- Top of page
The present study focuses on how candidates in the Dutch general elections of 2010 use Twitter, a popular microblogging and social networking service. Specifically the study focuses on explaining why some candidates are more likely to adopt Twitter, have larger networks, and show more reciprocation than other candidates. The innovation hypothesis, predicting that candidates from less established and smaller parties will use Twitter more extensively, is unsupported. This suggests that normalization of campaign practices is present on Twitter, not changing existing communication practices. The findings do show that being an early adopter of these new technologies is more effective than adoption shortly before Election Day.
- Top of page
Before turning to the tests of the hypotheses, we will first look at some descriptive analyses of the dependent variables, and the relations between these. Figure 1 shows the level of adoption is 32% on average (horizontal line). There are large differences in adoption levels between the parties: the levels range from 0% for two new fringe parties Blank (i.e. list without a name) and Lacié to 83.3% for the Green Party (abbr. GL). The parties are ordered on the X-axis from left to right according their past (2006) electoral success. From party Nieuw Nederland (abbr. NN) further to the right are parties that have never been elected to parliament before3.
Figure 1. Level of adoption of Twitter by candidates per political party Note: N = 682; Gov = government party, Opp = opposition party, Fr = fringe party. The order of the parties on the x-axis is according the order as determined by the Electoral Council based on the election outcome of the last general elections of 2006. See Appendix A for full names of political parties.
Download figure to PowerPoint
Reviewing adoption rates along the x-axis, there is no clear pattern as to whether the adoption rate increases or decreases moving from left to right, suggesting that the level of adoption is unrelated to the number of seats in parliament. However, there are some political parties that show higher than average levels of adoption (CDA, PvdA, VVD, GL, D66). What is striking is that candidates from two of the three prominent parties in parliament show less than average adopter rates are considered populist parties, but also historically had strong leaders4 (leftwing SP (former) party leader Jan Marijnissen and right-wing PVV party leader Geert Wilders). On the one hand, being a populist party would suggest candidates would want to listen and connect to the larger part of the electorate, yet, the low adoption rate suggests otherwise. On the other hand, having strong leaders suggests they want to control external communication, centralizing it, and to focus on party and specifically party leader communication instead of general candidate communication. This suggests that these parties utilize personalization strategies for all party candidates to a lesser extent (Caprara, Barbaranelli & Zimbardo, 1999) as indicated by the lower adoption rate.
To further understand how the Twitter activities are interrelated we produced Figures 2 to 5. Figure 2 shows that there is a positive relation between the number of people following the candidate and the number of people followed by the candidate. However, at a certain point, saturation occurs: Even though a candidate is followed by more people, the number of people followed by the candidate is not likely to increase to the same extent or even decreases. The rate between the number of followers and following is .56: For each person that is followed by the candidate, the candidate is followed by two persons5.
Figure 2. v The relation between the number of people followed by the candidate and the number of people following the candidate. Note: linear trend r = .300, curvilinear trend r = .451
Download figure to PowerPoint
According to Figure 3 the number of reciprocal relations lags behind the number of followers (rate = .26: for each four followers one relation is reciprocated). When the number of people following the candidate (followers) is still low, the relation between the number of followers and the number of reciprocated relations is linear: the number of reciprocated relations increases roughly at the same level as the number of followers. However, when candidates have a certain number of followers, the amount of reciprocated relations levels off and subsequently decreases. This suggests that as candidates become more popular, indicated by the number of followers, Twitter becomes less reciprocal and Twitter is used more as traditional mass media instead of a social medium. An interesting observation from Figure 4 is that some candidates, who have others following them on Twitter, do not themselves follow any others. These candidates are located on the x-axis and use Twitter solely as a mass medium, merely creating a following without following others or even reciprocating.
Figure 3. The relation between the number of followers and the number of reciprocated relations. Note: linear trend: r = .386; curvilinear trend r = .531
Download figure to PowerPoint
Figure 4. The relation between the number of people the candidate follows on Twitter and the number of reciprocated relations. Note: linear trend: r = .945
Download figure to PowerPoint
Figure 4 indicates there is a near perfect relation between the number of people the candidate follows and the number of reciprocated relations. The correlation of .945 indicates that the number of reciprocated relations can be predicted almost perfectly, based on the number of people followed by the candidate, or vice versa. The rate between the number of following and reciprocal relations is .56: For each two people a candidate follows one is reciprocated.
Figure 5 shows that the more followers a candidate has, the more tweets the candidate sent out. The rate is 1.90: For each follower, the candidate sends out two tweets. Whether increased tweeting leads to more followers, or more followers leads to more tweets, or whether that these processes take place simultaneously is not testable with these data. Strikingly, some candidates never tweeted, and yet had quite a large group of followers. These candidates used Twitter solely as a social networking tool rather than a communication tool. These Twitter accounts and those not following others can be considered dormant accounts: called into life at one point, but no longer used actively.
Figure 5. The relation between the number of people followed by the candidate and the number of tweets sent by the candidate. Note: linear trend: r = .324; curvilinear trend r = .394
Download figure to PowerPoint
Explaining activities on Twitter
Table 1 and Table 2 present the findings that aim to provide an answer to why candidates from different parties use Twitter in various ways. Table 1 on the adoption of (i.e. subscribing to) Twitter shows that candidates from established (i.e. older) parties are not more or less likely to subscribe to Twitter than those of younger parties. Party ideology is also unrelated to having adopted Twitter as a campaign tool. As such it appears that whether or not being a member of an established party does not affect the likelihood of subscribing to Twitter.
Table 1. Logistic regression analysis of subscribing to Twitter on party and candidate characteristics
|Party age (in years)||−.001|
|Candidate rank||−.017c |
|External shock (relative change in seats) Internal shock||−.207c |
|(leadership uncertainty no = 0, yes = 1)||−.162|
|Gender (female = 0, male = 1)||.166|
|Candidate's age (in years)||−.064c |
Table 2. Negative binomial regression analysis on key concepts of candidates' use of Twitter
|Intercept||5.404c ||5.450c ||3.959c ||3.224c |
|Party age (in years)||−.004||.020c ||.003||.006|
|Candidate rank||−.005||−.056c ||.011a ||.003|
|Shock|| || || || |
| External (relative change in seats)||−.050||.104a ||−.109a ||−.097|
| Internal (leadership issues no = 0, yes = 1)||−.231||.518a ||−.875c ||−.796b |
|Gender (female = 0, male = 1)||.042||.161||−.312||−.376|
|Candidate's age (in years)||−.008||.014||−.009||−.006|
|Twitter subscription (in years)||.437b ||1.523c ||1.020c ||1.259c |
|Likelihood Ratio χ2 ||7.276||154.706||57.657||66.030|
Candidates from parties that had lost seats in the last general elections of 2006 were more likely to subscribe to Twitter, suggesting they sought new ways to reach out to voters. Also, candidates from the parties with leadership problems were neither more nor less likely to subscribe to Twitter than those from other parties.
Candidates that were given lower priority by their party (i.e. having higher rank number) are less likely to subscribe to Twitter. Male and female candidates (RQ2) do not differ as to the likelihood of subscribing to Twitter. Older candidates appear less likely to subscribe to Twitter.
In Table 2 negative binomial regression analyses are presented in order to understand which factors explain various activities on Twitter.
Trying to explain why candidates use microblogs seems difficult. Of all the explanations proposed, only the length of the subscription shows a statistical significant effect: The longer the candidate has been signed up, the more the candidate tweeted during the 40-day period prior to Election Day. This suggests that having more experience is reflected in more tweeting activities later in time. All other explanations (party characteristics, personal characteristics) show no significant effects.
Candidates from older parties have more followers than those from younger parties, suggesting that established parties know best how to utilize new technology to reach out to people, or at least are able to bind an existing offline supporter base utilizing Twitter. Put differently, people know how to find candidates from older, established parties more easily on Twitter, in part due to the general mass media attention these high profile candidates receive. Ideology is unrelated to the number of followers: Candidates from the left-wing and the more progressive parties do not have significantly more or less followers online than the right-wing and conservative candidates. Candidates ranked higher by their party (lower rank number) have a significantly larger following than those ranked lower. This refutes the idea of intraparty competition in favor of less prioritized candidates competing with higher prioritized candidates for votes. It implies that the party leader and those candidates most likely to be elected to parliament have the most followers on Twitter.
Male candidates do not significantly differ from their female counterparts, regarding their number of followers. Age is unrelated to the number of followers, although there is a slight tendency for older candidates to have a larger following, suggesting that seniority and reputation is important in attracting followers. Candidates from parties with internal leadership issues did have a larger following, perhaps due to the additional media attention these problems created. Furthermore, candidates from parties that gained seats in the last general elections also had a larger following.
Following and reciprocal relations
The regression models for following and reciprocal relations are expected to produce similar results due to the large correlation between the two (r = .945, p<.001). As to the extent the candidate follows others or reciprocates a relation on Twitter, the findings in Table 2 show that party age and ideology are unrelated.
Candidates that had to deal with internal leadership problems followed fewer people on Twitter. Whether these candidates were more occupied resolving these internal party issues than bothering with soliciting people's support using social media is unclear. While the candidates that suffered an internal shock followed fewer people, candidates from parties that had lost seats in the last general elections followed more other people on Twitter. To turn around the downward trend candidates might have decided to utilize new media to communicate with the constituency and seek their support, or those who did win in the last elections became complacent and did not put much effort in retaining the electoral advantage.
Candidates who were given a lower propriety by the party did not follow other people more or less frequently. Male and female candidates are equal as to the number of other people they follow on Twitter. The same holds for age: The older candidates do not follow more or less others than the younger candidates do.
Summarizing the results in terms of the hypotheses, we find that hypothesis 1 on establishment (i.e. party age) receives little support, only for the number of followers. Hypothesis 2, predicting that less prioritized candidates use Twitter more extensively, finds no empirical support. Rather, higher prioritized candidates use Twitter most extensively, indicating there is little to no intraparty competition between candidates. As for ideology (RQ1), there appears to be no significant relation whatsoever with dependent variables. Hypothesis 3 on the external shock of having won or lost in the prior national elections, is supported regarding the adoption of Twitter and for following others. It is not supported regarding microblogging, number of followers, and reciprocal relations. Hypothesis 4 on leadership issues as internal shock shows a significant predicted effect for the number of followers, but a contradictory relation for the numbers of following and reciprocal relations. Hypothesis 5 on the candidate's age is supported for adopting Twitter, but for all other Twitter activities it receives no support. Furthermore, there are no differences between male and female candidates in how they use Twitter (RQ2). Candidates that adopted Twitter at an earlier stage are also the most active in terms of microblogging (tweeting) and online social networking.
- Top of page
This study focused on candidates' adoption of micro-blogging and online social networking activities using Twitter for e-campaigning during the general election of 2010 in the Netherlands. In particular, this study focused on whether Twitter as a microblogging and social networking service was able to equalize the political playing field, or whether the use was merely reflective of offline power structures. The findings show that the overall adoption level of Twitter among candidates increased sharply within one year: In the 2010 elections 32% subscribed to Twitter, compared to 12.5% in the 2009 EP elections (Vergeer, Hermans & Sams, forthcoming). Whether the 32% adoption rate constitutes a critical mass to entice the remaining politicians to adopt Twitter is unclear. We also see that “social” as in “social media” is limited: As candidates become more popular (i.e. have more followers) they tend to follow others or even reciprocate these relations less likely.
As for one key feature of SNSs, i.e. setting up and maintaining online networks, using Twitter at an early stage, long before the election campaign starts, ensures the candidate of a larger network to which to send messages. This is a network that both serves as an audience and also as a re-layer of messages to others through the use of retweeting. Over time, it is more likely these Twitter networks grow, instead of declining. Whereas on Facebook de-friending seems to be a practice, on Twitter this seems less likely for two reasons. First, de-friending by politicians or candidates seems, from a strategic point of view, less likely: Candidates need a large following to create an audience and to gain support. Second, Twitter is more liberal in setting up relations than Facebook: whereas Facebook requires mutual consent to set up a relation, Twitter allows for purely one-directional relations. However, Lewis and West (2009) find that youngsters on Facebook ignore friendship requests rather than explicitly turn them down. This practice of merely ignoring friend requests does not decrease online network sizes, merely slows down its growth.
Microblogging activities seem difficult to explain using the set of predictors in this study. Common attributes used in studies in political communications did not show significant differences, apart from being an early adopter of microblogging as indicated by the Twitter sign-up date. Other factors, such as psychological traits might play a role in the use of SNS. Previous research by Orchard and Fullwood (2010) suggests that introversion, extraversion, and neuroticism, three of the Big Five personality traits, play a role in the use of Twitter. For instance, previous research suggests that introversion is positively associated to more preference for online computer-mediated communication (CMC), whereas extraverts prefer face-to-face communication. Also, neuroticism seems to be negatively related to online discussions, suggesting that emotionally stable persons - and politicians need to be, or at least are expected to be, stable emotionally - like online discussions.
Even though these factors could play a role in the use of Twitter, studying psychological traits of politicians might be very difficult to organize, requiring a survey among politicians and candidates. It is highly unlikely politicians and candidates would agree to participate in a study to assess their psychological make-up. One of the few studies shows that female politicians can be characterized as being more extravert and open than voters (Caprara, Francescato, Mebane, Sorace & Vecchione, 2010). Alternatively, analyzing the content of their (online) communications, an unobtrusive way of measuring (e.g. lexical tradition of measuring; cf. Wiggings, 2001), might shed more light on their psychological make-up. Still, this is not without problems, because one needs to make sure that the author of the texts is indeed the politician and not, for instance, an assistant. This holds for official documents, but also for the tweets being sent out.
A final remark is that the use of Twitter as a campaign tool is only one of the many tools that are available to candidates. Apart from the traditional channels, such as television, newspapers and radio, candidates can use personal profiles page on the party website or opt for a personal website. As such, Twitter is not and will not be expected to be the single most important communication channel. Yet, it enables candidates to connect very directly on a regular basis to those that for some reason are interested in the candidate. Therefore, to fully understand today's online political campaigning, not only specific modal campaigning is needed (as in this study), also multi-modal research (e.g. websites, Facebook, Hyves) is required.
The implications of these findings from a political perspective are that, even though new campaign instruments such as new media technology may appear, to attribute any changes in the power structure to these tools is very hard. Furthermore, the notions of “birds of a feather flock together” (McPherson, Smith-Lovin & Cook, 2001) and “preaching to the converted” (Norris, 2003) suggests that the people that are reached with social media are those already interested in politics and a specific political party or candidate. If the electoral gain is to be achieved by convincing the floating voters social media campaigning might need to cross the boundaries of their homogeneous social networks and strive for pluralistic networks by focusing on different segments in the population. A similar argument has been made by Chadwick (2009) on hyperlinks on government websites, indicating that by providing these links the government's policymaking is pluralistic and inclusive. This, of course, could be extrapolated to political parties as well.
Furthermore, even though this study has shown no clear indication of innovation or even equalization, Twitter (and other online services such as YouTube and Facebook) may well have other effects. For instance, candidates' access to these communication tools allows them to campaign at an individual level, less attached to the overall party campaign. This might have consequences for parties' campaign strategies. These individualized campaigns coincides with an increasing trend in politics to personalize politicians and candidates, particularly to present candidates as regular people, maybe even likeable, and not as the distant politician far away in the countryás capital. The use of social media might be a means to achieve this goal. It is also important to note that there are good reasons to assume that process of normalization and equalization may be dependent on institutional factors as well. Anstead and Chadwick (2009) note that the type of electoral system (cf. Farrell, 2001), campaign finance (cf. Anstead, 2008), or the degree countries differ with respect to whether or not a few parties dominate the political playing field (cf. Rae's fractionalization, 1968) might affect the role of the Internet in electoral campaigning. Because these institutional factors are conceptualized at the country level, a cross-national comparative research approach show could be very productive in establishing the role of these factors.