TDP As the Party of Hope For AP Youth Under N Chandrababu Naidu’s Leadership
Breaking the Barrier: Interactive Election Campaign Communication on Twitter
1. Breaking the Barrier
Interactive Election Campaign
Communication on Twitter
during the German General Election 2009
Pascal Jürgens
U of Mainz, Germany
pascal.juergens@gmail.com
Andreas Jungherr
U of Bamberg, Germany
andreas.jungherr@gmail.com
3. 2
Twitter and Politics
The 2009 German General Election started
with the impression of Obama’s online
campaign fresh in mind
4. 2
Twitter and Politics
The 2009 German General Election started
with the impression of Obama’s online
campaign fresh in mind
Due to several high-profile incidents, German
media and politics focussed on Twitter at
least as much as on other social networks
8. Twitter Overview
»Micro-publishing« — publish short messages
Re-Tweet — quote message including attribution
Directed (@-)message — explicitly sent to a
recipient
3
9. Twitter Overview
»Micro-publishing« — publish short messages
Re-Tweet — quote message including attribution
Directed (@-)message — explicitly sent to a
recipient
Topic tags (#) — defines the topic of the message
3
10. Twitter Overview
»Micro-publishing« — publish short messages
Re-Tweet — quote message including attribution
Directed (@-)message — explicitly sent to a
recipient
Topic tags (#) — defines the topic of the message
High degree of mobile usage
(~15% in our dataset)
3
12. Data Collection
Bootstrap: List of political twitter users
(Assembled from several websites compiling lists of politicians on
twitter)
4
13. Data Collection
Bootstrap: List of political twitter users
(Assembled from several websites compiling lists of politicians on
twitter)
Growth: perpetual search for #tags
(Add new users to sample)
4
14. Data Collection
Bootstrap: List of political twitter users
(Assembled from several websites compiling lists of politicians on
twitter)
Growth: perpetual search for #tags
(Add new users to sample)
Capture: Collect all new tweets
(Also crawling archives for coverage over entire time range)
4
15. Data Collection
Bootstrap: List of political twitter users
(Assembled from several websites compiling lists of politicians on
twitter)
Growth: perpetual search for #tags
(Add new users to sample)
Capture: Collect all new tweets
(Also crawling archives for coverage over entire time range)
Graphs: Nightly snapshot of friend/followers
4
18. Dataset
Three months prior to General Election
A sample of Germany’s politically active
twitter users — 33 048 individuals
5
19. Dataset
Three months prior to General Election
A sample of Germany’s politically active
twitter users — 33 048 individuals
A complete archive of their communication
(public tweets) — 10 109 894 messages
5
20. Dataset
Three months prior to General Election
A sample of Germany’s politically active
twitter users — 33 048 individuals
A complete archive of their communication
(public tweets) — 10 109 894 messages
A temporal map of their connections
5
25. Defining Links
Followers are a questionable indicator of
influence (Cha et al: “The Million Follower Fallacy”)
9
26. Defining Links
Followers are a questionable indicator of
influence (Cha et al: “The Million Follower Fallacy”)
Intentional, meaningful interaction as a link:
@-messages and quotes (re-tweets)
9
27. Network Structure
10
0 2000 4000 6000 8000
0.00.20.40.60.81.0
Degree Distribution
in-degree
cumulativefrequency
1 10 100 1000 10000
1e-041e-031e-021e-011e+00
Degree Distribution
in-degree
highly connected
users are
infrequent
little connected
users make up
the majority
(Clustering Coefficient C = .045 Random Erdös-Rényi Graph C = .001)
Distribution of Incoming Links (cumulative)
30. Follower gain over sample timespan
correlates with intensity of interaction
(Spearman’s Rho rs = .54, two-tailed p ≃ 0)
12
«The Rich get Richer»
31. Follower gain over sample timespan
correlates with intensity of interaction
(Spearman’s Rho rs = .54, two-tailed p ≃ 0)
Corresponds to “preferential attachement”
theory on network growth
New participants attach (follow) to most visible nodes
12
«The Rich get Richer»
34. Content Analysis
Hand-coded a sample from each of the 50
most prominent users
Prominence: Volume of meaningful
communication (direct messages + re-tweets)
13
35. Content Analysis
Hand-coded a sample from each of the 50
most prominent users
Prominence: Volume of meaningful
communication (direct messages + re-tweets)
Coding for content topic, references, links
13
41. Results
Reach on twitter is very dependent on a small
group of users (new gatekeepers)
Preferential attachement makes entry into
twitter ecosystem difficult
16
42. Results
Reach on twitter is very dependent on a small
group of users (new gatekeepers)
Preferential attachement makes entry into
twitter ecosystem difficult
Politicians are only successful if they attach
to existing topics/trends/conventions? (e.g.
Jörg Tauss, Piratenpartei)
16
45. Results II
Dedicated “political” communities do not
play a significant role.
Dedicated “political” users do (mostly) not
play a significant role.
17
46. Results II
Dedicated “political” communities do not
play a significant role.
Dedicated “political” users do (mostly) not
play a significant role.
Political communication happens ad-hoc in
issue-driven topics among non-political
tweets and topics.
17
49. Politics as One Among Many Topics
19
Translated:
“I’ll be glad once the election campaigns
are over and we can all like each other
again. Especially once the dull discussions
come to an end.”
51. 21
List of hashtags identifying political topics
cducsu spd fdp gruene grüne piraten npd linke zensursula
bundestagswahl petition politik cduremix09 btw09 wahl sst
linkspartei union tvduell
52. 22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131
53. Explorative factor analysis not fruitful:
22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131
54. Explorative factor analysis not fruitful:
22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131
55. Explorative factor analysis not fruitful:
22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131
56. Explorative factor analysis not fruitful:
22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131
57. Explorative factor analysis not fruitful:
22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131
58. Explorative factor analysis not fruitful:
(p-value 8.870000×10^-154, Chi square of 727.04 on 6 degrees of freedom)
22
No Clear-Cut Style
Loadings:
Factor1 Factor2 Factor3 Factor4
life -0.902 -0.316 -0.105 -0.267
mind 0.973 -0.199
work 0.996
auto 0.993
polit 0.485 0.188 0.800 0.110
level 0.370 0.167 0.678
ext 0.156 0.152
diag -0.118 -0.154
self -0.131