the ‘likes’ on Facebook to the ballot boxes appears to be more complicated and a way less predictable than what online activists or marketing enthusiasts might assume” (Štětka et al. 2014:242).

To our knowledge, effective frameworks and standards for measuring and comparing the influence of politicians via social media are currently lacking. In addition, more attention should be paid to the effects of social media at a local scale. This chapter is a step towards the development of a theoretical framework and an algorithm that contributes to more reliable impact measurement of social media campaigns by politicians. In order to maximize the impact of time and effort spent on social media, members of political parties could benefit from understanding the effects of these tools on their work and political campaigns. This research focuses on the most local form of the political system, the municipal level. Municipalities and their councils are relatively close to their citizens in comparison to national parliaments. Yet politicians within such local political communities know little about the effects of their social media campaigns in this local setting. Local social media campaigns are very much different from national campaigns and less media-driven. Local concerns should be an explicit part of the social media strategy in order for it to be effective (Berthon et al. 2012; Bottles and Sherlock 2011).

Studies have shown that there is a potential relationship between social media use of national political candidates on the one hand and the voting outcome on the other hand (Effing et al. 2011; Spierings and Jacobs 2012). However, most research carried out so far used very simple metrics to measure the extent to which politicians use social media. As a result, there is still a lack of understanding of the exact impact of social media, especially in the local political context. To address this, the main question of this study is:
  • How to best measure and compare social media influence levels of local political candidates during elections?

The aim of this chapter is to test a theoretical framework and a related scoring algorithm for evaluating the influence of individual candidates via social media on their social environment. A theoretical framework could help to develop a more comprehensive view on the use of social media by politicians and to make use levels comparable to each other. A scoring algorithm is necessary to obtain a feasible method for indicating the impact of someone on social media. Both a theoretical framework and a scoring algorithm are developed and tested by applying these in an empirical pilot study based on the local 2014 municipal elections in the Netherlands.

The remainder of this chapter is structured as follows. First, the method section will follow. It includes the proposed framework, its theoretical backgrounds, and the scoring algorithm. Second, the results of an empirical pilot study of the 2014 local elections in the Netherlands will be presented as a test case to validate the framework and its scoring algorithm. Third, the conclusion is presented. The final section presents the discussion and observed limitations.

15.2 Method and Introducing the Theoretical Framework

To address the research question, it is necessary to develop a theoretical framework for measuring and comparing social media influence levels of politicians. When developing such a framework, it is of importance to obtain a theoretical background. After this theoretical background, the framework will be presented including its scoring instructions.

One of the key characteristics of social media channels is personal participation (Bonsón et al. 2012; Kaplan and Haenlein 2010). Since political campaigns nowadays tend to shift towards more personal campaigns, it is important to have instruments to assess personal influence levels via social media. The theoretical background partly draws upon theories from the field of e-participation. While existing frameworks (e.g. Participation Ladders in Grönlund 2009; Medaglia 2007; Sommer and Cullen 2009) to evaluate social media influence and participation are too high level, they need to be further developed for specific evaluation of social media practices (Dylko and McCluskey 2012; Sponder 2012; Stieglitz and Dang-Xuan 2012). Others who have investigated use and impact levels of social media have focused on single social media channels, like Twitter, or have used abstract indicators (e.g. Jungherr 2012). Social media is primarily about developing and maintaining personal relationships through computer-mediated communication. Therefore, we argue that the influence of a politician depends on the size his or her network of online relationships. Online followers and friends are indicators for influence (e.g. The theory of Three Degrees of Influence of Christakis and Fowler 2009). Anyone who follows someone is in some way acknowledging, if not directly influenced, the willingness to be reached and thereby the possibility of being influenced (Sponder 2012).

This section develops the revised Social Media Indicator 2 (SMI2) as an instrument used to address the construct of Social Media Use. The SMI2 is a revised version of and evolution of the first version of the SMI, as presented and employed in earlier studies (Effing et al. 2011, 2013; Withaar et al. 2013). Limitations were recognized and addressed in the new version. The SMI2 aims to contribute to the measurement and comparison of the use of social media by politicians.

The SMI2 draws upon the theoretical backgrounds of the Macintosh participation ladder (Grönlund 2009; Macintosh and Smith 2002; Medaglia 2007; Sommer and Cullen 2009) and the ROI model described by Hoffman and Fodor (2010). The theory of Macintosh consists of a hierarchy in electronic participation levels. These steps on the e-participation ladder are e-Enabling, e-Engaging, and e-Empowering (Grönlund 2009; Medaglia 2007; Sommer and Cullen 2009). Ann Macintosh created a three-step participation ladder, which is useful in describing the participation levels of the social media phenomenon at a high-level. The model from Macintosh was considered to be most suitable for social media. The participatory forms of communication related to social media fit well with the steps given in the ladder. Furthermore, the ladder’s boundaries are better defined than in other theories. The first step on this ladder is e-Enabling. In this step, party members provide access and information to citizens. The second step is e-Engaging. During this stage, party members give opportunities to citizens to interact with them and start a dialogue. Citizens are frequently consulted on certain projects, decisions or activities, for instance through discussion forums and polls. The third step is e-Empowering. This step is about members working together with citizens, empowering the citizens with responsibilities, tasks, and opportunities to collaborate with the party’s community. This theory is not specifically designed for social media, but it is a general theory helpful in understanding the various levels within electronic participation. The first two steps of the ladder are integrated in the SMI2 framework. The Contribution aspect is linked to the stage of e-Enabling. The Engagement aspect is linked to the step of e-Enabling. The highest step on the ladder of e-Empowering, however, cannot easily be captured by these quantitative metrics, but require additional content analysis.

In recent years, influence measurement of the individual users of social media has gained importance. Yet most of the available tools are based on the principle of monitoring keywords instead of monitoring individual influence levels. While keyword-based tools are highly relevant for understanding the impact of names of brands or organizations, they cannot help to understand the specific use levels of individual politicians. Stieglitz and Dang-Xuan (2012) have created a framework for social media analytics for political purposes, and among other categories, they distinguish between keyword-based tools and actor-based tools. An example of an existing actor-based tool is Klout.com. Klout, however, has been subject to criticism, since their method of calculation lacks transparency and is arguably biased (Edwards et al. 2013; Peters et al. 2013; Sponder 2012).

The SMI2 is an instrument that contributes to the actor-based category of social media analytics. The manual collection and analysis of social media parameters (such as number of messages, likes, and shares) can be time-consuming. The reliability of the observations can be questioned because, if conducted manually, errors are easily made. This problem of inaccuracy can be solved by developing a software tool that is capable of automatically analysing the social media profiles of all selected politicians. Another advantage of developing a software tool is the capability of monitoring all the participants’ social media profiles in a more dynamic way, with multiple measurement points in time.

In order to find a software tool that could reliably measure the influence-levels of individual social media users, existing applications were compared. Well-known examples, at the time of writing, were social media analytics services such as Radian6, Teezir, Coosto, SocialMediaCheck, and Klout. We are not striving to provide a complete list of examples here, but many solutions were investigated to understand their approaches. In the following Table 15.1, some examples are shown of existing measurement tools for social media analytics, but we do not strive for completeness.
Table 15.1

List of examples of existing social media analytics tools

Tool

Approach

Radian6

Keyword based monitoring

Teezir

Keyword based monitoring

Hootsuite

Keyword based monitoring

Social Mention

Keyword based monitoring

Coosto

Keyword based monitoring

SocialMediaCheck

Keyword based monitoring

Klout

Individual network influence score

The main problem in using the available tools for our needs is that most of them are designed from the perspective of running search queries with keywords. The keyword approach has apparent disadvantages in measuring influence in networks and is useless for the comparison of the use and influence levels of political candidates. The keyword approach is problematic for at least three reasons:
  1. 1.

    These keyword-based tools scan the number of times the name of a person was mentioned on social media channels. The risk of this approach is that the results can be irrelevant and unreliable. This is because a higher number of retrieved results does not necessarily relate to a higher level of use or influence by that person.

     
  2. 2.

    There can be spam messages among the retrieved data. Since a vast amount of online posts can be categorized as spam, these posts can easily bias the outcomes (Sponder 2012).

     
  3. 3.

    There are semantic problems. The keyword approach can easily cause validity issues as keywords can have various meanings in language. Words can have multiple meanings (e.g. Ajax in mythology, as a football club or a programming framework) and can have different meanings in different languages. Also, it is difficult to reliably identify politicians who have the same or similar names as others. While there are sophisticated algorithms deployed in some of the solutions to cope with these problems, they cannot solve these issues completely.

     

While keyword-based tools primarily provide information about the larger collective mass in discussing a topic, the alternative approach, as proposed in this study, focuses on the influence from groups (e.g. political parties) and their people (e.g. political candidates).

To measure the levels of use and influence, it is necessary to have a framework of metrics. According to Peters et al. (2013:283), “Social media metrics require theoretical grounding, completeness, and a diagnostic nature; they also need to be credible to management and reliable over time.” To meet these criteria, the revised SMI2 instrument is presented in Table 15.2. It comprises of the parameters (metrics) to measure personal influence in the social media channels of Facebook, Twitter, LinkedIn, YouTube, and Google+, while taking into account current technological limitations for each of them. Additionally, theoretical backgrounds are presented to sustain the matrix in the face of future changes in the application programming interfaces of social media channels.
Table 15.2

The Social Media Indicator-2 framework

Social media channel

Contribution (e-enabling) (C)

Engagement (e-engagement) (P)

Posts (X)

Network size (N)

Interaction (I)

Word of mouth (W)

Facebook page

# of posts

□ # of likes on page

□ # of talking abouta

● # of shares of all posts

● # of comments on post

● # of likes on post

Facebook profile

● List of postsb

● # of friends

● # of comments on postb

 

● # of subscribers

Twitter

■ # of tweets

■ # of followers

■ # of reply’s

■ # of retweeted

■ # of favourites

● List of mentionsc

YouTube

□ # of video’s

□ # of subscribers

□ # of ratesb

□ # of favouriteb

□ Rating averageb

□ # of viewsb

□ # of likesb

□ # of dislikesb

LinkedIn

■ # of updatesd

■ # of first degree connectionsd

 

■ # of recommendersd

■ # of all connectionsd

Google+

□ List of activity’se

□ List of activity’se

□ # of comments per activity

□ # of +1 per activity

□ # of re-shares per activity

a Number based on activity of the last 7 days

b Limited to retrieve 25 posts per request

c Limited to retrieve 20 tweets per request with a maximum of 800 tweets in total

d Limited to a maximum of 500 kb per request

e There are undefined limits for gathering activity per request

 Data publicly available to developers

 Data available to developers depending on privacy settings of the users and in certain cases user authentication is required

 Data only available with authentication of the user

Drawing upon theories such as Macintosh (Grönlund 2009; Medaglia 2007; Sommer and Cullen 2009), Sponder (2012), Hoffman and Fodor (2010), and others, the theoretical framework is presented in Table 15.2. Based on this framework, a scoring algorithm is designed that takes into account specific social network characteristics.

The SMI2 framework, as presented in Table 15.2, offers a distinction between two levels of participation. The SMI2 scores consist of both scores for Contribution and for Participation. Based on the metrics of important social media channels (Table 15.2), we can calculate scores. Basically, there is a score for contribution (C), which is based on the number of posts (X) related to the number of personal connections—the network (N)—reached through these channels. The size of the network (N), at the individual participant level, is mostly used as the corresponding social media metric (Peters et al. 2013). Hoffman and Fodor (2010) address this as the level of exposure. For each separate channel, the score is calculated and then the sum of all channels is being calculated. For example, the contribution score for Twitter is calculated based on the number of tweets in relationship to the number of followers (Table 15.2). The participation score (P) is a sum of metrics that indicate the extent to which someone is sparking interaction (I) and the word-of-mouth (W) via social media. For example, Twitter’s interaction score is the sum of replies, favourites, and mentions. The word-of-mouth score is based on the number of retweets.

The aim is to address every social media channel from the same perspective and to deliver an equivalent value for each channel, regardless of the number of available parameters. For each of these aspects of Posts (X), Network Size (N), Interaction (I), and Word of Mouth (W), metrics can be selected for specific social media channels. “As individuals may participate in several social media like Facebook and Twitter, any social network may not be fully understood in isolation” (Peters et al. 2013:289). Every social media channel has its own type of posts and ways of replying and sharing. The aim was to address every social media channel at the same level and return a number that was independent and comparable to all social media. The API’s1 provide us with the opportunity to directly connect to the databases of various social media channels. After that a list was made with all the parameters that were relevant and potentially measurable. They were included in the matrix to provide an overview based upon the types of parameters.

Software developers can register as official developers on various social media channels such as Facebook, Twitter, YouTube, LinkedIn, and Google. As a result access is provided to some of the database functions of these social media channels. There are no additional costs in obtaining the status of a developer on these included social media channels. Table 15.3 provides a list of developer links to the API of included social media channels.
Table 15.3

List of API and links to developer documentation

Social media channel

Hyperlink to documentation for developers

[API-1] Facebook

https://developers.facebook.com/docs/reference/apis/

[API-2] Twitter

https://dev.twitter.com/docs/api/1.1

[API-3] YouTube

https://developers.google.com/YouTube/

[API-4] LinkedIn

https://developer.linkedin.com/apis

[API-5] Google+

https://developers.google.com/+/api/

The availability of each parameter in the matrix (Table 15.2) is marked by a symbol. However, even the parameters that seem available for developers, by default, may require user authentication depending on the user’s privacy settings (Vitak 2012). Since most users decide themselves which posts they share publicly, the privacy control remains in the user’s hands (Boyd and Ellison 2008). The user decides what is shared publicly and what is accessible to the public. Since publicly available data is the primary source for retrieving statistics, it is argued that the privacy consequences of these measurements are at an acceptable level. Nevertheless, researchers should always be aware that there could be potential privacy risks and violations that are the result of combination of user data.

In addition to the theoretical framework, we are proposing a scoring algorithm to transform the metrics to comparative scores. The social media scores are proposed to be not merely the total sum of each metric of the SMI2. We propose a more refined scoring formula here to take into account the Network Size (N) of Posts (X).

We calculate the Contribution Score C as follows:
$$ C= \log \left({\displaystyle \sum}_{d=1}^D{X}_d\right)\cdot {N_D}^{\left(1+\frac{1}{N_D}\right)} $$
with
  • \( d\in \left\{1,,,2,,,\dots, D\right\} \) as an indicator for a day in the time horizon

  • D as the final day

  • Xd as the number of Posts on day d

  • ND as the Network Size on day D

To calculate scores that reflect the influence of a post on its audience, the network has to be included as a leverage of the post. Therefore, it was decided to include the Network Size (N) as a weighting factor in the scoring approach. It is assumed that, for measuring influence, the size of the network is more important than the number of posts.

Table 15.4 displays a series of examples for the outcomes of such a scoring approach for the Contribution Score.
Table 15.4

Example of contribution score for various network sizes

Posts (X)

Network size (N)

Contribution score (C)

5

150

108

10

150

155

50

150

264

100

150

310

5

500

354

10

500

506

50

500

860

100

500

1,013

5

2,000

1,403

10

2,000

2,008

50

2,000

3,411

100

2,000

4,015

For the Engagement (Participation score—P) part of the SMI2, we have to count the sum of all the interactions (I) and also add the sum of the Word of Mouth (W) count for a selected period of time. The Word of Mouth count is proposed to have higher weight in the total participation score. This is because of the viral effect of sharing contents of someone.

The participation score P can be written as follows
$$ P=W+I $$
With the word-to-mouth score W given by the following equation:
$$ W={\displaystyle \sum}_{d=1}^D{W}_d^2 $$
The interaction score I is given by the following equation:
$$ I={\displaystyle \sum}_{d=1}^D{I}_d $$

However, the empirical test as elaborated upon in the next section of results is necessary to test the soundness of these proposed formulas.

In the end, the total SMI2 scores have to be calculated in the following way.
$$ \mathrm{S}\mathrm{M}\mathrm{I}2=C+P $$

Yet the formulas have to be tested with more empirical data before deciding on their exact calculation. An initial exploratory test was carried out during the Municipal elections of March 19, 2014 in The Netherlands as elaborated upon in the Results section of this chapter. The aim was to retrieve and compare social media indicator scores of the electoral candidates of the seven largest municipalities in the province of Overijssel, The Netherlands. We started with developing software to collect the necessary data from the databases of Facebook, Twitter, and YouTube. The software development was a joint initiative of a new consortium called Social Indicator. The following three parties have collaborated in developing such a software tool: Saxion University of Applied Sciences, The University of Twente, and eLabbs. eLabbs is a Research and Development software company that is closely attached to these universities and is located on the campus area. As a result of this initial test, and of the theoretical background (as discussed above), we propose the SMI2 score to be a useful indicator in measuring the use and influence of individual politicians and their parties.

The revised SMI2, as presented above, is expected to contribute to having more feasible measures to understand the social media influence of politicians. The SMI2 can easily be adapted to future social media channels that gain in popularity. This is because the SMI2 provides the theoretical foundation to select metrics of other (future) media channels that have similar functionality.

15.3 Results of the Pilot Study During Local Municipal Elections in the Netherlands

Within the Netherlands, the elections are to a large extent based on votes on individual political candidates. These candidates are at the same time representatives of political parties. These political parties present themselves as collectives. However, the voting system is based on preference votes for specific candidates of these political parties. Voters elect candidates for the council of municipalities or Houses of parliament. As our qualitative interviews with politicians earlier in 2012 suggested, voters are increasingly making choices for certain politicians instead of their parties. The personal campaigns of politicians are therefore becoming more relevant. Voters can give exactly one preference vote to the candidate of their choice. This situation is different from many other countries. As a result, the focus in this study was on individual politicians of parties and not the parties as groups.

It was decided to collect SMI2 data for the seven largest municipalities of the province of Overijssel during the campaign period of the municipal elections of March 19, 2014. The data collection method was based on the framework and scoring instructions as presented above. The threshold for selection was that the municipality should have more than 45,000 citizens in order to limit the number of subject in this pilot study. Table 15.5 shows the included 7 municipalities of the total of 25 municipalities of Overijssel that were included in this pilot study. The province of Overijssel in the Netherlands has 1,139,697 citizens in total (overijssel.databank.nl).
Table 15.5

Selected municipalities and number of citizens

Municipality

Number of citizens in 2014a

Zwolle

123,159

Kampen

51,092

Hengelo

80,957

Hardenberg

59,577

Enschede

158,586

Deventer

98,322

Almelo

72,459

aoverijssel.databank.nl

A software tool was developed (based on Java and APIs) to automatically retrieve the metrics and scores based on the framework and algorithm. Real world empirical data was gathered by applying the tool during the campaign period of the local (municipal) elections in the Netherlands held on March 19th, 2014. The data is primarily used to evaluate the feasibility of the framework and its algorithm.

The empirical results consist of a state of the field for a specific local area (Europe, Netherlands, Overijssel), a comparison of scores for a sample of individual politicians (n = 202), and its overarching political parties. In addition, to evaluate the usefulness of the SMI2 measure, correlations are calculated between the social media scores of the political parties (based on scores of individual candidates) with the election outcome (based on individual preference votes).

We have collected data for the political candidates and their parties in a pre-defined period, from March 5 to March 17, which preceded the elections on March 19. We followed the SMI2 framework and developed software with connections to the APIs2 of Twitter, Facebook, and YouTube to retrieve the appropriate metrics. The software included the metric for both the social media channels Facebook and Twitter. YouTube, LinkedIn, and Google Plus were not yet included in this pilot study due to time constraints for software developing. Nevertheless, the included social media channels already provide us with data for the purposes of this pilot study. Furthermore, as our first exploratory observations and prior research made clear, channels such as Google Plus are still scarcely used in the Netherlands by political candidates in the province of Overijssel. Nevertheless, future versions of the software will be extended with all listed social media channels of the framework. We have focused on retrieving data from the politicians who had the highest positions on their party lists. The reason was that collecting and entering the “urls” for all members on the list would be impractical and time-consuming. We limited the selection to the top five members of each party in all municipalities. As a result, the number of included subjects was 202 (n = 202). A more detailed overview of the number of included political party candidates is displayed in Table 15.6.
Table 15.6

Selected politicians as subjects

Municipality

Selected politicians in this study

Zwolle

37

Kampen

24

Hengelo

20

Hardenberg

14

Enschede

47

Deventer

30

Almelo

30

Below, in Table 15.7, we present the top 10 candidates with the highest SMI2 scores in a Table.
Table 15.7

Top 10 political candidates with highest SMI2 scores

Name

Political party

Municipality

Contribution score (C)

Participation score (P)

Total SMI2 score

Hilde Palland Mulder

CDA

Kampen

2,081

9,586,964

9,589,045

Ronald Klappe

PVDA

Kampen

4,977

2,111,572

2,116,549

Jan Brink

D66

Zwolle

5,650

1,626,871

1,632,521

Elske Mooijman

Groenlinks

Hengelo

2,157

589,127

591,284

Ayfer Koç

CDA

Enschede

7,585

475,922

483,507

Wim van der Noordt

Enschede Solidair

Enschede

629

407,097

407,726

Martin Ekker

VVD

Kampen

3,421

404,186

407,607

Silvia Bruggenkamp

Swollwacht

Zwolle

1,171

390,759

391,930

Niels Jeurink

Groenlinks

Kampen

962

150,715

151,677

Leo Janssen

Pro Hengelo

Hengelo

3,545

146,636

150,181

Of each political party that was active in more than one municipality, we could calculate the total SMI2 scores within the entire province of Overijssel. The top 10 of that ordered list is displayed below:
  1. 1.

    CDA

     
  2. 2.

    PvdA

     
  3. 3.

    D66

     
  4. 4.

    VVD

     
  5. 5.

    GroenLinks

     
  6. 6.

    Enschede Solidair

     
  7. 7.

    Swollwacht

     
  8. 8.

    SP

     
  9. 9.

    ProHengelo

     
  10. 10.

    ChristenUnie

     

In order to evaluate the usefulness and feasibility of the SMI2 scores, we have made a comparison with an independent variable. In this case, we compare the social media scores with the number of preference votes in the voting outcome. We have used SPSS to calculate Spearman’s rho correlations to explore potential relationships between the variables. Although there are many other factors that influence the number of preferential votes, these calculations could reveal whether a higher social media influence score could be associated with obtaining more votes.

Table 15.8 presents the Contribution scores in relationship to voting outcome and the correlations.
Table 15.8

SMI contribution scores (C) related to preferential votes

Municipality

Spearman’s rho correlation

Almelo

−.036

Deventer

.391

Enschede

.245

Hardenberg

.679

Hengelo

.050

Kampen

.517

Zwolle

.345

In four of the seven municipalities, we could reveal a positive relationship (Spearman’s rho > .3) between the scores for Contribution (C) and the individual preference votes. In three municipalities, such evidence could not be found.

The SMI2 consists also of a score for engagement: the Participation score. Table 15.9 presents the Participation scores of political parties and its members in relationship to voting outcome.
Table 15.9

SMI participation scores (P) related to preferential votes

Municipality

Spearman’s rho correlation

Almelo

.670

Deventer

.518

Enschede

−.100

Hardenberg

−.286

Hengelo

.267

Kampen

.383

Zwolle

−.200

In four of the seven municipalities, we could not reveal a significant positive relationship between the political party’s scores for Participation (P) and their party members’ individual preference votes. On the other hand, however, within three municipalities there was a positive relationship revealed between the Participation score of the political parties and the votes.

There are only a few minor differences between the SMI2 score calculations and the Participation score calculations, as a comparison between Tables 15.9 and 15.10 makes clear. The weight of the participation score in the total SMI2 score is very dominant. Therefore, the scoring algorithm for the Participations should be revisited and improved to resolve this issue.
Table 15.10

SMI2 total scores (SMI2) related to preferential votes

Municipality

Spearman’s rho correlation

Almelo

.610

Deventer

.473

Enschede

−.191

Hardenberg

−.286

Hengelo

.267

Kampen

.383

Zwolle

−.200

In the end, the pilot test showed that we were capable to retrieve and compare social media influence scores of a large number of political candidates during the 2014 local elections. The Infographic as displayed in Fig. 15.1 shows an example of how the result of a SMI2 study can be presented to a broader audience of practitioners.
Fig. 15.1

Infographic of the social media influence in Overijssel

15.4 Discussion

More studies and further refinement of the measure are still necessary. Especially the algorithm to calculate the participation score has to be refined to limit its weight in the total SMI2 scores. The scoring algorithm has shown to be a feasible approach for comparing social media influence scores, especially when the Participation score algorithm is adjusted.

Five limitations have to be taken into account. First, the challenge remains in how to deal with members who have more than one affiliation in their social media profiles. For example, a local politician could use his or her social media profiles for more than one affiliation. A politician could have a large network of relationships based on his daytime job. It might be justified to exclude certain members in studies where the other affiliations are expected to cause a bias in the results.

Second, because of privacy settings, it remains challenging to obtain all relevant data from the social media databases. Moreover, it could be a difficult task to keep the software up to date and to stay connected to the Application Programming Interfaces of these included social media channels. For example, the API of Facebook will change in 2015, which will have consequences for the data collection in the software.

Third, as became apparent in this pilot test, not many politicians had a Facebook Page (which is most appropriate for public persons such as politicians). Most of them created a personal/private Facebook profile instead. Due to privacy settings of the Facebook API, we were not allowed to take data from these personal profiles. As a result, the scores in this pilot study were predominantly based on Twitter activities of the candidates.

Fourth, an already existing large offline network size of a politician can influence the SMI score because the network size is often mirrored in online connections. Someone with a rich social life in the offline world tends to have more online connections as well.

Finally, the scores as provided with the SMI2 only indicate the influence, and due to advanced algorithms and advertising programs of Facebook (Edgerank) and Twitter, the order and relevance of posts and comments are continuously influenced by these companies. As a result, the real number of followers reached by each post is smaller than the total size of the network. Future version could integrate these kinds of factors in the scoring algorithm.

15.5 Conclusion

The SMI2 framework contributes to more reliable impact measurement of social media in the political field. It was useful to collect and compare social media scores of a large number of politicians during the local elections in the Netherlands. In a first pilot test, the scoring algorithm for the Contribution score (C) seems to be quite appropriate to compare influence scores. We were already able to relate the social media contribution scores to preference votes and reveal positive correlations in certain municipalities. A positive correlation makes clear that politicians and parties that had a higher influence score based on the SMI2 also received more preference votes during elections. Our measure can potentially have predicting value in future elections. However, other factors influence voting outcome as well and social media is not solely responsible for the voting outcome. For example, some parties were in a losing spiral during these elections in the Netherlands and using social media seems not to change public sentiment. When the public sentiment regarding a certain party is negative, social media does not help greatly to change that.

The SMI2 measure comprises of a set of social media metrics that can be used and revised in many types of research where it is important to compare the social media influence levels of politicians. The scoring algorithm and metrics of the SMI2 can be revised in future studies to relate the independent variable of social media influence to various dependent variables. The SMI2 can be adjusted to specific situations in various countries and can be updated quite easily as social media channels change in popularity over time.

Software companies can apply the SMI2 measure in new software solutions to monitor and measure both the impact and interaction caused by social media use. This provides them with new business opportunities. This contributes to more justified solutions in the field of social media analytics. In preparation of this research, a consortium was initiated to develop a software tool for this end. The open and scientific approach of measuring and comparing use levels of social media can provide politicians with a more reliable approach to social media monitoring tools and social dashboards. This can help to gain knowledge about effective social media campaigns and strategies.

Footnotes

  1. 1.

    Application Programming Interfaces, open, public or semi-open interfaces to access functions with a programming language for accessing social data from the social media databases.

  2. 2.

    Application Programming Interfaces.

Notes

Acknowledgments

The research projects mentioned in this chapter are initiated, supported, and funded by the Social Indicator Consortium as a collaborative project from Saxion University of Applied Sciences, University of Twente and eLabbs in Enschede, The Netherlands.

References

  1. Bayraktutan, G., Binark, M., Çomu, T., Doğu, B., İslamoğlu, G., & Telli Aydemir, A. (2014). The use of Facebook by political parties and leaders in the 2011 Turkish general elections. In B. Pătruţ & M. Pătruţ (Eds.), Social media in politics (Vol. 13, pp. 165–199). New York: Springer.CrossRefGoogle Scholar
  2. Berthon, P. R., Pitt, L. F., Plangger, K., & Shapiro, D. (2012). Marketing meets Web 2.0, social media, and creative consumers: Implications for international marketing strategy. Business Horizons, 55(3), 261–271.CrossRefGoogle Scholar
  3. Bonsón, E., Torres, L., Royo, S., & Flores, F. (2012). Local e-government 2.0: Social media and corporate transparency in municipalities. Government Information Quarterly, 29(2), 123–132.CrossRefGoogle Scholar
  4. Bottles, K., & Sherlock, T. (2011). Who should manage your social media strategy? Physician Executive, 37(2), 68–72.Google Scholar
  5. Boyd, D. M., & Ellison, N. B. (2008). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13(1), 210–230.CrossRefGoogle Scholar
  6. Christakis, N. A., & Fowler, J. H. (2009). Connected: The surprising power of our social networks and how they shape our lives. New York: Little Brown and Company.Google Scholar
  7. D’heer, E., & Verdegem, P. (2014). An intermedia understanding of the networked twitter ecology. In B. Pătruţ & M. Pătruţ (Eds.), Social media in politics (Vol. 13, pp. 81–96). New York: Springer.CrossRefGoogle Scholar
  8. Deželan, T., Vobič, I., & Maksuti, A. (2014). Twitter campaigning in the 2011 national election in Slovenia. In B. Pătruţ & M. Pătruţ (Eds.), Social media in politics (Vol. 13, pp. 141–163). New York: Springer.CrossRefGoogle Scholar
  9. Di Fraia, G., & Missaglia, M. (2014). The use of Twitter in 2013 Italian political election. In B. Pătruţ & M. Pătruţ (Eds.), Social media in politics (Vol. 13, pp. 63–77). New York: Springer.CrossRefGoogle Scholar
  10. Dylko, I., & McCluskey, M. (2012). Media effects in an era of rapid technological transformation: A case of user-generated content and political participation. Communication Theory, 22(3), 250–278.CrossRefGoogle Scholar
  11. Edwards, C., Spence, P. R., Gentile, C. J., Edwards, A., & Edwards, A. (2013). How much Klout do you have. A test of system generated cues on source credibility. Computers in Human Behavior, 29(5), A12–A16.CrossRefGoogle Scholar
  12. Effing, R., Van Hillegersberg, J., & Huibers, T. W. C. (2011). Social media and political participation: Are Facebook, Twitter and YouTube democratizing our political systems? In Electronic participation: Proceedings of the Third IFIF WG 8.5 International Conference ePart 2011 (Vol. 6847, pp. 25–35). Delft, The Netherlands: Springer.Google Scholar
  13. Effing, R., Van Hillegersberg, J., & Huibers, T. W. C. (2013). Social media participation and local politics: A case study of the Enschede council in the Netherlands. In Proceedings of Electronic Participation, ePart 2013 (pp. 57–68). Koblenz, Germany: Springer.Google Scholar
  14. Goode, P. (2013). The value of data in an evolving digital world. Retrieved from http://www.comscore.com/Insights/Presentations_and_Whitepapers/2013/The_value_of_data_in_an_evolving_digital_world
  15. Grönlund, Å. (2009). ICT is not participation is not democracy—eParticipation development models revisited. In A. Macintosh & E. Tambouris (Eds.), Electronic participation: First international conference, ePart 2009 (pp. 12–23). Berlin, Germany: Springer.CrossRefGoogle Scholar
  16. Hoffman, D., & Fodor, M. (2010). Can you measure the ROI of your social media marketing? MITSLoan Management Review, 52(1), 41–49. Retrieved November 6, 2013, from http://www.emarketingtravel.net/resources/can you measure the ROI of your Social media marketing.pdf
  17. Jungherr, A. (2012). Online campaigning in Germany: The CDU online campaign for the general election 2009 in Germany. German Politics, 21(3), 317–340.CrossRefGoogle Scholar
  18. Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of social media. Business Horizons, 53, 59–68.CrossRefGoogle Scholar
  19. Lipsman, A., Lella, A., & Dreyer, K. (2014). U.S. digital future in focus 2014. Whitepaper comScore, comScore. Retrieved from http://www.comscore.com/dut/Insights/Press-Releases/2014/4/comScores-Latest-White-Paper-2014-US-Digital-Future-in-Focus-Available-for-Download
  20. Macintosh, A., & Smith, E. (2002). Citizen participation in public affairs. In R. Traunmueller & K. Lenk (Eds.), Proceedings of Electronic Government (pp. 256–263). Berlin, Germany: Springer.CrossRefGoogle Scholar
  21. Medaglia, R. (2007). Measuring the diffusion of eParticipation: A survey on Italian local government. Information Polity, 12, 265–280.Google Scholar
  22. Montero, M. D. (2009). Political e-mobilisation and participation in the election campaigns of Ségolène Royal (2007) and Barack Obama (2008). Quaderns del CAC, 33, 27–34.Google Scholar
  23. Nielsen. (2012). State of the media: The social media report 2012 (Public report Nielsen, New York). Retrieved from http://www.nielsen.com/us/en/insights/reports/2012/state-of-the-media-the-social-media-report-2012.html
  24. Peters, K., Chen, Y., Kaplan, A. M., Ognibeni, B., & Pauwels, K. (2013). Social media metrics—A framework and guidelines for managing social media. Journal of Interactive Marketing, 27(4), 281–298.CrossRefGoogle Scholar
  25. Ren, J., & Meister, H. P. (2010). Drawing lessons from Obama for the European context. The International Journal of Public Participation, 4(1), 12–30.Google Scholar
  26. Rustad, E., & Sæbø, Ø. (2013). How, why and with whom do local politicians engage on Facebook? In M. A. Wimmer, E. Tambouris, & A. Macintosh (Eds.), Electronic participation: 5th IFIP WG 8.5 International Conference, ePart 2013 (pp. 69–79). Berlin, Germany: Springer. Retrieved June 18, 2014, from http://link.springer.com/chapter/10.1007/978-3-642-40346-0_7
  27. Sommer, L., & Cullen, R. (2009). Participation 2.0: A case study of e-participation within the New Zealand Government. In R. H. Sprague (Ed.), Proceeding of 42nd Hawaii International Conference on System Sciences 2009 (pp. 1–9). Piscataway, NJ: IEEE.Google Scholar
  28. Spierings, N., & Jacobs, K. (2012). Marginal, but significant: The impact of social media on preferential voting. In ECPR Joint Sessions Workshop Parties and Campaigning in the Digital Era, 10–15 April 2012, Antwerp.Google Scholar
  29. Sponder, M. (2012). Social media analytics: Effective tools for building, interpreting, and using metrics. New York: McGraw-Hill.Google Scholar
  30. Štětka, V., Macková, A., & Fialová, M. (2014). A winding road from “Likes” to votes. In B. Pătruţ & M. Pătruţ (Eds.), Social media in politics (Vol. 13, pp. 225–244). New York: Springer.Google Scholar
  31. Stieglitz, S., & Dang-Xuan, L. (2012). Social media and political communication: A social media analytics framework. Social Network Analysis and Mining, 3(4), 1277–1291.CrossRefGoogle Scholar
  32. Van der Veer, N., & Boekkee, S. (2014). Social media in perspectief: Whitepaper nationale social media onderzoek 2014. Public whitepaper. Enschede, The Netherlands: Newcom. Retrieved from http://www.newcom.nl/socialmedia
  33. Vitak, J. (2012). The impact of context collapse and privacy on social network site disclosures. Journal of Broadcasting & Electronic Media, 56(4), 451–470. Retrieved November 6, 2013, from http://www.tandfonline.com/doi/abs/10.1080/08838151.2012.732140
  34. Withaar, R. J., Ribeiro, G. F., & Effing, R. (2013). The social media indicator 2: Towards a software tool for measuring the influence of social media. In M. A. Wimmer, M. Janssen, A. Macintosh, H. J. Scholl, & E. Tambouris (Eds.), Joint Proceedings of Ongoing Research of IFIP EGOV and IFIP ePart 2013: Lecture Notes in Informatics (LNI), Koblenz, Germany (pp. 200–207).Google Scholar

Copyright information

© Springer International Publishing Switzerland 2016

Authors and Affiliations

  • Robin Effing
    • 1
  • Jos van Hillegersberg
    • 2
  • Theo Huibers
    • 2
  1. 1.Saxion University of Applied SciencesEnschedeThe Netherlands
  2. 2.University of TwenteEnschedeThe Netherlands

Personalised recommendations