Like me if you like me
Personalization effects of political communication
on Facebook
Dominique Prescher 10607374
Master’s Thesis
Graduate School of Communication Political Communication
Supervisor: Dr. Marjolein Moorman
Abstract
Adopting a two-wave longitudinal content analysis of Facebook messages in the context of the 18th German federal parliament, acknowledging variations during election campaigns, this study provides further insights into the effect of personalization of political coverage in general and of political communication on social media in particular. Analysing 3334 Facebook messages of 61 members of parliament (MP) the study reveals that
personalized messages generate the most “likes”. Indeed, the perception of personalized messages depends on contextual differences of both politicians and their Facebook audience. In contrast to findings of personalization studies in traditional media, followers of MP’s Facebook profiles are likely to be homogenous by party lines, resulting in a positive depiction of privatized messages due to perceived similarities. However, a silent majority exists.
Furthermore, contextual differences influence success of political communication on Facebook in general. Especially politicians younger than 35, in opposition to the generation older than 65, seem to be capable to carry out a “like”-oriented approach.
Introduction
In today´s western societies the influence of the media on the individual, the society and generally on public opinion is almost undisputed. It is widely acknowledged that the media sets the boundaries of our thinking and what we perceive as reality, especially outside our own experiences (Jamieson & Campbell, 2000). What we do and not do, buy, love, hate is to a great extent an outcome of what we are shown by the media. For decades success of political communication in traditional media was highly dependent on the will of journalists to let political communication be successful. Before reaching the public, political messages were often going through a journalistic filter of affection or objection (Neveu, 2002).
Within the last decade political communication with the public changed. Technological advancement brought a new player on the field, making individuals capable to reach a
decisive part of the public: social media. Created in 2003, MySpace was the first social medium with a perceptible public influence, yet Facebook, created in 2004, soon dominated the western market with 1.32 billion users in 2014 (Tagesschau.de, 2014). During the 2006 US midterm election campaign Facebook firstly opened its communication channels to politicians (Williams & Girish, 2012) and is since an increasingly important instrument to spread political messages, especially in the US election context (Bronstein, 2013; Gulati & Williams 2013).
Consequently, the vast amount of research, with respect to social media as a political communication tool, focuses on the US context. Probably the most studied case in this regard is the 2008 Obama presidential campaign, which was the first election campaign actively using social media and creating a benchmark in the field. The campaign used social media as an interactive platform on which volunteers and supporters were able to codetermine
campaign messages (Gibson, 2012). Obamas decentralized, interactive and collaborative communication approach had two effects: A direct and an indirect “two-step mobilization effect”, meaning that the political communication on social media activated supporters who
then mobilised their on- and offline networks (Norris & Curtis, 2008).
The success of the 2008 Obama social media campaign attracted other politicians around the world to utilize the mobilization effects of social media, mostly of Facebook. In contrast to most of traditional campaigning tools, the political adoption of Facebook does not require any financial resources and is “free to air” (Chen & Walsh, 2010). It is a channel that offers new ways of communication between the political and the civic sphere regardless of differences in budgets. Facebook offers the possibility to directly communicate with citizens and furthermore, to receive an immediate reaction via the “like” button. Consequently, scholars observed an increasing and by 2014 almost universal adoption among individual politicians of social media in general and Facebook in particular (Gulati & Williams, 2013).
Due to their increasing relevance among politicians, scholars became interested in social media’s broader political impact. In a series of articles political scientists found that the size of social media networks is capable to predict election outcomes (Metaxas, Mustafaraj & Gayo-Avello, 2011; Cameron, Barrett & Stewardson, 2013; Franch, 2013. Subsequently, political scientists discovered a positive correlation between the accumulated “likes” of a politician´s Facebook messages and actual election outcomes (Giglietto, 2012). This might not only have consequences on public opinion measurements, but also on the political relevance of social media. It seems that the amount of “likes” of a message of a politician is related to his or her election result and eventually, to the success of political communication. Arguably, the success of political communication on Facebook is measurable.
Research on personalization and social media. The adoption of social media by individual politicians is a subject of interest. An individual approach of political communication is part and parcel of presidential systems, yet novel in essentially party focused multi-party systems. In multi-party systems the source and focus of political communication are parties themselves and top-politicians (McAllister, 2007). Facebook challenged this conventional approach by
shifting the focus towards individual politicians from all levels of politics. Through Facebook politicians are independently capable to mobilize and to publish messages directly reaching the interested public. Subsequently, the centralized communication monopoly of parties in multi-party systems is extended by a decentralized, individual approach to political
communication (Enli and Skogerbø, 2013).
Interestingly, the rise of a decentralized, individual approach to political communication on social media falls in a scientific debate about the existence of a
personalization trend of political coverage of traditional, western, multi-party system media. According to this debate two main factors are responsible for an increase of personalized political coverage: Firstly, because traditional ties between parties and voters diminished, resulting in citizens who base their voting decision increasingly on individual differences of the candidates (Jebril, Albæk & de Vreese, 2013). Secondly, because of the changing media environment that represents politics as a competition between individual politicians (Schulz, Zeh & Quiring, 2005). With regard to the latter, the scholars’ emphasise clearly lies on
television rather than social media. So far, the debate did not provide coherent evidence, since both an increase (Reinemann & Wilke, 2007) and stagnation (Kriesi, 2012) of a
personalization trend was observed in similar contexts. Potential consequences of a trend are disputed as well. Some empirical evidence suggests that personalization in political coverage decreases cynicism among those less interested in politics (Jebril et al., 2013), while other scholars see parliamentary democracy endangered by personal political authority and personalized mandates, calling for drastic institutional reforms (McAllister, 2007).
Altogether findings on personalization trends of political coverage in multi-party systems are at best mixed. The new player on the field, social media, might help to give new insights into both, the actual trend and potential consequences and furthermore, expand the field to new developments in political communication.
Personalization on Facebook. A Facebook profile of a politician is inherently personalized. The politician is the assumed source and the focus of the messages published. Nevertheless, the content of the posts can differ and could ignore personalization. It is therefore possible to measure personalization of messages in a similar way as personalization in traditional
political coverage, with a longitudinal content analysis. Furthermore, due to the positive relationship between “likes” and votes (Giglietto, 2012) it is possible to measure the
communication success of each Facebook message and thus the success of messages that are personalized. For the sake of convenience, Facebook message success, the relative amount of “likes”, is from here on abbreviated FMS.
Scholars are discordant about the consequences of personalization in political coverage. At the same time, Facebook gives the opportunity to measure the success of personalized political communication. This study aims to provide insights into the effects of personalization of political social media messages in western multi-party systems. For this purpose the context of Germany was chosen, since it is a large multi-party system with an almost universal adoption of Facebook by federal members of parliament (MPs) (Elter, 2013). Accordingly, the first research question is as follows:
RQ1: Does personalization of Facebook messages of German members of parliament influence FMS?
As elaborated above, previous research on personalization did not focus on social media but on political coverage of traditional media. Jebril et al. (2013) showed that personalized political coverage increases cynicism among politically interested, but decreases cynicism among politically uninterested citizens. Politically uninterested citizens are not likely to expose themselves to Facebook profiles of politicians, in fact, gathering political information is a positive predictor of political behaviour (Gil de Zúñiga, Jung & Valenzuela, 2012). Accordingly, personalized Facebook messages are likely to have less FMS than
personalized messages. However, the different audience of social media with respect to traditional media does not allow this interpretation. Siri, Melchner and Wolff (2012) found, without offering empirical data, that users of Facebook prioritize personalized messages over rational political communication, because the audience tends to be homogenous with respect to party affiliation. The lack of empirical data for both Facebook messages and the audience leaves room for interpretation of the study. Furthermore, a clear conceptualization of
personalization is missing. Yet this is necessary, because the concept of personalization is multidimensional.
In a conceptual review of the literature Van Aelst, Sheafer and Stanyer (2012)
established two dimensions of personalization in relation to traditional news content. The first dimension, individualization, describes the shift of media focus from parties as central actor in politics towards the individual politicians. Within the first dimension one can conceptually distinguish between either focus on politicians in general or on the top-leaders. Privatization, the second dimension, refers to the shift in media focus from politicians in their official, public role towards politicians as private persons. Again, two sub-dimensions can be found: Firstly, focusing on personal traits of politicians and secondly, on their private life.
According to research on personalization in traditional media politicians might strategically use personalization in order to appeal to the voters with perceived similarities (Garzia, 2011). Perceived similarities are barely situated on a political, but on a private level. E.g. if a politician shows that he or she walks a dog, her or she appeals to voters that like dogs. Hence, an effect of privatized messages on FMS can be assumed.
H1: Privatized Facebook messages have higher FMS than messages that are individualized or not personalized.
Demand and supply of politicians´ Facebook messages. As mentioned above, the influence of personalization might differ among groups with divergent levels of political interest.
Following the argumentation of previous research, not every Facebook user is likely to depict personalized messages positively. The same might be true for messagesds that are not
personalized. Consequently, other factors than personalization might have an influence on FMS. Furthermore, these factors might interact with personalization.
In pluralistic countries, like Germany, citizens differ and so do their representatives. Consequently, both the politicians and respective Facebook followers may come from different backgrounds. Zittel (2009) concluded that the decentralized communication
approach on social media is better suited to react to the information demands of citizens than traditional media. Hence, this study looks into factors of demand and supply of political information on Facebook, called by the umbrella term contextual differences.
One of the demand factors is certainly political interest (Jebril et al., 2013). Many studies propose education as an indicator for this factor. According to Siever (2012), 90.2% of German citizens with higher education use the internet, compared to only 60.5% with lower education. Subsequently, politicians competing in electoral districts or regions with lower educated inhabitants might have a disadvantage on social media, resulting in smaller FMS.
Furthermore, citizens in different periods of life might react different to Facebook messages. The concept of the digital divide implies that German citizens younger than 35 are more likely to gather political information on social media than other age groups (Siever, 2012). This could result in a disadvantage for politicians from regions with numerous elderly people and an advantage for those representing many younger voters.
On the supply side not every politician is likely to provide personalized messages that appeal to perceived similarities with the voters. It might be the case that some politicians do not even see a purpose in trying. The German case offers an interesting distinction in this regard. Candidates of federal elections compete either for a direct, personalized mandate or a party based, listed mandate. A direct candidate is considered a stronger parliamentarian since he or she is a direct representatives from an electoral district, whereas a county-based, listed
candidate is considered a representative of a party (Gschwend & Zittel, 2014). Consequently, direct candidates might put more emphasis on a personalized social media strategy than listed candidates in order to fulfil their representative function.
The case of Germany further offers a dynamic political environment. Since the
reunification of Germany the traditional dominance of the two largest parties, CDU and SPD, is decreasing and developed towards a pluralistic party system. Additionally, due to the massive decline of party memberships, Germany´s political environment is affected by increasingly intense electoral competition with increasing budgets involved (Korte, 2009). Wattal, Schuff, Mandviwalla and Williams (2010) found that competition is an indicator of the amount and frequency of social media tools a politician uses. Challengers are more likely to use more social media tools more often in order to compensate vis-a-vis the incumbent candidate. Therefore the candidacy status, whether a politician is an incumbent or a challenger, could have influence on his or her FMS. Challengers are most active in the political marketing competition on social media (Enli & Skogerbø, 2013) and might be more professionalized in social media than incumbents.
These contextual differences can be complemented by characteristics derived from Germany´s party system that is, as the state itself, organized in a representative democratic order. Consequently, politicians have differing internal party status´. Due to the federal structure of the state members of the German parliament are not only top level politicians. Previous studies suggest that in an election campaign context the status within a party is an indicator of professionalization of political communication (Strömbäck, & Kiousis, 2014). In the case of social media communication, a higher internal party status means a higher
likeliness to employ a social media officer. Accordingly, those with a high internal party status might have an advantage over lower level politicians with regard to FMS.
Furthermore, the question arises whether there is a difference in FMS among members of the five parties represented in the German parliament. The traditionally combined fractions 8
of the federal CDU (Christian Democratic Union) and the purely Bavarian based CSU (Christian Social Union) represent with 311 seats the largest portion in the 18th Bundestag. 193 seats are owned by the SPD (Social Democrats), 64 by Die Linke (Socialists), and 63 by Bündnis 90/Die Grüne (Greens). The differences between the parties with regard to social media adoption are marginal (Elter, 2013). An exception is the Green party, which is due to its history and its traditionally inherent grass-root organization more active on social media than other parties in the German parliament (Vergeer, Hermans, & Sams, 2013). As a consequence members of Bündnis 90/Die Grüne might be more professionalized in social media, thus pursuing a “like”-oriented approach.
During election campaign times both parties and individual politicians increase their political communication efforts. The same is true for political communication on social media in general and Facebook in particular, which is highly campaign sensitive. More specifically, this means that political communication in times of election campaigns is higher, in terms of message frequency, compared to non-election campaign times (Enli & Skogerbø, 2013). Since during election campaigns politicians aim to mobilize voters, their efforts on Facebook might increase in professionalization, influencing FMS.
These contextual differences on the supply, the politicians, and the demand side, the voters, are assumed to have an influence on FMS. Therefore, the second research question this thesis tries to answer is:
RQ2: Do contextual differences of German members of parliament influence FMS?
Additionally, contextual differences and personalization might interact with respect to FMS. E.g., since citizens interested in politics are more likely to show an increase in cynicism when exposed to personalized political news coverage, personalized Facebook messages of
politicians representing higher educated regions might be less successful than those from regions with less educated inhabitants. Therefore, an interaction of both sets of variables
needs to be studied:
RQ3: Do personalization of Facebook messages and contextual differences of German members of parliament interact with respect to FMS?
Methodology
In order to answer the research questions the Facebook profiles of German members of the 18th federal parliament (MPs), the Bundestag, are chosen to be the unit of analysis. The units of observation are the individual Facebook messages of the MPs.
The research design chosen joins other studies in the field of political communication on the internet and studies on social media: longitudinal content analysis. Two time frames are chosen in order to avoid a possible threat to validity. The first time frame in which
observations are made are the 31 days running up to the German federal election 2013, 22nd of August to 22nd of September 2013. A second time frame of 31 days was chosen randomly after the election (3rd of April to 4th of May, 2014).
First wave of content analysis. A two-wave content analysis was conducted. The complete codebook can be seen in Appendix H. The first wave coded all 631 MPs of the German Bundestag according to seven characteristics. Data was gathered on the official websites of the party fractions1, official data from the federal election supervisor (Bundeswahlleiter)2 and from official data from the federal institute of demographic research (Bundesintitut für Bevölkerungsforschung)3.
1) Age
The age of the politicians was coded in 5 cohorts, whereby the first represent the strata of
1
CDU/CSU: http://www.cducsu.de,; SPD: http://www.spdfraktion.de; Die Linke: http://www.linksfraktion.de ; Bündnis 90/Die Grünen: http://www.gruene-bundestag.de
2 http://www.bundeswahlleiter.de 3 http://www.bib-demografie.de
10
digital natives, the politicians younger than 35. The following cohorts consist of each 10 years and end with the fourth cohort, politicians older than 65.
2) Party
Party affiliation was coded according to the five parties present in the Bundestag after the 2013 election: CDU, CSU, SPD, Die Linke and Bündnis 90/Die Grüne
3) Mandate status
Whether a politician was elected as a direct or a listed candidate was coded respectively 4) Candidacy status
The dichotomous, nominal variable “candidacy status” was coded 1 if a politician was a challenging candidate before the election and 0 in case the politician already owned a seat in the 17th parliament.
5) Internal party status
In order to be coded as having a high internal party status, a MP needs to, within one of the two time frames, hold an elected office within his or her party on a federal or county level and/or hold an elected office within the parliament or government. An elected office is a position on a federal or county level, that needs to be preceded by a democratic election within either a party, or the federal parliament, or federal parliamentary groups. Furthermore, positions on the government level are considered high internal party status’.
6) Digital natives
Direct mandates represent certain electoral districts and listed mandates are elected by county lists. An electoral district or a county is coded inhabiting many digital natives, if the average number of people younger than 35 exceeds the national average of 36.6% (Bundesinstitut für Bevölkerungsforschung, 2014). Regions with fewer digital natives are coded respectively.
7) Education
Electoral districts and counties inhabiting more than the national average of 33.9% owning a diploma from German secondary school qualifying for university admission or matriculation
(“Abitur” or “allgemeiner Hochschulabschluß) (Bundesinstitut für Bevölkerungsforschung, 2014) are coded 1, high education. Regions with less graduate degrees are coded respectively. Additionally, the gender of MPs was included in the codebook.
In order to test the reliability of the coding process 2 additional coders coded, after a brief training of 15 minutes, a sub-sample of 20 politicians each. Krippendorff´s Alpha was chosen to calculate inter-coder reliability, resulting in a perfect reliability of α=1.0 for the first wave.
Second wave of content analysis. A random sample of 61 MPs was taken and a total number of 3334 Facebook posts analysed in the second wave of content analysis. Firstly, Facebook profiles of all 61 MPs were looked up, based on links on official websites of the respective politicians and their parties. Each post originating from the politicians on their respective Facebook profiles dated within the two time frames, were coded (N=3334). The codebook for the second wave focused on personalization and FMS.
Within the nominal variable “personalization” an individualized post was coded 1. An individualized post means that one or more of the following conditions apply:
- The post is explicitly written from the perspective of the politician - Usage of first-person pronouns
- The name of the politician is included in the text or audio/visual elements - The politician is subject of audio/visual elements
Code 2 was applied if a post contains private elements. Private elements are prioritized, meaning that if the following condition applies, privatization is prioritized over
individualization.
- The main message of a text or audio/visual element concerns the private life of the politician, meaning content that is not explicitly related to the work of politicians Since the borders between private and political life tends to be blurry, it is necessary to further
explain the conditions. If the main message of a post is drinking a morning coffee in the office, it is a privatized post. The main message is not focused on the work of the politician, but on having a coffee. If the main message of a post is drinking a coffee with citizens at a campaign event, it is not a privatized post. In this case the focus of the message is on the campaign and on the fulfilling of a representative function. The perceived main message of a post, interpreted by the coder, according to the conditions above is the pivotal reason for assigning a code to a post.
Lastly, if none of the conditions above apply, the post was coded as not personalized. As measures of success, the number of “likes” of each message is divided by the total number of “likes” of a respective Facebook profile, at the time of coding (21st of December, 2014 to 10th of January, 2015). Popular politicians, or politicians with more total profile “likes”, are expected to have more message “likes” than those with less total profile “likes”. By dividing message “likes” by total profile “likes” a relative measure of success was established, ruling out a potential bias of popularity. For reasons of convenience the relative measure is multiplied by 100.
As wave 1, the second content analysis wave was also tested for reliability. After a brief training of 20 minutes, three reliability coders coded a sub-sample of in total 100 messages of a random sample of 10 MPs. Again, Krippendorff´s Alpha was chosen to calculate inter-coder reliability, resulting in a strong reliability of α=0.87.
With the help of the statistical software SPSS several multiple regressions were conducted. In order to conduct regressions, the variables which are not dichotomous (age, party affiliation, personalization) were recoded into dummy variables.
Appendix A shows the model of the research.
Results
The random sample of 61 MPs fairly reflects the characteristics of members of the German 13
Bundestag and is considered representative. Appendix B shows a table that compares characteristics of both the sample (n=61) and the total population (N=631). All contextual characteristics are represented fairly compared to the population of MPs.
Out of a total of 3334 posts, 2032 were individualized, 298 privatized and 1004 posts were not personalized. Appendix C shows the summary of personalization of messages according to contextual differences. Individualized messages are by far the most published in every contextual difference group, whereas privatized messages always make up the minority.
During the first time frame 2507 messages, while during the second time frame only 827 messages were published.
The highest FMS reached was 47.35 by a message posted on Election Day by Rudolf Henke (CDU) and the lowest FMS was 0 by 170 messages posted by diverse MPs.
Consequently a maximum of 47.35% of followers reacted to a Facebook message of a politician in the sample. The mean FMS of all messages is 1.28.
Relationship of Personalization with FMS. In order to answer the first research question, a multiple regression was conducted. Appendix D shows the results of the model. The
regression model with FMS as dependent variable and individualization and privatization as independent variables is significant, F(2, 3331) = 35.79, p < .001. Therefore, the regression model can be used to predict FMS, but the strength of the prediction is weak: 2.1 per cent of the variation in FMS can be predicted on the basis of individualization and privatization (R2 = .021). Individualization, b* = 0.50, t = 6.86, p < .001, 95% CI [0.36, 0.64] and privatization, b* = 0.90, t = 7.24, p < .001, 95% CI [0.66, 1.15], have a significant association with FMS. This means that an individualized message is associated with a FMS of 0.5 points higher than other messages. A privatized message is associated with a score of 0.9 FMS points higher than other messages. For these effects other independent variables are assumed to be held constant.
Relationship of contextual differences with FMS. A similar calculation was conducted in order to answer the second research question. Appendix E shows the results of the model. The regression model with FMS as dependent variable and the contextual characteristics age, party affiliation, mandate status, candidacy status, internal party status, digital natives,
education and gender as independent variables proofed to be significant, F(14, 3319) = 24.84, p < .001. This means that the regression model can be used to predict FMS. It is important to note that messages of members of the CDU and of politicians in the age cohort 45 – 54 did not correlate with FMS and were therefore eliminated from the model. The strength of the prediction of the remaining variables is moderate: 10.1 per cent of the variation in FMS can be predicted on the basis of the variables of contextual differences (R2 = .101). The strongest positive association with FMS was found for messages of MPs younger than 35, b* = 0.91, t = 3.19, p < .01, 95% CI [0.35, 1.46], followed by internal party status, b* = 0.79, t = 7.20, p < .001, 95% CI [0.57, 1.00], candidacy status, b* = 0.69, t = 6.71, p < .001, 95% CI [0.49, 0.89], education, b* = 0.65, t = 6.29, p < .001, 95% CI [0.49, 0.85], MPs in the age cohort 55 - 64, b* = 0.41, t = 4.60, p < .001, 95% CI [0.23, 0.58] and time frame, b* = 0.19, t = 2.57, p < .05, 95% CI [0.05, 0.33]. The strongest negative association was found for message of members of Bündnis 90/Die Grüne, b* = -1.21, t = -5.31, p < .001, 95% CI [-1.66, -0.76], followed by MPs older than 65, b* = -1.06, t = -7.34, p < .001, 95% CI [-1.35, -0.78] and gender, b* = -0.46, t = -4.14, p < .001, 95% CI [-0.67, -0.24]. Consequently, messages of MPs younger than 35 are associated with 0.91 more FMS points than messages from MPs of other age groups. A high internal party status means that messages have 0.69 FMS points more than those of MPs with a low internal party status. Challenging candidates publish message that are associated with 0.69 FMS points more than messages of incumbent candidates. Messages from MPs representing electoral districts or counties with many highly educated inhabitants are 0.65 FMS points more successful than those from less educated regions. MPs in the age between 55 and 64 publish messages that are associated with 0.41 higher FMS than other age
groups. Furthermore, after election message are associated with 0.19 higher FMS than
messages before election. MPs of Bündnis 90/Die Grüne are least successful in terms of FMS. Their messages score 1.21 FMS points less than those of MPs of other parties. MPs above 65 years publish messages that are associated with 0.46 less FMS than those of other age groups. Lastly, Facebook messages of female MPs score 0.46 less FMS points than message of male MPs.
Relationship of interaction of personalization and contextual differences with FMS. Before analysing the data for interactions, the variables were centred in order to avoid multicollinearity. Each value of a variable was subtracted by the mean of the variable. This way, a centred variable was created. In the following, each contextual difference variable was multiplied with each personalization variable in order to create potential two-way interactions. Lastly, two multiple regressions, one with contextual differences interacting with privatization and one interacting with individualization as independent variables and FMS as dependent variables, were conducted in order to answer the third research question.
Appendix F shows the results of the first interaction model. The regression model proofed to be significant, F(15, 3318) = 3.43, p < .001, and found eight significant interactions of contextual differences and privatization predicting FMS. Nevertheless, the strength of the prediction of the variables is weak: 1.8 per cent of the variation in FMS can be predicted on the basis of the interaction variables of contextual differences and privatization (R2 = .018). The strongest positive association with FMS was found for privatized messages published by MPs in the age cohort 55 - 64, b* = 2.76, t = 5.18, p < .001, 95% CI [1.72, 3.78], followed by the age cohort 35 - 44, b* = 2.21, t = 3.38, p < .01, 95% CI [0.93, 3.50], MPs older than 65, b* = 2.18, t = 2.67, p < .01, 95% CI [0.58, 3.78], MPs of the CSU, b* = 1.98, t = 2.67, p < .01, 95% CI [0.52, 3.43] and the interaction of privatized messages and MPs within the age from 45 to 54, b* = 1.70, t = 2.81, p < .01, 95% CI [0.51, 2.88]. The strongest
negative association with FMS was found in the interaction of privatized messages and MPs younger than 35, b* = -9.05, t = -3.34, p < .01, 95% CI [-14.36, -3.73], followed by internal party status, b* = 1.20, t = 2.74, p < .01, 95% CI [2.05, 0.34] and gender, b* = 0.24, t = -3.49, p < .001, 95% CI [-0.37, -0.10]. This means that privatized messages of MPs between 55 and 64 years are associated with 2.76 more FMS points than those of other age groups. Privatized messages of MPs in the age cohort 35 – 44 have 2.21 and those of MPs older than 65 have 2.18 more FMS points than other age cohorts. MPs of the CSU publish privatized messages that are associated with 1.98 more FMS points than MPs of other parties. 1.70 FMS points higher than other age groups are scored by privatized messages of MPs in the age from 45 to 54. Privatized messages of the youngest age group among German MPs, those younger than 35, are associated with 9.05 less FMS than privatized messages of other age groups. Lower internal level MPs publish privatized messages that are 1.20 FMS points less
successful than those of high internal level MPs. Lastly, privatized messages of female MPs are associated with 0.24 less FMS points than those of male MPs.
Also the second regression model is significant, F(15, 3318) = 2.70, p < .001. 4 significant interactions of contextual differences and individualization predicting FMS were found. Appendix G shows the results of the second interaction model. The strength of the prediction of the variables is weak: 1.2 per cent of the variation in FMS can be predicted on the basis of the interaction variables of contextual differences and individualization (R2 = .012). The strongest positive association with FMS was found for individualized messages published by MPs in the age cohort 55 - 64, b* = 0.86, t = 4.12, p < .001, 95% CI [0.45, 1.27], followed by candidacy status, b* = 0.52, t = 2.32, p < .05, 95% CI [0.80, 0.96], MPs between 35 and 44 years old, b* = 0.49, t = 3.05, p < .01, 95% CI [-1.14, 1.50] and MPs in the age cohort 45 – 54, b* = 0.44, t = 3.35, p < .01, 95% CI [0.18, 0.70]. Hence, individualized messages of MPs that are between 55 and 64 years old are associated with 0.86 FMS points more than other age groups. Challenging MPs publish individualized messages that are 0.52
FMS points more successful than those of incumbent MPs. Furthermore, individualized messages of MPs from 35 to 44 years of age are associated with 0.52 and MPs in the age group 45 – 54 are associated with 0.49 FMS points higher than those of other age groups.
Discussion
The silent majority. The aim of this thesis is to shed light on personalization effects by analysing the content of Facebook messages in the context of social media. Generally, Facebook messages of members of the German parliament that are personalized generate more “likes” than message that are not personalized. Thus, the assumption of Jebril et al. (2013), that personalization of political coverage increases cynicism, cannot be supported in the context of social media. “Likes” do not express cynicism, yet it seems unlikely that exposure to a personalized Facebook message encourages a user to express a positive depiction of the same, while actually disliking it. From the perspective of Siri et al. (2012), this result was predictable, since social media audiences tend to be homogenous in terms of party affiliation. However, several comments included opinions that criticised the respective MP and his or her party, some comments openly promoted competing parties. Furthermore, only a fraction of all followers expressed their feeling and opinions to a message. An example for this participation gap is in terms of FMS by far the most successful message that was published by Rudolf Henke (CDU). The individualized message stimulated a reaction of 47% of his 340 followers and was posted on Election Day, containing a personal acknowledgement to voters and followers. Subsequently, a silent majority among Facebook followers exist. Since Facebook limits immediate expression means to a positive reaction, the positive association of personalization with FMS might only apply to the homogenous group among the followers. Consequently, it is necessary to expand this field of research towards the social media audience of political communication.
Privatized messages generated the most FMS among the personalization variables. 18
Therefore H1 is supported. The hypothesis that politicians publishing privatized messages are most successful, in terms of FMS, because they are able to appeal to perceived similarities with the voters is confirmed by the analysis.
FMS and contextual differences. According to the analysis, contextual differences generally explain more variations in FMS than personalization of messages. Most interestingly, MPs younger than 35, or those that are part of the generation of digital natives, have the most success in terms of FMS. It seems that they are most capable to communicate via Facebook, because they are inherently professionalized with regard to social media and follow a “like”-oriented approach. A diametrical opposed finding is that MPs older than 65 generate the least FMS. It seems logical that it is more difficult for the older generation to communicate via social media, because they are confronted with new technologies and fast communication developments in a later period of life, rather than growing up with it. Another reason for a smaller FMS of older MPs might be that, even though messages were individualized, many messages were explicitly written by a third person. Consequently, a “face-to-face” approach was not always visible, which might have reduced FMS. Since differences in content and style, other than personalization, are not analysed, further research could focus on the age gap with regard to the substance of the messages.
Furthermore, the advantage of a high internal party status can be observed. The assumption that top level politicians benefit from a professional social media consultant or employee is supported by the data. Another reason for this FMS difference might lie in the fact that top level politicians are generally more known to the public, since they are depicted by traditional media more often. This thesis tried to rule out a popularity bias, yet due to the phenomenon of political fandom, followers of popular politicians might be more inclined to give a “like”.
A professionalized, “like”-oriented approach might also be the reason for a higher 19
FMS of direct candidates compared to listed candidates as well as challenging candidates compared to incumbent candidates. In both cases, direct and challenging status, MPs might take more effort in their Facebook presentation than their counterparts. It is more important for direct candidates to appeal to citizens of their respective electoral district than for listed candidates, since they are direct representatives of the same and highly depend on the votes of a specific group of citizens. It might be interesting to investigate by future studies whether the followers of a direct candidate are mainly inhabitants of the respective electoral district. The Facebook audience of challenging candidates is also of interest in this regard. For challenging candidates it is generally more important to generate attention vis-à-vis an incumbent
candidate, since the inherent underdog position often include a disadvantage with regard to communication channels, e.g. official parliament channels. Consequently, a professionalized, “like”-oriented Facebook usage might counterbalance this disadvantage.
A surprising result was observed with regard to Facebook message of MPs of Bündnis 90/Die Grüne. Against the assumption that they are most successful they scored the least FMS. A reason for this outcome might be the small amount of respective MPs analysed (n=6). Another reason might lie in a political shift. Their traditional grass-root organizations tend to lose impact since the political orientation of the party shifted slowly since 1998 and
considerably since the 2013 election from the progressive left to the conservative centre (Mohr, 2012; Niedermayer, 2013). Yet the clear gap with regard to FMS compared to other parties cannot be explained by this. Future research could base their research on this unexpected finding.
The outcome showed campaign sensitivity, however not as expected due to the higher frequency of social media messages posted during campaign times. A higher FMS was observed for messages after election, even though fewer messages were published. In this regard it is noteworthy that some politicians posted an immense amount of messages without stimulating any or any considerable reaction from the users in the pre-election time frame.
This can be seen as an indicator of a non-professional usage of Facebook that principally aims at being active on Facebook, rather than pursuing a “like”-oriented political communication on social media. MPs applying a “like”-oriented strategy might be those who continued their activity on Facebook, resulting in a higher FMS in the second time frame. Furthermore, before election, many MPs published messages that portrayed the political opponent in a negative way. These posts were highly debated among users and barely resulted in a mentionable amount of “likes”.
As suggested by theory, messages of MPs from regions with many educated people have higher FMS than those from less educated electoral districts or counties. Consequently, MPs seem to react to an information demand of highly educated citizens and might therefore be more professionalized, pursuing a “like”-oriented approach. The effect could also root in the fact that respective MPs have an inherent advantage with respect to FMS, since highly
educated people are more likely to gather political information on social media. The gender of a MP proofed to be a significant factor, yet the reasons behind a higher
FMS of messages of male MPs compared to female MPs remains unclear.
The Bavarian exception and other anomalies. Privatized messages of all age groups except messages of MPs younger than 35, are more successful in terms of FMS than individualized or non-personalized messages. One reason for this outcome might be that younger MPs are not successful in appealing to perceived similarities with citizens. An age gap might exist between the MPs and his or her Facebook audience, resulting in shifted similarities.
Unfortunately, a test for an interaction effect between privatized messages, MPs younger than 35 and regions with many digital natives did not produce a significant outcome. Consequently future research could focus on the audience of MPs’ Facebook profiles in general to shed light on this relationship.
Interestingly, the exploitation of traditional, cultural habits might be the reason for the 21
success of privatized messages among MPs of the CSU. The party, solely based in Bavaria, puts much emphasis on the Bavarian culture in their election campaigns and general
communication. The majority of privatized posts of CSU members highlight Bavarian traditions, like traditional clothing, Bavarian folk festivals like the Oktoberfest and the
consumption of Bavarian beer. Culturally and politically, Bavaria is a specific German county, because regional traditions are given an important value (Berg-Schlosser & Schissler, 1987; Wehling; 2006). It seems that CSU members are successful in appealing to perceived similarities with their Facebook followers.
Furthermore, high level politicians publish more successful privatized messages than lower level MPs. Once again a reason might be the phenomenon of political fandom. Insights into the private lives of popular politicians might be more interesting to the followers, since the perceived societal status gap with the citizens is higher.
With regard to individualized messages MPs of all age groups between 35 and 65 are more successful than the youngest and the oldest MPs. Since 3-way interactions with demand factors do not provide any significant results, this finding is not allegeable in the light of the theory and seems to be an anomaly.
On the other hand, that challenging MPs publish individualized messages that are more successful can be explained due to their underdog status, making it necessary to depict the respective politician and his or her personal achievements in a positive light.
Individualization of Facebook messages are a proper mean in this regard which seems to be appreciated by the audience.
Unfortunately, interactions of personalized messages in general and demand factors did not prove to be significant. This seemed especially interesting due to possible differences of perception of personalized messages among different groups of society. Since the data did not prove to be significant in this regard, future research could put special emphasise on the audience of Facebook profiles of MPs, as mentioned above.
Limitations. This thesis has several limitations. The most striking is certainly the limited amount of demand-driven factors. In order to explain interactions between the variables properly, it is necessary to include more variables that explain a political information demand. Furthermore, the thesis did not analyse the actual audience of the MPs Facebook profiles. This would have been benefited to the explanatory power of the study, especially in the light of a silent majority among followers.
Considering the methodological part, some measurements are limited. The demand driven factors education and digital natives are oriented on the average of the German population. Both variables could be based on scale measurements in order to depict a more accurate picture.
Furthermore, measurements of personalization could be focused more on the content of the messages, e.g. considering the explicit author of the messages, the MP or a third person.
Conclusion
This thesis aimed to bring new insights to research on personalization of political coverage in the context of social media. By conducting a content analysis among Facebook profiles among members of the 18th German federal parliament, findings showed that personalization mattered with respect to success of political communication on social media. Furthermore the study showed that contextual differences of MPs have influence, not only on message success in general, but also on the depiction of personalized messages in particular. Consequently, as suggested by theory, the backgrounds of the audience and the politicians themselves have not only an effect on the perception of personalized political coverage in traditional media, but in social media, too. Politicians that are aware of their social media audience are therefore able to react to the respective information demand and could optimise their social media
presentation. Especially promising in this regard is the privatized approach. A homogenous 23
audience, appealing to perceived similarities with the same could lead to a higher success of political communication, resulting in higher attention for the respective politician and
eventually, to an increased mobilization effect. However, a silent majority among followers of MP’s Facebook profiles exist that could not be analysed due to the lack of data. All in all, not only traditional media shapes our perception of politics, culture and reality in general, but social media does as well. The difference is that social media equips politicians with the possibility to add a new shape themselves.
References
Baron, R. M. and Kenny, D. A. (1986). The Moderator-Mediator Variable Distinction in Social Psychological Research – Conceptual, Strategic, and Statistical Considerations. Journal of Personality and Social Psychology, 51 (6), 1173–1182.
Berg-Schlosser, D., & Schissler, J. (1987). Politische Kultur in Deutschland. VS Verlag für
Sozialwissenschaften, 11-26.
Bronstein, J. (2013). Like me! Analyzing the 2012 presidential candidates’ Facebook pages. Online Information Review, 37(2), 173-192.
Bundesinstitut für Bevölkerungsforschung (2014) Bevölkerungsbilanz und Altersstruktur, Retrieved from
http://www.bib-demografie.de/DE/ZahlenundFakten/02/bevoelkerungsbilanz_und_altersstruktur_node .html;jsessionid=B7D49FDAE7FE63EC04A45B2CD52835C9.2_cid331
Cameron, M. P., Barrett, P., & Stewardson, B. (2013). Can Social Media Predict Election Results? Evidence from New Zealand, 13(8).
Chen, P. J., & Walsh, L. (2010). E-election 2007? Political competition online. Australian
Cultural History, 28(1), 47-54.
Effing, R., van Hillegersberg, J., & Huibers, T. (2011). Social media and political participation: are Facebook, Twitter and YouTube democratizing our political systems?. Electronic participation, 25-35.
Elter, A. (2013). Interaktion und Dialog? Eine quantitative Inhaltsanalyse der Aktivitäten deutscher Parteien bei Twitter und Facebook während der Landtagswahlkämpfe 2011. Publizistik, 58(2), 201-220.
Enli, G. S., & Skogerbø, E. (2013). Personalized campaigns in party-centred politics: Twitter and Facebook as arenas for political communication. Information, Communication &
Society, 16(5), 757-774.
Franch, F. (2013). Wisdom of the crowds 2: 2010 UK election prediction with social media. Journal of Information Technology & Politics, 10(1), 57-71.
Garzia, D. (2011). The personalization of politics in Western democracies: Causes and consequences on leader–follower relationships. The Leadership Quarterly, 22(4), 697-709.
Gibson, R. K., & McAllister, I. (2011). Do online election campaigns win votes? The 2007 Australian “YouTube” election. Political Communication, 28(2), 227-244.
Gibson, R. (2012). From brochureware to ‘MyBo’: An overview of online elections and campaigning. Politics, 32(2), 77-84.
Giglietto, F. (2012). If Likes Were Votes: An Empirical Study on the 2011 Italian Administrative Elections. ICWSM.
Gil de Zúñiga, H., Jung, N., & Valenzuela, S. (2012). Social media use for news and individuals' social capital, civic engagement and political participation. Journal of Computer‐Mediated Communication, 17(3), 319-336.
Gschwend, T., & Zittel, T. (2014). Do constituency candidates matter in German Federal Elections? The personal vote as an interactive process. Electoral Studies.
Gulati, G. J., & Williams, C. B. (2013). Social Media and Campaign 2012 Developments and Trends for Facebook Adoption. Social Science Computer Review, 31(5), 577-588.
Halpern, D., & Gibbs, J. (2013). Social media as a catalyst for online deliberation? Exploring the affordances of Facebook and YouTube for political expression. Computers in Human Behavior, 29(3), 1159-1168.
Jamieson, K. H., & Campbell, K. K. (2000). The interplay of influence: News, advertising, politics, and the mass media.
Jebril, N., Albæk, E., & de Vreese, C. H. (2013). Infotainment, cynicism and democracy: The effects of privatization vs personalization in the news. European Journal of
Communication, 28(2), 105-121.
Korte, K. (2009). Die Amerikanisierung der Wahlkämpfe, Retrieved from
http://www.bpb.de/politik/wahlen/bundestagswahlen/62584/amerikanisierung Kriesi, H. (2012). Personalization of national election campaigns. Party Politics,18(6),
825-844.
Larsson, A. O., & Moe, H. (2012). Studying political microblogging: Twitter users in the 2010 Swedish election campaign. New Media & Society, 14(5), 729-747.
Larsson, A. O., & Kalsnes, B. (2014). ‘Of course we are on Facebook’: Use and non-use of social media among Swedish and Norwegian politicians. European Journal of Communication.
Macnamara, J., & Kenning, G. (2011). E-electioneering 2010: Trends in social media use in Australian political communication. Media International Australia, Incorporating Culture & Policy, (139), 7.
McAllister, I. (2007) The personalization of politics. The Oxford handbook of political
behaviour, 571-588.
Metaxas, P. T., Mustafaraj, E., & Gayo-Avello, D. (2011). How (not) to predict elections. Privacy, security, risk and trust (PASSAT), 2011 IEEE third international conference on and 2011 IEEE third international conference on social computing (SocialCom), 165-171.
Mohr, K. (2012). Von „Welfare to Workfare “? Der radikale Wandel der deutschen Arbeitsmarktpolitik. Arbeitsmarktpolitik in der sozialen Marktwirtschaft, 57-69. Neveu, E. (2002). Four generations of political journalism. Political journalism: New
challenges, new practices, 22-44.
Norris, P. & Curtice, J. (2008), ‘Getting the Message Out: A Two-Step Model of the Role of the Internet in Campaign Communication Flows during the 2005 British General Election’, The Journal of Information Technology and Politics 4(4), pp. 3–13.
Niedermayer, O. (2013). Die Entwicklung des bundesdeutschen Parteiensystems. In Handbuch der deutschen Parteien, 111-132.
Reinemann, C., & Wilke, J. (2007). It's the Debates, Stupid! How the Introduction of Televised Debates Changed the Portrayal of Chancellor Candidates in the German Press, 1949—2005. The Harvard International Journal of Press/Politics, 12(4), 92-111.
Schulz, W., Zeh, R., & Quiring, O. (2005). Voters in a Changing Media Environment A Data-Based Retrospective on Consequences of Media Change in Germany. European
Journal of Communication, 20(1), 55-88.
Siever, T. (2012). Zugänglichkeitsaspekte zur Kommunikation. Sprache und Kommunikation im technischen Zeitalter: Wieviel Internet (v)erträgt unsere Gesellschaft?, 2, 7. Siri, J., Melchner, M., & Wolff, A. (2012). The Political Network. Parteien und politische
Kommunikation auf Facebook. Phänomen „Facebook “. Sonderausgabe von kommunikation@ gesellschaft, 13(6).
Statista.com (2014). Nutzer von Facebook in Deutschland bis 2014. Retrieved from http://de.statista.com/statistik/daten/studie/70189/umfrage/nutzer-von-facebook-in-deutschland-seit-2009/
Stieglitz, S., & Dang-Xuan, L. (2012). Political communication and influence through microblogging--an empirical analysis of sentiment in twitter messages and retweet
behavior. System Science (HICSS), 2012 45th Hawaii International Conference, 3500-3509.
Strömbäck, J., & Kiousis, S. (2014). Strategic Political Communication in Election Campaigns.
Tagesschau.de (2014) Mobiler Markt macht Facebook reich, Retrieved from http://www.tagesschau.de/wirtschaft/facebook-108.html
Van Aelst, P., Sheafer, T., & Stanyer, J. (2012). The personalization of mediated political communication: A review of concepts, operationalizations and key findings. Journalism, 13(2), 203-220.
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.
Vitak, J., Zube, P., Smock, A., Carr, C. T., Ellison, N., & Lampe, C. (2011). It's complicated: Facebook users' political participation in the 2008 election. CyberPsychology,
behavior, and social networking, 14(3), 107-114.
Wattal, S., Schuff, D., Mandviwalla, M., & Williams, C. B. (2010). Web 2.0 and Politics: The 2008 US Presidential Election and an E-Politics Research Agenda. MIS Quarterly, 34(4).
Wehling, H. G. (2006). Föderalismus und politische Kultur in der Bundesrepublik Deutschland. Landespolitik in Deutschland, 87-107.
Williams, C. B., & Girish, J. (2012). Social networks in political campaigns: Facebook and the congressional elections of 2006 and 2008. new media & society
Zittel, T. (2009). Lost in technology? Political parties and the online campaigns of
constituency candidates in Germany's mixed member electoral system. Journal of Information Technology & Politics, 6(3-4), 298-311.
Appendix A
Figure 1. Model of the thesis. Contextual characteristics
- age
- party affiliation - mandate status - candidacy status - internal party status - digital natives - education - gender Personalization - Individualization - Privatization - No personalization
Facebook message success (FMS) INDE P E NDE N T V AR IA B L E S DE P E NDE NT V AR IAB L E Interaction 29
Appendix B Table 1
Sample and population in absolute numbers (%) according to contextual characteristics Sample (n=61) Population (N=631) Gender Male 40 (65.6) 403 (63.9) Female 21 (34.4) 228 (36.1) Age <35 4 (6.5) 35 (5.5) 35 – 44 15 (24.6) 154 (24.4) 45 – 54 20 (32.8) 212 (33.6) 55 – 64 17 (27.9) 179 (28.4) >65 5 (8.2) 51 (8.1) Party CDU 25 (41.0) 255 (40.4) CSU 5 (8.2) 56 (8.9) SPD 19 (31.2) 193 (30.6) Die Linke 6 (9.8) 64 (10.1)
Bündnis 90/Die Grüne 6 (9.8) 63 (10.0) Internal party status Top level 17 (27.9) 185 (29.3)
Low level 44 (72.1) 446 (70.7)
Candidacy status Incumbent 43 (70.5) 451 (71.5)
Challenger 18 (29.5) 180 (28.5)
Mandate status Direct mandate 29 (47.5) 294 (46.6) Listed mandate 32 (52.5) 337 (53.4) Digital Natives Many Digital Natives 28 (45.9) 282 (44.7) Few Digital Natives 33 (54.1) 349 (55.3)
Education High education 34 (55.7) 353 (55.9)
Low education 27 (44.3) 278 (44.1)
Appendix C Table 2
Personalization of messages in absolute numbers according to contextual characteristics (N=3334) Personalization Individualization Privatization No personalization Gender Male 1426 166 582 Female 606 132 422 Age <35 76 17 12 35 – 44 440 110 283 45 – 54 848 116 347 55 – 64 486 54 266 >65 162 16 101 Party CDU 991 119 410 CSU 104 13 61 SPD 651 134 343 Linke 160 29 156 Grüne 106 18 39 Internal Party status Top level 371 47 167 Low level 1661 251 837
Candidacy status Incumbent 1530 254 834
Challenger 502 44 170
Mandate status Direct mandate 1374 166 622
Listed mandate 658 132 382
Digital Natives Many digital natives
1088 175 564
Few digital natives
944 123 440
Education High education 1346 210 661
Low education 686 88 343
Time Frame Before election 1569 197 741
After election 463 101 263
Appendix D Table 3
Regression model to predict FMS with personalization (N=3334) FMS b* Personalization Constant 0.90*** Individualization 0.50*** Privatization 0.90*** Note. * p <.05. ** p <.01. *** p <.001. 32
Appendix E Table 4
Regression model to predict FMS with contextual differences (N=3334) FMS b* Constant <35 35 – 44 55- 64 >65 CSU SPD Die Linke
Bündnis 90/ Die Grüne mandate status
candidacy status internal party status digital natives Education Gender Time Frame 0.162 0.91** -0.05 0.41*** -1.06*** -0.16 0.17 0.00 -1.21*** 0.05 0.69*** 0.79*** 0.02 0.65*** -0.46*** 0.19* Note. * p <.05. ** p <.01. *** p <.001. 33
Appendix F Table 5
Regression model to predict FMS with contextual differences interacting with privatization (N=3334) FMS b* Constant <35 * Privatization 35 – 44 * Privatization 45 – 54 * Privatization 55 - 64 * Privatization >65 * Privatization CDU * Privatization CSU * Privatization SPD * Privatization Die Linke * Privatization
Bündnis 90/ Die Grüne * Privatization mandate status * Privatization
candidacy status * Privatization internal party status * Privatization digital natives * Privatization Education * Privatization Gender * Privatization Time Frame * Privatization
1.29*** -9.05** 2.21** 1.70** 2.76*** 2.18** 0.43 1.98** 0.00 -0.76 0.41 0.63 -0.14 -1.20** 0.49 -0.64 -0.24*** 0.64 Note. * p <.05. ** p <.01. *** p <.001. 34
Appendix G Table 6
Regression model to predict FMS with contextual differences interacting with individualization (N=3334) FMS b* Constant <35 * Individualization 35 – 44 * Individualization 45 – 54 * Individualization 55 - 64 * Individualization >65 * Individualization CDU * Individualization CSU * Individualization SPD * Individualization Die Linke * Individualization
Bündnis 90/ Die Grüne * Individualization mandate status * Individualization
candidacy status * Individualization internal party status * Individualization digital natives * Individualization Education * Individualization Gender * Individualization Time Frame * Individualization
1.28*** 0.18 0.49** 0.44** 0.86*** 0.57 0.00 0.23 0.05 0.20 -0.32 0.04 0.52* 0.33 -0.16 0.24 -0.21 0.09 Note. * p <.05. ** p <.01. *** p <.001. 35
Appendix H Code Book
Wave 1 – Contextual differences 1) Age 1: <35 2: 36 – 45 3: 46 – 55ix 4: 56 – 65 5: >65 2) Education 1: High education
o More than 33.9% (average Germany) of inhabitants of electoral district/county with owning a diploma from German secondary school qualifying for
university admission or matriculation (“Abitur” or “allgemeiner Hochschulabschluß)
2: Low education
o Less than 33.9% (average Germany) of inhabitants of electoral district/county with owning a diploma from German secondary school qualifying for
university admission or matriculation (“Abitur” or “allgemeiner Hochschulabschluß)
3) Digital natives
1: Many digital natives
o more than 36.6% (average Germany) in electoral district/county 2: few digital natives
o less than 36.6% (average Germany) in electoral district/county 4) Party affiliation
1: CDU 2: CSU 3: SPD 4: Die Linke
5: Bündnis 90/ Die Grüne 5) Internal party status
1: High internal party status
o holds an elected office in party, government or parliament 2: Low internal party status
o does not hold an elected office in party, government or parliament 6) Mandate status 1: direct mandate 2: listed mandate 7) Candidacy status 1: incumbent 2: challenger 36
8) Gender 1: male 2: female
Wave 2 – Personalization of messages 1) Time frame 1: Before election 2: After election 2) Frequency – date of message – # of candidate message 3) Success of message – # of likes 4) Success of profile – # of followers 5) Personalization 1: Individualized post
o The post is explicitly written from the perspective of the politician o Usage of first-person pronouns
o The name of the politician is included in the text or audio/visual elements o The politician is subject of audio/visual elements
2: Privatized post
o The main message of a text or audio/visual element concerns the private life of the politician
o Private content that is not explicitly related to the work of politicians o Examples:
If the main message of a post is drinking a morning coffee in the office, it is a privatized post
If the main message of a post is drinking a coffee with citizens at a campaign event, it is not a privatized post
3: No Personalization
o None of the conditions above apply
