The Rise of Germany's AfD: A Social Media Analysis

Abstract

In 2017, a far-right party entered the German parliament for the first time in over half a century. The Alternative für Deutschland (AfD) became the third largest party in the government. Its campaign focused on Euroscepticism and a nativist stance against immigration. The AfD used all available social media channels to spread this message. This paper seeks to understand the AfD's social media strategy over the last years on the full gamut of social media platforms and to verify the effectiveness of the party's online messaging strategy. For this purpose, we collected data related to Germany's main political parties from Facebook, Twitter, YouTube, and Instagram. This data was subjected to a unified multi-platform analysis, which relies on four measures: party engagement, user engagement, message spread, and acceptance. This analysis proves the AfD's superior online popularity relative to the rest of Germany's political parties. The evidence also indicates that automated accounts contributed to this online superiority. Finally, we demonstrate that as part of its social media strategy, the AfD avoided discussion of its economic proposals and instead focused on pushing its anti-immigration agenda to gain popularity.

Full text

The Rise of Germany’s AfD: A Social Media Analysis
Juan Carlos Medina Serrano
Bavarian School of Public Policy
Munich, Bavaria
juan.medina@tum.de
Morteza Shahrezaye
Bavarian School of Public Policy
Munich, Bavaria
morteza.shahrezaye@tum.de
Orestis Papakyriakopoulos
Bavarian School of Public Policy
Munich, Bavaria
orestis.papakyriakopoulos@tum.de
Simon Hegelich
Bavarian School of Public Policy
Munich, Bavaria
simon.hegelich@hfp.tum.de
ABSTRACT
In 2017, a far-right party entered the German parliament for the rst
time in over half a century. The Alternative für Deutschland (AfD)
became the third largest party in the government. Its campaign fo-
cused on Euroscepticism and a nativist stance against immigration.
The AfD used all available social media channels to spread this mes-
sage. This paper seeks to understand the AfD’s social media strategy
over the last years on the full gamut of social media platforms and
to verify the eectiveness of the party’s online messaging strategy.
For this purpose, we collected data related to Germany’s main polit-
ical parties from Facebook, Twitter, YouTube, and Instagram. This
data was subjected to a unied multi-platform analysis, which relies
on four measures: party engagement, user engagement, message
spread, and acceptance. This analysis proves the AfD’s superior
online popularity relative to the rest of Germany’s political parties.
The evidence also indicates that automated accounts contributed
to this online superiority. Finally, we demonstrate that as part of its
social media strategy, the AfD avoided discussion of its economic
proposals and instead focused on pushing its anti-immigration
agenda to gain popularity.
CCS CONCEPTS
•Networks →Social media networks
;
•Human-centered com-
puting →Social network analysis;
KEYWORDS
political campaigns, social media, AfD, multi-platform, Twitter,
Facebook, Instagram, YouTube
ACM Reference format:
Juan Carlos Medina Serrano, Morteza Shahrezaye, Orestis Papakyriakopou-
los, and Simon Hegelich. 2019. The Rise of Germany’s AfD: A Social Media
Analysis. In Proceedings of International Conference on Social Media and
Society, Toronto, ON, Canada, July 19–21, 2019 (SMSociety ’19), 10 pages.
https://doi.org/10.1145/3328529.3328562
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than ACM
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specic permission and/or a
fee. Request permissions from permissions@acm.org.
SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada
©2019 Association for Computing Machinery.
ACM ISBN 978-1-4503-6651-9/19/07. . . $15.00
https://doi.org/10.1145/3328529.3328562
1 INTRODUCTION
The rise of the Alternative für Deutschland (AfD) represented
a schism in German politics. The AfD originally emerged as an
anti-Euro party and then gradually adopted the language of right-
wing populism [
54
]. The AfD has advocated for anti-Euro, anti-
immigration and anti-refugee policies, and has been outspoken
about other previously taboo topics in German politics. Its anti-
establishment rhetoric parallels that of other EU right-wing populist
parties, such as the National Front in France, the Party for Freedom
in Netherlands and the Lega Nord in Italy. The recent surge in
far-right voting in Europe calls for careful study of the roots of this
new political trend.
The AfD was founded as a eurosceptic party by a group of uni-
versity professors and former politicians in February 2013. Their
proposals were centered on economic liberalism, ordoliberalism
and free market ideas. Though the AfD was originally a single-issue
party, it soon found support from right-wing groups and started
shifting toward an anti-immigration ideology. Before the AfD, right-
wing populist parties had achieved only limited electoral success
in Germany. The AfD overcame this burden by distancing itself
from previous right-wing ideologies and presenting itself as a party
with economic expertise and scientic authority [
22
]. Moreover, it
formed a stable nation-wide organization [
4
]. In the 2013 federal
elections, the AfD missed the 5% threshold for entering parliament
by only 0.3%. Nevertheless, by the next year, the party won seven
seats in the European Parliament and later entered three state par-
liaments.
This rapid increase in electoral success would not have been pos-
sible without the AfD’s wide base of supporters. At its beginnings,
the AfD’s constituents were mostly well educated, high-income citi-
zens [
5
]. Following 2014, the party enjoyed a great surge of support
from low-income citizens [
42
]. According to the latest study of the
issue, the 2017 election report from the Infratest dimap Institute [
10
],
the largest demographic group that voted for the AfD was East Ger-
man men. Kim
[30]
investigated why lower socio-economic groups
like blue-collar workers and the unemployed would support a party
that advocates radical market-oriented policies, which would not
benet them. Kim argues that the AfD strategically avoided discus-
sion of the party’s economic proposals to prevent divisions among
its supporters.
The AfD’s base of support tripled in recent years, from 5% in 2015
to 15% in 2018. This inection in the opinion polls started in Septem-
ber 2015, at the beginning of the refugee crisis. The AfD’s popularity
214

SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada J.C. Medina Serrano et al.
in polling grew over one year, and then the polls remained stable
until several months before the 2017 federal elections. The party’s
support increased further after the elections, indicating approval
of the party’s work in parliament.
The rise of the AfD cannot be understood fully without taking its
online presence into consideration. This study aims to describe the
AfD’s social media strategy and compare its eectiveness with that
of other German parties by examining various platforms over mul-
tiple years. First, we give an overview of political parties’ activity
on social media and then we focus on the AfD’s user engagement
strategy. We propose a unied multi-platform analysis for the evalu-
ation of all the dierent social media channels. This analysis covers
the four social media platforms that are used regularly by German
political parties to examine how political messages spread across
them. To further understand the AfD’s strategy, we compare its
social media discourse with their campaign proposals. Moreover,
we show that the spread of the AfD’s messages was boosted by
automated accounts on at least one of the platforms.
2 THEORETICAL BACKGROUND
2.1 Political Parties’ Activity on Social Media
The use of social media by political parties has transformed political
communication. Since the World Wide Web provides a plurality of
possibilities for communicating with the electorate [
43
], political
parties have developed new methods and tools for externalizing
party attitudes and evaluating the reactions of potential voters [
59
].
In social media, candidates and parties have ocial pages and ac-
counts, through which they make political statements and declare
their positions on salient issues. These messages are diused fur-
ther on the platforms by journalists and other users [
64
]. Voters
can then respond to political actors, providing them with rapid and
granular feedback; the directness of this political dialog was not
possible when only traditional media was available [
44
]. This form
of political interactivity has been proven to benet political actors,
making them more favorable to the electorate [
63
]. Additionally, in
accord with existing trends in data collection [
34
], political parties
are using social media to collect, monitor, and analyze voter reac-
tions to political messaging [
56
]. These analyses help the parties to
design, correct, and strategically adapt campaign activities. Finally,
political parties use social media as spaces for political microtar-
geting [
45
], sending personalized messages to users to encourage
support. Social media has become so critical to political campaign-
ing that social media has become ’environmental’[
58
]: parties and
candidates cannot avoid or neglect its use, as social media platforms
are now a cornerstone of political communication.
Given the constant ow of information on social media, politi-
cal parties and candidates no longer present a static or complete
overview of their views and positions to the users. They tend to
comment and respond on topics that were made prominent by ex-
ogenous events, like economic crises or natural catastrophes, or
they adapt rapidly to the topics in the agenda set by mass-mediated
public debates and news-media platforms [
25
]. In addition, they
choose to address topics about which they hold strong and inuen-
tial positions and that are of concern to voters [
68
], while avoiding
other topics that might decrease their popularity. Finally, the parties
concentrate on issues and strategies that are tailored to the audi-
ence on each social media platform [
62
]. This behavior means that
the image of the political parties that is presented to the electorate
often deviates from the ocial party positions that are expressed
in political manifestos [25].
Although the above behaviors and strategies are characteristic of
all parties, the opportunities available on social media are of greater
importance for outsider parties. As they have limited access to tradi-
tional mass media, they use these new communication channels to
overcome disadvantages in communication and to contact potential
voters [
31
]. Therefore, outsider parties emphasize their online pres-
ence and intensify their interactions with users to achieve eective
communications [
27
]. Due to their limited resources, they are even
more selective about the topics that they express opinions about and
they deploy strategies that promote only their strongest arguments
[
68
]. For example, left-wing populist parties tend to concentrate on
economic issues on social media, while right-wing populist parties
tend to focus on issues that resonate with xenophobic voters [14].
2.2 The AfD’s Social Media Strategy
As an outsider party, social media has been an important commu-
nication channel for the AfD since its foundation, because social
media platforms provided a space to inuence public opinion out-
side of the traditional media. In recent years, the AfD has been
eective on social media as reported on media channels and in
previous research: Arzheimer
[1]
analyzed Facebook posts from
2013 and 2014 and found that the AfD used more populist rhetoric
on Facebook than it did on other communication channels; Schelter
et al
. [52]
evaluated the Facebook posts of six political parties in
Germany in 2014 and 2015 and reported that social media was a
major factor in the success of the AfD; and both Hegelich
[23]
and
Medina Serrano et al
. [36]
studied social media campaigns in the
months leading up to the 2017 German federal election.
From the literature, we deduce three main factors that help ex-
plain the AfD’s eectiveness on social media:
•Alternative media:
The AfD relies on social media plat-
forms to spread its message. The party leaders have blamed
traditional media for presenting them in a negative light
and obscuring their intentions. Using social media as an
alternative ecosystem, the AfD reached a part of the Ger-
man population that felt disenchanted with conventional
communication channels. Indeed, a study from the Otto-
Brenner-Stiftung [
9
] conrmed that followers of the AfD
place less trust in the German media. Hence, they prefer
to obtain information from social media platforms. Further-
more, a 2018 Pew Research Center report [
46
] ascertained
that people with populist preferences in Germany tend to
have less trust in the media. The right-wing political party
has taken advantage of this fact by employing a strong social
media campaign.
•High online activity:
The AfD’s strategy is to make use of
social media as much as possible and get its content to go
viral. To achieve this, the AfD regularly asks its supporters to
share content. Furthermore, it uses a provocative tone, which
together with its critical position on political correctness
[
40
], encourages users to engage and reply with positive
215

The Rise of Germany’s AfD SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada
or negative comments. Another factor that stimulates the
high user response is the negative and aggressive tone in
the AfD’s anti-establishment and anti-immigrant stances,
following the work of Fan et al
. [15]
that showed that hate
spreads faster on social media.
•Online Manipulation:
The right-wing party was not alone
in spreading its message, as pro-AfD social bots were active
on Facebook [
3
] and Twitter [
41
]. Social bots [
16
] are au-
tomated fake accounts that are fashioned to look like real
users and whose purpose is to viralize topics and manipu-
late trends. Even though Neudert et al
. [41]
found low levels
of automation in the time leading up to the 2017 German
federal election, they did nd that social bots in their sample
were working in favor of the AfD. Although it is not possible
to track the origin of these bots, two online communities
—Infokrieg and Reconquista Germania— had the explicit goal
of trolling social media in support of the AfD [
12
]. While
the eects of online manipulation attempts on public opin-
ion are dicult to quantify, these automated accounts likely
amplied the online reach of the AfD’s message.
These three points are in line with the social media activity of
other populist parties in Europe [
11
,
35
,
53
]. Social media has given
populist political actors greater freedom to articulate their ideology
and spread their message [
14
]. Social media acts as a sort of people’s
voice in populist movements [
19
] facilitating the reinforcement of
the anti-establishment ideology that is common to populist parties
[
37
]. Furthermore, social media has no gatekeepers to fact-check
the information, which gives populists a fertile space to spread their
rhetoric [29].
2.3 Multi-Platform Schema
In order to understand and evaluate the AfD’s strategies on so-
cial media, a long-term multi-platform analysis is needed. Even
though research on digital campaigns over the last two decades
has been extensive, few studies have focused on how candidates
and parties use multiple social media channels [
50
,
57
]. The so-
cial media environment itself has become more complex, and the
fast pace of technology and digitalization calls for researchers to
adapt their methods to cross-platform research. Indeed, researching
media platforms separately ignores the reality of today’s contempo-
rary media experience [
7
]. Already in 1968, when the term media
ecology was rst coined, researchers realized that media should
not be considered in isolation [
48
]. However, researchers tend to
conveniently study the types of social media that can be accessed,
without taking into consideration the political relevance of specic
aspects of diverse media platforms [28].
Rains and Brunner
[49]
found that between 1997 and 2013 more
than two-thirds of studies on social media were limited to a single
platform, most often Facebook. However, research on Twitter has
become more predominant in the last few years [
26
], given the easy
availability of the data in comparison to other platforms. Moreover,
Facebook decided to restrict the access to its API, which is likely to
impede future research on this platform.
On the other hand, there exist only a handful of studies about
political campaigns that focus on data from YouTube [
33
,
66
] or
Instagram [
17
,
39
]. However, according to a 2018 Pew Research
Center report [
47
], these two platforms have become the most pop-
ular among younger citizens, which may make them more relevant
to politics in the near future. Additionally, politicians are not bound
to use only the thoroughly-researched platforms. Germany’s chan-
cellor, Angela Merkel, for instance, has an Instagram account but
not one on Twitter.
Bossetta
[8]
presents a framework for comparing dierent plat-
forms and how their idiosyncratic features aect political campaign
strategies on social media. He exhibits dierences between plat-
forms in terms of their network structures, functionalities, and
algorithms. Our multi-platform analysis does not focus on the plat-
forms’ structure but instead deals with online interactions between
political parties and users. We hope that this focus will prove help-
ful in advancing understanding of the overall social media strategy
of political parties.
3 DATA AND METHODS
We propose a multi-platform approach that unies the dierent
features of each social media channel. The analysis is based on the
following four categories:
•Party engagement:
Quanties the activity of a political
party on a social media channel. It constitutes the main
source of interaction with the users.
•User engagement:
Represents the amount of users’ online
interaction with the political party. Interactions may be in
the form of a direct response to the social media content or a
message sent to the party’s account. User engagement does
not imply direct support for a party, given that it may have
a positive or a negative tone.
•User support:
Determines the level of users’ acceptance of
and alignment with the party’s social media content.
•Message dissemination:
Quanties the success of the party
in spreading its message across a social media channel, which
is one of the main purposes of an online political campaign.
These measures give a general overview of the elds in which a
political party is performing better than others. The specic mea-
sures used on each platform and their association with one of the
four categories are listed in Table 6. Each of these is explained
thoroughly in the next section.
In order to apply this multi-platform approach, we collected
data from Facebook, Twitter, YouTube, and Instagram using the
application programming interfaces (APIs) of each platform. We
also collected the AfD’s 2017 manifesto as a reference for the party’s
ocial ideology and proposals. From the social media channels,
we collected data for the AfD and for the other six main political
parties in Germany: CDU, Germany’s main conservative party; CSU,
the sister party of the CDU in Bavaria; Bündnis90/Die Grünen, the
green party in Germany; FDP, a neo-liberal party; SPD, Germany’s
social-democratic party; and Die Linke, the radical left party. This
allowed us to compare and measure the AfD’s eectiveness on
social media against the eectiveness of the other parties’ online
activity.
For Facebook, we retrieved posts made by the political parties in
the period from January 2015 to May 2018. These posts amounted to
a total of 12,912 posts. The data included all comments and reactions
to the posts and their respective comments. The number of posts
216

SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada J.C. Medina Serrano et al.
is smaller in comparison to those collected by Arzheimer
[1]
or
Schelter et al
. [52]
. This is due to changes in the Facebook API.
Previously, the API allowed access to drafts, status changes, and
post modications. Our data only includes the nal posts written
by each party, which is helpful, since we are not interested in
considering every modication made by the page administrators.
For Twitter, we collected the tweets from political parties’ Twitter
accounts over the one-year period, starting in July 2017. We also
included tweets from users that mentioned or retweeted the political
parties. This dataset includes 1,961,318 tweets. We also collected
tweets that included the name of one or more of the political parties.
Overall, we gathered 30,437,991 tweets.
We obtained the tweets from the parties using Twitter’s Search
API, which allows access to the last 3,200 tweets from an account.
We gathered the rest of the tweets with an automated procedure
that continuously accesses data from Twitter’s Streaming API. In
contrast to the Search API, the Streaming API allows the retrieval
of tweets in near real-time based on certain criteria like hashtags,
keywords or geolocations. The limitation of this API is that it only
provides a sample of the complete tweets. The enterprise version
of the API, called Firehose, makes it possible to query the entire
Twitter history, but its cost is prohibitive. Morstatter et al
. [38]
analyzed the dierences between Firehose and the Streaming API
and showed that samples gathered from the public API are biased.
For YouTube, we used its Data API to collect metadata from
videos published between October 2016 and May 2018. We focused
on the videos published on the YouTube channels belonging to the
political parties. The AfD has two channels, namely AfD Kompakt
and AfD-Fraktion Bundestag, while each of the other parties only
has one. The former was created in October 2016 and the latter in
December 2017, two months after this party entered the parliament.
From their two channels, we also extracted the videos’ subtitles
using the Linux package YouTube-dl. Some of the videos do not
include dialogue, but only have a written message. We manually
transcribed these videos for further analysis.
For Instagram, we collected 4,155 Instagram posts from the po-
litical parties’ accounts before the capabilities of Instagram’s public
API were diminished. The period of time matches the period we
used for collecting Facebook data. For Twitter, the period of time is
shorter, given that historical data cannot be collected with Twitter’s
public API. For YouTube, we only collected data beginning with
the creation of AfD’s rst YouTube channel.
In order to nd insights in the collected data, we applied the
following methods:
Exploratory Data Analysis. We rst performed simple qualitative
and quantitative analyses on the data. We gathered and summarized
the interactions for each platform. For Facebook and Twitter, we
also included a time series of how the parties interacted over time.
For the four dierent measures in the multi-platform schema, we
performed Kolmogorov-Smirnov tests to determine the goodness
of t to the log-normal distribution and used Vuong tests [
67
] to
compare with similar distributions. The results provide a better
intuition about how the dierent social media measures behave.
Bot Detection. A growing body of literature deals with social
bots and their inuence on politics [
60
]. Most of these studies ana-
lyze the percentage of bot activity in a given narrative. For example,
Neudert et al
. [41]
analyzed the users who tweeted hashtags related
to German political parties a month before the 2017 elections. They
found that tweets with AfD-related hashtags showed the highest
percentage of users that behaved as if they were automated. How-
ever, even though hashtags are helpful to characterize the Twitter
conversation, their use does not directly indicate support for the
party since they can be attached to both positive and negative mes-
sages. In contrast, we concentrate on users who spread the political
parties’ messages directly and look for bot behavior. We can per-
form such an analysis only on Twitter, since information about
who is sharing which political content on the other platforms is
not available.
We categorized the users who had retweeted the parties’ original
contents with the help of Botometer
1
, which is a public bot detection
framework created by the University of Indiana. The framework
implements a machine learning algorithm that has been trained
on tens of thousands of labeled examples [
65
]. For a given user,
it returns a score from 0 to 1, which determines the probability
that the user is a bot. We selected 0.5 as the threshold to classify
bots. Additionally, we only used Botometer’s language independent
features for the classication, given that the other features can only
give accurate predictions when the content is written in English.
Topic Modeling. Topic modeling algorithms are based on statis-
tical models that discover topics from a text corpus. We selected
Latent Dirichlet Allocation (LDA) given its extensive use in the
literature [
6
]. LDA takes a group of documents and treats each
document as a combination of topics. Each topic is then dened
from a collection of words. After creating a list of topics, the trained
model assigns the probability of belonging to each topic to a doc-
ument. Within each document, the probabilities sum up to one.
The probabilities and topic distributions are helpful in comparing
dierent large text corpora. Hence, we used this method to contrast
the AfD’s online content against the proposals included in their
party manifesto.
For LDA to work, the number of topics (K) must be predened.
We decided to train the model on twenty topics. The algorithms for
calculating the optimal number of topics did not converge, so we
chose the most appropriate K after experimenting with dierent
parameters. Moreover, the model requires two hyperparameters:
α
, the prior of the topic distribution; and
β
, the prior of the word
distribution. We set
α=K/
20 and
β=
0
.
01, as suggested by
Griths and Steyvers
[21]
. For the implementation, we relied on
nltk and tmtoolkit, two Python toolkits used for natural language
processing.
In the next section, we discuss the data analysis and the results
of these methods. First, we consider each social media platform
separately and then we analyze them together with the proposed
multi-platform approach.
4 FINDINGS
4.1 Facebook
For years, German political parties have used Facebook as their
main form of online communication. They interact with users by
1https://botometer.iuni.iu.edu/#!/
217

The Rise of Germany’s AfD SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada
creating posts that state their views and ideology on group pages.
Of the seven parties, the AfD has the page with the most fans, with
two times as many fans as the pages of CDU and SPD, which are
the current ruling parties. After initially losing support in early
2015, the AfD’s number of fans increased sharply when the refugee
crisis arose [
11
]. The AfD’s fan count almost doubled in a single
year, from around 140,000 to 260,000 followers. This parallels the
increase in support of the AfD shown in opinion polls. Both online
and oine, the AfD grew to be an important political force in
Germany.
Table 1 presents the results of our analysis of Facebook data. The
most active party was the AfD, with 2,363 posts. At the same time,
its post received the most comments. Each post had an average
of 420 comments. In comparison, the CDU page had an average
of 160 comments per post. Comments can have a positive or neg-
ative connotation. Hence, a high number of comments does not
directly translate into party support. Negative comments were both
in favor and against the posts’ messages. Sentiment analysis of
the comment corpus would not suce to determine party support
since the methods can only classify text into positive and negative
categories. The context is necessary for understanding the nature
of the comments.
Table 1: Facebook statistics for the German political parties
in the period from January 2015 to May 2018.
posts comments likes shares
AfD 2,363 994,191 4,168,022 2,891,377
CDU 1,690 272,155 483,924 153,131
CSU 2,162 406,804 1,897,622 634,153
Die Grünen 1,127 142,473 625,689 411,073
Die Linke 1,367 140,489 903,629 437,920
FDP 2,211 118,277 755,000 192,974
SPD 1,992 247,095 892,198 421,025
The number of shares is more representative of the party’s reach.
When a user shares a post, it appears on the timelines of the user’s
Facebook friends. Posts with more shares have reached more users
on the platform. The number of shares of the AfD’s posts is larger
than the sum of the shares of all posts from the rest of the parties.
This is a clear signal of the wide reach of the AfD on Facebook and
of its online popularity.
The CSU comes in second in terms of the numbers of comments
and shares. The CSU is a conservative party that operates only in
the state of Bavaria, while its close counterpart the CDU operates in
the rest of Germany. The CSU is more conservative than the CDU
on social issues and is closer to the political spectrum of the AfD
[
18
]. These results suggest that users with right-wing ideologies are
more politically active on Facebook. Although the CSU performed
well in terms of Facebook activity, it has lost voter support over
time. Since 2015, the CSU’s approval rating has gone down ten
points in the opinion polls.
Figure 1 shows the number of posts per month made by the pages
of the German parties. The pattern is similar for all the parties.
The large peak in the plot corresponds to the month of the 2017
parliamentary elections. In the months following the election, the
AfD continued to post content on Facebook, whereas activity from
the rest of the parties declined.
Figure 1: German political parties’ Facebook activity: Num-
ber of posts per month between January 2015 to May 2018.
We further analyzed the AfD’s posts. The format of these posts
consists of a message and an image that combines a short text
with a picture. The tone of these messages tends to be provocative
and sometimes is even sensationalist. The topics discussed are
controversial, which encourages users to engage with the posts and
express personal opinions.
To perform a quantitative analysis of the posts, we preprocessed
the text by removing stop words and punctuation marks. The most
frequent nouns are AfD, Germany, politics, EU, Merkel, Euro, Ger-
man people, SPD, and citizen. Indeed, the general message is that
the AfD is on the side of Germany and its citizens, and it is against
the Euro and the establishment parties, which are represented by
Merkel and SPD. Another word appearing on several posts is the
verb teilen (to share). This suggestion was the fth-most used verb
in the posts, indicating that part of the AfD’s strategy is to viralize
its content by explicitly asking fans to share it.
4.2 Twitter
All of the German political parties have a Twitter account that
interacts with politicians, journalists, and other users. In contrast
to Facebook, the AfD has the fewest followers on Twitter [
55
].
Nevertheless, this lack of followers does not imply that they are
less successful on this platform. For example, more than 50% of the
political conversation on Twitter on the day of the 2017 federal
election was related to the AfD [23].
As in our analysis of Facebook data, we selected four correspond-
ing measures on Twitter: number of tweets, likes, mentions and
retweets (Table 2). We divided the number of tweets into two cate-
gories: all tweets and original tweets, the latter of which are those
with content from only the party account, not including retweets
from other users. The format of the AfD’s original tweets resem-
bles that of the Facebook posts, as a brief message together with
a picture. Most of the tweets include a link to the corresponding
Facebook post. However, given the character limit, the message is
shorter on Twitter. The tweets also include hashtags. The top-used
218

SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada J.C. Medina Serrano et al.
hashtags by the AfD Twitter account are AfD, TrauDichDeutsch-
land, BTW17, Bundestag, Merkel, SPD, Gauland, CDU, FDP, and
GroKo. TrauDichDeutschland is AfD’s campaign slogan, BTW17
refers to the 2017 election, Gauland is one of the leaders of the AfD,
and GroKo is the grand coalition between CDU, CSU, and SPD.
Table 2: Twitter statistics in the period from July 2017 to July
2018.
original
tweets tweets mentions likes retweets
AfD 9,193 2,092 368,005 638,886 269,445
CDU 4,911 3,097 345,192 117,437 39,726
CSU 2,886 1,622 233,012 70,474 14,812
Die
Grünen
2,492 1,295 157,213 124,371 42,183
Die
Linke
6,809 1,776 208,047 136,440 45,856
FDP 3,149 1,730 189,687 121,220 31,487
SPD 7,480 1,782 260,056 119,507 41,803
On Twitter, the AfD was also the most active party with the
largest number of tweets. 77 percent of these tweets were retweets
and most of them were from other regional AfD Twitter accounts
or AfD politicians. SPD and Die Linke followed a similar pattern.
CDU published more original tweets than the other parties, with
66 percent of the tweets being original content. Figure 2 shows the
tweet activity over the one-year period. The AfD was more active
than the other parties most of the time. The activity of all the parties
went up during the election month and went down afterward. In
contrast to the post activity on Facebook, the AfD did not continue
to tweet at the same pace as during the months before the election.
Figure 2: German political parties’ Twitter activity: Number
of tweets per month during a one-year period.
Like Facebook comments, mentions allow users to reply to a
tweet or send a message directly to the party account by using
the @ symbol and the screen name of the account. We excluded
retweets that included a mention of a political party from this
analysis. Additionally, if a tweet mentioned more than one party,
each mention was counted separately. The results show that the
AfD received the most mentions and CDU came in the close second
place. Even though Die Grünen has the most followers, it had the
fewest mentions. Note that the mentions in our data come from the
sample that the Streaming API provides, which is only a selection
of all mentions published to the platform over the period of interest.
While mentions can contain positive or negative messages about
the party, likes and retweets serve as measures of support [
24
].
For both these measures, there is a large dierence between the
AfD values and those of the other parties. Like shares on Facebook,
the AfD’s tweets were retweeted more than the tweets of all other
parties combined. Retweets and likes can only originate from the
account’s original tweets. On average, each of the AfD’s original
tweets was retweeted 129 times and had 305 likes. For compari-
son, each CDU tweet was, on average, retweeted 13 times and had
38 likes. This corresponds to a dierence of one whole order of
magnitude.
Not only the total number of retweets is of relevance, but also
the information on which users retweeted the party accounts. With
this data, we could investigate how many of the party retweets
were published by social bot accounts. We obtained the Botometer
score from all the users who had retweeted a political party during
the month of September 2017. This subset of data included 111,919
retweets from 22,396 unique users. We assigned accounts that were
closed by Twitter directly to the bot category. After classifying
the users, we calculated the percentage of retweets from the bot
accounts. Table 3 shows the results for each party. With almost 33
percent, the AfD is the party with the maximum number of bots
retweeting each party’s content. The CDU, CSU, and FDP follow
with bot retweets between 20 and 25 percent. The parties with the
lowest percentage of bot retweets are SPD, Die Grünen and Die
Linke.
Table 3: The number of retweets in September 2017 and
the percentage of those that belong to accounts classied as
bots.
retweets bots(%)
AfD 43,633 32.92%
CDU 14,603 21.15%
CSU 3,738 24.02%
Die Grünen 15,440 15.68%
Die Linke 12,888 12.97%
FDP 7,905 23.45%
SPD 13,712 12.99%
Note that these results do not provide an exact quantity of bot
activity since the bot detection methods are not always accurate.
Even the term bot is dened loosely across the literature [
20
]. Nev-
ertheless, the comparison between parties is of relevance since the
percentages vary considerably. Interestingly, center, center-right,
219

The Rise of Germany’s AfD SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada
and right-wing political parties have a higher proportion of bot
retweets. Hence, we conclude that automated accounts on Twitter
were more likely to spread messages from the center and right side
of the German political spectrum.
4.3 YouTube
YouTube is dierent from the other platforms in its focuses on
videos instead of text and images. YouTube allows political parties
to publish video content that can then be shared on other social
networks. In the rst years after its conception, the AfD did not
use a YouTube channel to spread its message. The channel AfD
Kompakt was rst created in October 2016 and is closely related
to the AfD’s member magazine of the same name. It claims in its
description to be the AfD’s ocial YouTube channel. Three months
after the 2017 elections, the AfD faction in parliament created a
new channel called AfD-Fraktion Bundestag. We considered both
channels in our analysis since both seek to spread ocial AfD
content. The other parties have only one YouTube channel with
ocial videos.
Table 4 shows the party activity and user response to the parties
on YouTube. The party activity varies, with Die Grünen having
published the most videos in the period of interest. The dierence
between parties is greater when taking into account the number of
comments on the videos. The dierence is not entirely due to user
interaction, but is aected by the fact that the comment sections of
some videos were disabled. For CSU, FDP and Die Grünen, several
videos have comments deactivated, and the AfD Kompakt channel
does not allow any comments on its videos. The 8,080 comments
on the AfD’s videos are therefore all from its second channel. Even
though this channel has been active only since December 2017, the
AfD-Fraktion Bundestag’s videos have more comments and likes
than any of the other parties’ channels.
Table 4: Youtube statistics for the German political parties
in the period from October 30 2016 to May 2018.
videos comments likes dislikes views
AfD 454 8,080 60,375 2,458 2,049,008
CDU 264 4,988 5,574 9,926 385,262
CSU 166 44 2,410 1,728 482,586
Die Grü-
nen
479 665 27,480 26,199 5,283,833
Die Linke 204 5,791 20,043 6,912 814,219
FDP 63 910 341 90 853,673
SPD 236 3,573 22,469 22,348 2,418,132
A unique YouTube feature is the option to dislike a video. A
dislike depends on the context of the video and does not always
translate into opposition to the party. The AfD channels have the
lowest ratio of dislikes to likes. Both Die Grünen and SPD have
nearly equal numbers of likes and dislikes, whereas the CDU has
considerably more dislikes than likes.
Video popularity is measured by the number of views. As with
the publishing activity, Die Grünen’s videos have the most views.
The SPD has the channel with most views after Die Grünen even
though they posted less than half the number of the videos. The
AfD comes in third place for this measure. From all the parties,
the most-seen videos are the campaign commercials. Most of the
videos are of press conferences, campaign talks, and appearances
in the German parliament by politicians of the party. These videos
receive less attention than the others.
4.4 Instagram
Little research has been published about Instagram’s potential in
political campaigns [
51
]. Thomson and Greenwood
[61]
found that
users in the US are less likely to engage with political images on
Instagram. In contrast, a survey in Germany by Eckerl and Hahn
[13]
showed that the platform holds great potential for political
communication. In the collected data, we observe active political
campaigns from the German political parties. Each has an active
Instagram account.
Table 5 shows that the AfD takes second place in terms of activity
after the CSU. Even so, user interaction with the AfD’s posts is
greater than interaction with the CSU’s posts. Die Linke has a
similar number of posts and likes as the AfD, but their posts attract
signicantly fewer comments. Indeed, the number of comments
on the AfD’s posts is larger than the sum of the comments on the
posts of all the other parties.
The AfD’s strategy on Instagram is like its strategy on Facebook.
Most of the party’s Instagram posts are a subset of the Facebook
posts. They include an image that embeds a short text and have
a longer text in the image description. Since Instagram is more
image oriented, the highlight of the information is embedded in
the image. The top nouns used in the descriptions are the same as
those used in the Facebook posts but are ranked in a dierent order.
The subset of messages in these Instagram posts is representative
of AfD’s messages on Facebook.
Table 5: Instagram statistics for all the German political par-
ties in the period from January 2015 to May 2018
posts comments likes
AfD 870 68,399 469,380
CDU 229 4,702 122,353
CSU 958 4,880 154,291
Die Grünen 514 16,689 317,048
Die Linke 825 13,313 418,208
FDP 584 8,397 325,941
SPD 175 4,098 105,067
4.5 Social Media Comparison
In order to compare and summarize the results of the previous
subsections, we applied the aforementioned multi-platform schema.
Table 6 lists which features of each of the social media platforms
220

SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada J.C. Medina Serrano et al.
Table 6: Features on each social media channel divided into the four categories from the multi-platform analysis. The col-
lected data from the features in bold letters are more likely to follow a log-normal distribution and give better ts than the
exponential, Poisson and power-law distribution according to the KS and Vuong tests. This is also the case for the features in
italics, with the only dierence that the log-normal distribution did not give a better t than the exponential distribution.
Facebook Twitter YouTube Instagram
party engagement posts tweets videos posts
user engagement comments mentions comments comments
user acceptance likes likes likes likes
message dissemination shares retweets views —
falls into each one of the four categories. The only missing entry
on the table is the message dissemination measure for Instagram.
This would correspond to the number of times the pictures from a
party account have been observed. This information is private and
is only accessible to the owner of a business account, and is even
then restricted to the account’s own posts.
With the aid of the four measures introduced in this table, we
proceed to compare the level of eectiveness of the AfD’s activ-
ity on social media. Unfortunately, a quantitative cross-platform
comparison is not possible, since the period of data collection dif-
fers between the social media platforms. Across the social media
channels, the AfD has a high party engagement, taking either the
rst or second place in this category. This is part of their aforemen-
tioned social media strategy. In terms of user engagement, the AfD
surpasses the rest of the parties by a considerable margin. As men-
tioned before, the tone of the party’s message, together with the
controversial topics they tend to discuss, such as immigration and
the economic crisis, encourage users to engage more with the posts.
The same superiority is shown in user support. In this category,
the dierence between the AfD and the rest of the parties is even
larger than the dierence in user engagement. This suggests that
the user base that supports the party is also regularly active on so-
cial media. Regarding the last category, the AfD spread its message
on Facebook and Twitter more eectively than the other German
political parties by an order of magnitude. The same dominance
is not observed on YouTube. Nevertheless, the fact that the AfD
decided to open a channel after entering the parliament suggests
that they now consider YouTube as part of the party’s social media
strategy. In the next few years, the AfD may prove to be as success-
ful in spreading their message on YouTube as it has been on the
other two platforms.
For the measures of user engagement, user acceptance and mes-
sage dissemination categories, we provided aggregated results that
were independent of the party activity. We deliberately did not
focus on eciency measures like the average number of Facebook
likes per post or the average number of views per YouTube video in
this analysis since the social media features are far from following a
normal distribution. The great majority of social media posts attract
little user engagement, in contrast to the very few posts that draw
large amounts of attention. Indeed, research has shown that most
complex social media interactions follow a log-normal distribution
[
2
,
32
,
69
]. We assessed this statement by testing the distributions of
the features per unit of social media activity. This analysis consid-
ered the four social media channels and all seven German political
parties
2
. We used bootstrapped Kolmogorov-Smirnov tests with 50
samples and a p-value of 0.05. In all cases, the tests failed to reject
the null hypothesis that the data is generated from a log-normal
distribution. We further compared the log-normal distribution with
the exponential, power-law and Poisson distribution by implement-
ing Vuong tests with a p-value of 0.05. In all cases, the log-normal
distribution was a better t than the Poisson or the power-law dis-
tributions. However, in some cases, the Vuong test failed to show
that the log-normal distribution gave a better t than the expo-
nential distribution. The features with at least one case of a failing
Vuong test are shown in Table 6 in italic letters. On the other hand,
the features in bold letters t better to the log-normal distribution
than to the other three distributions for the data from all of the
political parties.
4.6 Discourse Comparisons
The last step in our analysis consisted of exploring the dierence
between how the AfD presents its goals to followers online and
the content of party’s explicitly stated political intentions and mo-
tives. For this analysis, we only considered the Facebook posts,
YouTube videos and Instagram posts created during the same one-
year period over which we collected the Twitter data, so that the
period of interest for all four channels was the same. With the help
of topic modeling, we compared the topics between the dierent
communication channels with the topics included in the party’s
manifesto.
For this analysis, we treated each post, tweet and video as a
document. We divided the text of the manifesto into paragraphs
and dened each of these paragraphs as a document. The resulting
corpora include 1,113 Facebook posts, 9,213 tweets, 436 YouTube
video captions, 866 Instagram posts, and 395 paragraphs from the
manifesto. Preprocessing was applied to eliminate stop words, punc-
tuation marks, and applying the Snowball stemming algorithm to
the remaining words. We additionally removed the string "RT" from
the retweets.
Since the topics are created algorithmically, the interpretation of
each topic relies on human curation. To compare the ve corpora,
we selected the topics that were directly connected to the economy
2
We did not evaluate the YouTube comments since many videos have them blocked,
and the Twitter mentions since they do not always represent a response to a tweet.
221

The Rise of Germany’s AfD SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada
or immigration. We then calculated the percentage of documents
that included these topics. For this calculation, we summed up the
probabilities of each document from the selected topics and then
divided by the number of topics. Table 7 shows that economy and
immigration are treated equally in the manifesto, whereas in the
AfD’s Facebook, Twitter and Instagram content, immigration topics
are discussed signicantly more than economic topics.
In the case of YouTube, only 18% of the total discourse is related
to immigration or economic topics, and these are discussed in a
similar proportion. This exception can be attributed to the fact that
in contrast to the other social media platforms, there is a lack of
platform-specic generated content. Most of the videos on the AfD’s
channels are political speeches, which represent the oral discourse
of the right-wing party which is not necessarily related to its social
media message. Further research into discourse comparisons could
focus on the dierences between the speeches given at campaign
events and those given during parliament sessions.
By comparison, immigration topics are discussed more frequently
on Twitter and Instagram than on Facebook. Retweets were in-
cluded in the Twitter data, perhaps explaining the 6% dierence be-
tween the Twitter and Facebook corpora. For Instagram, the AfD’s
content is mostly a subsample of the Facebook posts, which indi-
cates that the AfD deliberately favored immigration-related content
on this platform. With this discourse analysis, we prove quantita-
tively that the AfD-generated content on social media downplayed
the party’s economic proposals and focused on immigration topics,
which validates the analysis published by Kim [30].
Table 7: Percentage of documents related to topics dis-
cussing economy or immigration on dierent platforms.
economy immigration
Manifesto 21% 19.2%
Facebook 4.5% 16.1%
Twitter 4.7% 21.8%
Youtube 9.9% 8.2%
Instagram 7.6% 28%
5 CONCLUSION
Since its foundation, the AfD has used social media as its primary
communication tool. In this paper, we proved that the AfD’s online
activity prompted the most user interactions of any German politi-
cal party. Our analysis covers four social platforms over a longer
period of time than has been considered in previous research. We
mapped the AfD’s social media strategy and illustrated some of the
dierences between the party’s online discourse and its manifesto.
We also conrmed that that the spread of the AfD’s message was
boosted by automated accounts to some extent.
Although we cannot prove a direct connection between poll
gains and social media dominance, we conclude that the AfD suc-
ceeded in spreading its message on social media. This message has
also entered the limelight in traditional media and has permeated
Germany’s public agenda. The success of the AfD’s social media
campaign together with traditional media coverage was essential
in spreading and stimulating anti-establishment sentiment, which
partially explains the rise of a far-right party in Germany.
The AfD’s eective social media campaign raises the question
of what strategies the other political parties will take to improve
their online activity. There is no simple solution that can attend
to this question. Digital campaigns have to evolve and take into
consideration the new trends. In particular, they have to be creative
to attract the younger citizens and at the same time reach those
citizens that are disappointed with current politics.
Overall, the results suggest that there exists a shift in Germany’s
online political communication induced by AfD’s social media dom-
inance. Future research should explore the causes and eects of
this shift. For example, it is plausible that the shift has led to an
increase in online polarization or to user discussions becoming
more aggressive. These phenomena have to be analyzed taking
into consideration the potential biased reality caused by automated
accounts.
The main contribution in this paper is the unied multi-platform
analysis that classies social media features into four categories.
We hope that this cross-platform analysis can help scholars in the
study of political parties and their campaigns around the globe. We
emphasize the necessity for continuous and rigorous research in this
eld for two reasons: to cope with recent changes in digital media
that directly aect online political interaction, and to understand
the emergence and rise to dominance of right-wing populist parties
around the world.
REFERENCES
[1]
Kai Arzheimer. 2015. The AfD: Finally a Successful Right-Wing Populist Eu-
rosceptic Party for Germany? West European Politics 38, 3 (2015), 535–556.
[2]
Sitaram Asur, Bernardo A. Huberman, Gábor Szabó, and Chunyan Wang. 2011.
Trends in Social Media : Persistence and Decay. 5th International AAAI Conference
on Weblogs and Social Media (2011).
[3]
Justus Bender and Marvin Oppong. 2017. Frauke Petry und die Bots. FAZ.
Retrieved March 01, 2019 from https://www.faz.net/aktuell/politik/digitaler-
wahlkampf-frauke-petry-und-die-bots-14863763.html.
[4]
Nicole Berbuir, Marcel Lewandowsky, and Jasmin Siri. 2016. The AfD and its
Sympathisers: Finally a Right-Wing Populist Movement in Germany? German
Politics 24 (Jan. 2016), 154–178.
[5]
Knut Bergmann, Matthias Diermeier, and Judith Niehues. 2016. Die AfD - eine
Partei der Besserverdiener? (2016).
[6]
David M. Blei, Andrew Y. Ng, Michael I. Jordan, and John Laerty. 2003. Latent
Dirichlet allocation. Journal of Machine Learning Research 3 (2003), 993–1022.
[7]
Leticia Bode and Emily K. Vraga. 2018. Studying Politics Across Media. Political
Communication 35, 1 (2018), 1–7.
[8]
Michael Bossetta. 2018. The Digital Architectures of Social Media: Comparing
Political Campaigning on Facebook, Twitter, Instagram, and Snapchat in the 2016
U.S. Election. Journalism & Mass Communication Quarterly 95, 2 (2018), 471–496.
[9]
Oliver Decker, Alexander Yendell, Johannes Kiess, and Elmar Brähler. 2017. Polar-
isiert und radikalisiert? Medienmisstrauen und die Folgen für die Demokratie.
[10]
Dimap.de. 2017. WahlREPORT Bundestagswahl. Retrieved March 01, 2019 from
https://www.infratest-dimap.de/umfragen-analysen/bundesweit/wahlreport-
deutschland/2017/.
[11]
Paul-Jasper Dittrich. 2017. Social networks and populism in the EU: Four things
you should know.
[12]
Julia Ebner. 2017. How Germany’s far right took over Twitter - and
tilted the election. The Guardian. Retrieved March 01, 2019
from https://www.theguardian.com/commentisfree/2017/sep/26/germany-far-
right-election-afd-trolls.
[13]
Thomas Eckerl and Oliver Hahn. 2018. Die Sele-Seite der Macht: Instagram
in der politischen Kommunikation in Deutschland. In Strategische Politische
Kommunikation im digitalen Wandel: Interdisziplinäre Perspektiven auf ein dy-
namisches Forschungsfeld, Michael Oswald and Michael Johann (Eds.). Springer
Fachmedien Wiesbaden, Wiesbaden, 237–257.
[14]
Sven Engesser, Nicole Ernst, Frank Esser, and Florin Büchel. 2017. Populism
and social media: How politicians spread a fragmented ideology. Information,
222

SMSociety ’19, July 19–21, 2019, Toronto, ON, Canada J.C. Medina Serrano et al.
Communication & Society 20, 8 (2017), 1109–1126.
[15]
Rui Fan, Jichang Zhao, Yan Chen, and Ke Xu. 2013. Anger is More Inuential
Than Joy: Sentiment Correlation in Weibo. arXiv preprint arXiv:1309.2402 (2013).
[16]
Emilio Ferrara, Onur Varol, Clayton A. Davis, Filippo Menczer, and Alessandro
Flammini. 2016. The rise of social bots. Commun. ACM 59, 7 (2016), 96–104.
[17]
Kirill Filimonov, Uta Russmann, and Jakob Svensson. 2016. Picturing the Party:
Instagram and Party Campaigning in the 2014 Swedish Elections. Social Media +
Society 2, 3 (2016).
[18]
Simon Franzmann. 2014. Die Wahlprogrammatik der AfD in vergleichender Per-
spektive. Mitteilungen des Instituts für Deutsches und Internationales Parteienrecht
und Parteienforschung 20 (Jan. 2014), 115–124.
[19]
Paolo Gerbaudo. 2018. Social media and populism: an elective anity? Media,
Culture & Society 40, 5 (2018), 745–753.
[20]
Robert Gorwa and Douglas Guilbeault. 2018. Unpacking the Social Media Bot: A
Typology to Guide Research and Policy. Policy & Internet (2018).
[21]
T. L. Griths and M. Steyvers. 2004. Finding scientic topics. Proceedings of the
National Academy of Sciences 101, Suppl. 1 (April 2004), 5228–5235.
[22]
Robert Grimm. 2015. The rise of the German Eurosceptic party Alternative für
Deutschland, between ordoliberal critique and popular anxiety. International
Political Science Review 36, 3 (2015), 264–278.
[23]
Simon Hegelich. 2018. Die digitale Revolution der Meinungsbildung Social Media
im Wahlkampf. Hans Seidel Stiftung, Munich, Germany.
[24]
Simon Hegelich and Morteza Shahrezaye. 2015. The Communication Behavior of
German MPs on Twitter: Preaching to the Converted and Attacking Opponents.
European Policy Analysis 1 (2015), 155–174.
[25]
Marc Helbling and Anke Tresch. 2011. Measuring party positions and issue
salience from media coverage: Discussing and cross-validating new indicators.
Electoral Studies 30, 1 (2011), 174–183.
[26]
Andreas Jungherr. 2016. Twitteruse in ele ction campaigns: A systematic literature
review. Journal of Information Technology & Politics 13, 1 (2016), 72–91.
[27]
Bente Kalsnes. 2016. The social media paradox explained: comparing political
parties’ Facebook strategy versus practice. Social Media+ Society 2, 2 (2016).
[28]
David Karpf. 2017. Analytic Activism: Digital Listening and the New Political
Strategy. Oxford University Press, Oxford, UK.
[29]
Majid Khosravinik. 2017. Right Wing Populism in the West: Social Media Dis-
course and Echo Chambers. Insight Turkey 19 (Jan. 2017).
[30]
Juho Kim. 2018. The radical market-oriented policies of the Alternative for Ger-
many (AfD) and support from non-beneciary groups – discrepancies between
the party’s policies and its supporters. Asian Journal of German and European
Studies 3, 1 (21 May 2018), 6.
[31]
Anders O Larsson and Bente Kalsnes. 2014. ’Of course we are on Facebook’:
Use and non-use of social media among Swedish and Norwegian politicians.
European Journal of Communication 29, 6 (2014), 653–667.
[32]
Kristina Lerman and Rumi Ghosh. 2010. Information Contagion: an Empirical
Study of the Spread of News on Digg and Twitter Social Networks. arXiv preprint
arXiv:1003.2664 (2010).
[33]
Azi Lev-On. 2012. YouTube Usage in Low-Visibility Political Campaigns. Journal
of Information Technology & Politics 9, 2 (2012), 205–216.
[34]
Viktor Mayer-Schnberger. 2013. Big Data: A Revolution That Will Transform How
We Live, Work and Think. Viktor Mayer-Schnberger and Kenneth Cukier. John
Murray Publishers, UK.
[35]
Gianpietro Mazzoleni and Roberta Bracciale. 2015. Socially mediated populism:
the communicative strategies of political leaders on Facebook. Palgrave Commu-
nications 4, 50 (2015).
[36] Juan Carlos Medina Serrano, Simon Hegelich, Morteza Shahrezaye, and Orestis
Papakyriakopoulos. 2018. Social Media Report: The 2017 German Federal Elections.
TUM University Press, Munich, Germany.
[37]
Yves Meny and YvesSurel. 2002. Democracies and the Populist Challenge. Palgrave
Macmillan UK.
[38]
Fred Morstatter, Jürgen Pfeer, Huan Liu, and Kathleen M Carley. 2013. Is the
Sample Good Enough? Comparing Data from Twitter’s Streaming API with Twit-
ter’s Firehose. In ICWSM. Proceedings of the seventh International Conference
on Weblogs and Social Media, Cambridge, Massachusetts.
[39]
Caroline Lego Munoz and Terri L. Towner. 2017. The Image is the Message:
Instagram Marketing and the 2016 Presidential Primary Season. Journal of
Political Marketing 16, 3-4 (2017), 290–318.
[40]
Christian Nestler and Jan Rohgal. 2014. Eine deutsche Angst - Erfolgreiche
Parteien rechts von der Union: Zur AfD und den gegenwärtigen Gelegen-
heitsstrukturen des Parteienwettbewerbs. Zeitschrift für Politik 61, 4 (Dec. 2014),
389–413.
[41]
Lisa-Maria Neudert, Bence Kollanyi, and Philip N. Howard. 2017. Junk News
and Bots during the German Parliamentary Election: What are German Voters
Sharing over Twitter?
[42]
Oskar Niedermayer and Jürgen Hofrichter. 2016. Die Wählerschaft der AfD:
Wer ist sie, woher kommt sie und wie weit rechts steht sie? Zeitschrift für
Parlamentsfragen 47 (Jan. 2016), 267–285.
[43]
Pippa Norris. 2003. Preaching to the Converted?: Pluralism, Participation and
Party Websites. Party Politics 9, 1 (2003), 21–45.
[44]
Karoline Oelsner and Linette Heimrich. 2015. Social media use of German
politicians: towards dialogic voter relations? German Politics 24, 4 (2015), 451–
468.
[45]
Orestis Papakyriakopoulos, Simon Hegelich, Morteza Shahrezaye, and Juan Car-
los Medina Serrano. 2018. Social media and microtargeting: Political data pro-
cessing and the consequences for Germany. Big Data & Society 5, 2 (2018).
[46]
Pew Research Center: Global Attitudes & Trends. 2018. News Media
and Political Attitudes in Germany. Retrieved March 01, 2019
from https://www.pewglobal.org/fact-sheet/news-media-and-political-attitudes-
in-germany/.
[47]
Pew Research Center: Internet & Technology. 2018. Teens, Social
Media & Technology 2018. Retrieved March 01, 2019 from
https://www.pewinternet.org/2018/05/31/teens-social-media-technology-2018/.
[48]
Neil Postman. 1970. The reformed English curriculum. In High school 1980: The
shape of the future in American secondary education, Alvin C. Eurich (Ed.). Pitman
Publishing Corporation, New York, N.Y., 160–2168.
[49]
Stephen A Rains and Steven R Brunner. 2015. What can we learn about social
network sites by studying Facebook? A call and recommendations for research
on social network sites. New Media & Society 17, 1 (2015), 114–131.
[50]
Patricia Rossini, Je Hemsley, Sikana Tanupabrungsun, Feifei Zhang, and Jennifer
Stromer-Galley. 2018. Social Media, Opinion Polls, and the Use of Persuasive
Messages During the 2016 US Election Primaries. Social Media + Society 4, 3
(2018).
[51]
Uta Russmann and Jakob Svensson. 2017. Interaction on Instagram?: Glimpses
from the 2014 Swedish Elections. IJEP 8 (2017), 50–66.
[52]
Sebastian Schelter, Felix Biessmann, Malisa Zobel, and Nedelina Teneva. 2016.
Structural Patterns in the Rise of Germany’s New Right on Facebook. In ICDM
Workshops. IEEE Computer Society, Barcelona, Spain, 440–445.
[53]
Franzisca Schmidt. 2018. Drivers of Populism: A Four-country Comparison of
Party Communication in the Run-up to the 2014 European Parliament Elections.
Political Studies 66, 2 (2018), 459–479.
[54]
Rüdiger Schmitt-Beck. 2017. The ’Alternative für Deutschland in the Electorate’:
Between Single-Issue and Right-Wing Populist Party. German Politics 26, 1 (2017),
124–148.
[55]
Statista. 2017. Ranking der beliebtesten Parteien bei Twitter nach der Anzahl
der Follower in Deutschland im August 2017. Retrieved March 01, 2019
from https://de.statista.com/statistik/daten/studie/659207/umfrage/anzahl-der-
twitter-follower-von-parteien-in-deutschland/.
[56]
Stefan Stieglitz and Linh Dang-Xuan. 2013. Social media and political communi-
cation: a social media analytics framework. Social Network Analysis and Mining
3, 4 (2013), 1277–1291.
[57]
Sebastian Stier, Arnim Bleier, Haiko Lietz, and Markus Strohmaier. 2018. Election
Campaigning on Social Media: Politicians, Audiences, and the Mediation of
Political Communication on Facebook and Twitter. Political Communication 35,
1 (2018), 50–74.
[58]
Jesper Strömbäck and Peter Van Aelst. 2013. Why political parties adapt to the
media: Exploring the fourth dimension of mediatization. International Communi-
cation Gazette 75, 4 (2013), 341–358.
[59]
Jennifer Stromer-Galley. 2014. Presidential campaigning in the Internet age. Oxford
University Press, Oxford, U.K.
[60]
Andree Thieltges, Orestis Papakyriakopoulos, Juan Carlos Medina Serrano, and
Simon Hegelich. 2018. Eects of Social Bots in the Iran-Debate on Twitter. arXiv
preprint arXiv:1805.10105 (2018).
[61]
T.J. Thomson and Keith Greenwood. 2017. I "Like" That: Exploring the Character-
istics That Promote Social Media Engagement With News Photographs. Visual
Communication Quarterly 24, 4 (2017), 203–218.
[62]
Anke Tresch, Jonas Lefevere, and Stefaan Walgrave. 2018. How parties’ issue
emphasis strategies vary across communication channels: The 2009 regional
election campaign in Belgium. Acta Politica 53, 1 (2018), 25–47.
[63]
Sonja Utz. 2009. The (potential) benets of campaigning via social network sites.
Journal of computer-mediated communication 14, 2 (2009), 221–243.
[64]
José van Dijck and Thomas Poell. 2013. Understanding Social Media Logic. Media
and Communication 1, 1 (2013), 2–14.
[65]
Onur Varol, Emilio Ferrara, Clayton A. Davis, Filippo Menczer, and Alessandro
Flammini. 2017. Online Human-Bot Interactions: Detection, Estimation, and
Characterization. arXiv preprint arXiv:1703.03107 (2017).
[66]
Lucia Vesnic-Alujevic and Soe Van Bauwel. 2014. YouTube: A Political Advertis-
ing Tool? A Case Study of the Use of YouTube in the Campaign for the European
Parliament Elections. Journal of Political Marketing 13, 3 (2014), 195–212.
[67]
Quang H. Vuong. 1989. Likelihood Ratio Tests for Model Selection and Non-
Nested Hypotheses. Econometrica 57, 2 (1989), 307–333.
[68]
Markus Wagner and Thomas M Meyer. 2014. Which issues do parties emphasise?
Salience strategies and party organisation in multiparty systems. West European
Politics 37, 5 (2014), 1019–1045.
[69]
Fang Wu and Bernardo A Huberman. 2007. Novelty and collective attention.
Proceedings of the National Academy of Sciences of the United States of America
104, 45 (2007).
223