Updated Bundestag presentation at the SciCAR19 conference in Dortmund on September 10, 2019.

1. Social Network Analysis –
Insights from mapping the Twitter network of
the German Bundestag
by Harald Meier - harald@smrfoundation.org
SciCAR19 – 10. September 2019 – Dortmund

2. Social Network Analysis (SNA)
▪ Measuring and mapping collections of connections
▪ Describing the position of an individual within a network
NETWORKS AND SOCIAL NETWORK ANALYSIS
Network
A network consists of VERTICES and EDGES.
An EDGE is a connection between two VERTICES.
Twitter Network
VERTEX Twitter User
EDGE tweet, retweet, mention, reply, favourite, follow

3. SOCIAL NETWORK ANALYSIS
3
Network Overview
• Density / Modularity
Group Analysis
• Cluster Algorithm
• Density
Vertex Metrics
• Centrality: Betweenness,
Closeness, Eigenvector, …
Content Analysis
• Top hashtags, words, URLs, …
• Sentiment, time series
Layout Algorithms
• Group-In-A-Box: Treemap
• Harel-Koren Fast Multiscale

4. 4

5. 5

6. [Divided]
Polarized Crowds
[Unified]
Tight Crowd
[Fragmented]
Brand Clusters
[Clustered]
Community Clusters
[In-Hub & Spoke]
Broadcast Network
[Out-Hub & Spoke]
Support Network
TWITTER NETWORK SHAPES
PEW Report: Mapping Twitter Topic Networks: From Polarized Crowds to Community Clusters. PEW Research Report 2014:
http://www.pewinternet.org/2014/02/20/mapping-twitter-topic-networks-from-polarized-crowds-to-community-clusters/

7. 1
[Divided]
Polarized Crowds
[Unified]
Tight Crowd
[Fragmented]
Brand Clusters
[Clustered]
Community Clusters
[In-Hub & Spoke]
Broadcast Network
[Out-Hub & Spoke]
Support Network
TWITTER NETWORK SHAPES
PEW Report: Mapping Twitter Topic Networks: From Polarized Crowds to Community Clusters. PEW Research Report 2014:
http://www.pewinternet.org/2014/02/20/mapping-twitter-topic-networks-from-polarized-crowds-to-community-clusters/

8. TWITTER NETWORK PERSPECTIVES
Twitter Search API: Open space
8
Twitter Users API: Restricted space
▪ past 10 days / max. 18,000 tweets per query ▪ max. 3,200 tweets per user

9. 19. BUNDESTAG: TWITTER USEAGE
9
Party Color Seats
Twitter
users
Twitter users
per seat
CDU/CSU 246 131 53 %
SPD 153 123 80 %
AfD 92 85 92 %
FDP 80 72 90 %
Die Linke 69 60 87 %
B90/Die Grünen 67 64 96 %
no affiliation 2 2 100 %
All 709 537 76 %
https://www.bundestag.de/parlament/plenum/sitzverteilung_19wp

10. METHODOLOGY AND DATASETS
Methodology
1. Create a list with Twitter accounts
2. Download network data
3. Social network, content and visual analysis
4. Create several network subsets
Network Datasets
▪ 10-day time frames:
▪ Oct-12-2018, Dec-19-2018, Feb-20-2019
▪ 1-month time frames:
▪ June 2019, July 2019, August 2019
▪ All datasets available in NodeXL Graph Gallery
10https://nodexlgraphgallery.org/Pages/Default.aspx?search=%23nxlbundestag

11. NETWORK MAPS CREATED FROM ONE DATASET
11
1. Full Network Map
2. Internal Network Map
3. Party Network Map
4. Influencer Map
5. Party Interaction Map

12. 0
2
4
6
8
10
12
14
16
18
20
May
Sep
Dec
Mar
Jun
Sep
Dec
Mar
Jun
Sep
Feb
Jun
Sep
Dec
Mar
Jun
Sep
Dec
Mar
Jun
Sep
Dec
Mar
Jun
Sep
Feb
May
Aug
Nov
Feb
May
Sep
Dec
Mar
Jun
Sep
Dec
Mar
Jun
Oct
Jan
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
VERTEX METADATA: JOINED TWITTER DATE
12
Election
Sep-27-2009
Election
Sep-22-2013
Election
Sep-24-2017

13. Rank Name (MdB) Party Twitter Handle Followers Following Tweets Tweets/Day
1 Martin Schulz SPD martinschulz 698,054 3,261 4,629 1.2
2 Sahra Wagenknecht Die Linke swagenknecht 408,580 67 1,399 0.4
3 Christian Lindner FDP c_lindner 351,977 1,322 14,232 4.1
4 Gregor Gysi Die Linke gregorgysi 343,027 139 2,107 0.8
5 Heiko Maas SPD heikomaas 330,973 3,283 5,622 1.5
6 Sigmar Gabriel SPD sigmargabriel 271,903 585 3,333 1.2
7 Peter Altmaier CDU/CSU peteraltmaier 251,633 981 10,997 3.8
8 Peter Tauber CDU/CSU petertauber 188,935 1,219 22,929 6.0
9 Katrin Göring-Eckardt B90/Die Grünen goeringeckardt 148,895 1,214 14,605 5.6
10 Cem Özdemir B90/Die Grünen cem_oezdemir 148,870 1,052 9,538 2.5
MOST POPULAR BY NUMBER OF FOLLOWERS (SEP 2019)
13

14. Rank Name (MdB) Party Twitter Handle Followers Following Tweets Tweets/Day
1 Udo Hemmelgarn AfD udohemmelgarn 6,263 3,082 32,839 37.3
2 Johannes Kahrs SPD kahrs 23,428 4,370 129,823 34.4
3 Anke Domscheit-Berg Die Linke anked 31,271 2,314 88,147 22.1
4 Jörg Schneider AfD schneider_afd 3,855 1,127 18,028 14.6
5 Saskia Esken SPD eskensaskia 9,811 1,526 30,752 13.4
6 Stephan Brandner AfD stbrandner 6,572 689 16,038 12.8
7 Dieter Janecek B90/Die Grünen djanecek 10,041 1,522 40,942 11.0
8 Renate Künast B90/Die Grünen renatekuenast 47,044 1,534 25,231 11.0
9 Uwe Schummer CDU/CSU uweschummer 8,035 2,108 28,249 9.9
10 Sebastian Steineke CDU/CSU steinekecdu 5,469 1,663 28,374 9.2
TOP TWEETERS (BY DAY)
14

15. TWEET FREQUENCY – FEB-20-2019
15

16. TWEETING STRATEGIES BY PARTY
Network Edge Relationships:
▪ Tweet
→ No interaction
▪ Retweet
→ Amplification
▪ Replies to
→ Talking to someone
▪ Mentions
→ Talking about someone
16
Dataset: 1000 tweets per User / Sep 2, 2019 / 581,088 network edges

17. TWEETING STRATEGIES: @C_LINDNER (FDP)
17https://nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208710
Vertices : 884
Relationship:
Mentions : 1316
Replies to : 663
Tweet : 123
Retweet : 112

18. TWEETING STRATEGIES: @KAHRS (SPD)
18http://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208819
Vertices : 457
Relationship:
Replies to : 847
Mentions : 488
Tweet : 76
Retweet : 52

19. TWEETING STRATEGIES: @UDOHEMMELGARN (AFD)
19http://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208719
Vertices : 302
Relationship:
Retweet : 503
Mentions : 434
Replies to : 323
Tweet : 215

20. 20
Dataset:
Oct-12-2018
Internal Network

21. Dataset:
Oct-12-2018
Full Network
21

22. 22
Dataset:
Dec-19-2018
Internal Network

23. 23
Dataset:
Dec-19-2018
Full Network

24. 24
Dataset:
Feb-20-2019
Internal Network

25. 25
Dataset:
Feb-20-2019
Full Network

26. 26
Search term AfD
Feb-05-2019
Full Network

27. 27
Search terms
219a OR Abtreibung
Dec-19-2018
Full Network

28. 28
Dataset:
June 2019
Internal Network

29. 29
Dataset:
June 2019
Full Network

30. 30
Dataset:
July 2019
Internal Network

31. 31
Dataset:
July 2019
Full Network

32. 32
Dataset:
August 2019
Internal Network

33. 33
Dataset:
August 2019
Full Network

34. PARTY INTERACTION MAPS
34
Dataset 1: Oct-12-2018 Dataset 2: Dec-19-2018 Dataset 3: Feb-20-2019

35. PARTY INTERACTION MAPS
35
June 2019 July 2019 August 2019

36. GROUP NETWORK METRICS
36
Reciprocated Vertex Pair Ratio
Party Oct-12-2018 Dec-19-2018 Feb-20-2019 June 2019 July 2019 August 2019
AfD 14.7 % 14.1 % 8.5 % 9.2 % 9.1 % 13.2 %
B90/Die Grünen 23.8 % 21.7 % 21.0 % 24.4 % 23.0 % 17.6 %
CDU/CSU 5.8 % 11.1 % 13.0 % 15.3 % 14.1 % 18.4 %
Die Linke 14.8 % 9.9 % 15.8 % 22.6 % 6.2 % 9.4 %
FDP 18.8 % 14.2 % 15.8 % 20.3 % 20.4 % 17.0 %
SPD 11.4 % 8.8 % 10.8 % 10.7 % 14.6 % 16.1 %
Graph Density
Party Oct-12-2018 Dec-19-2018 Feb-20-2019 June 2019 July 2019 August 2019
AfD 5.8 % 5.8 % 5.7 % 4.5 % 4.2 % 3.8 %
B90/Die Grünen 8.1 % 9.6 % 8.7 % 13.3 % 7.5 % 7.9 %
CDU/CSU 2.0 % 2.6 % 2.5 % 4.0 % 4.0 % 3.3 %
Die Linke 7.5 % 4.6 % 6.0 % 11.0 % 5.9 % 4.8 %
FDP 4.3 % 4.8 % 5.1 % 7.3 % 7.3 % 7.6 %
SPD 2.7 % 2.9 % 3.2 % 4.7 % 2.9 % 3.5 %

37. GROUP SENTIMENT
37
Negative Word Percentage (%)
Party June 2019 July 2019 August 2019 June 2019 July 2019 August 2019
AfD 4.4 % 4.3 % 4.5 % 2.8 % 2.9 % 2.9 %
B90/Die Grünen 5.4 % 5.5 % 6.3 % 2.3 % 2.3 % 2.0 %
CDU/CSU 6.2 % 6.4 % 6.2 % 1.9 % 1.8 % 1.7 %
Die Linke 4.6 % 5.2 % 4.5 % 2.9 % 2.9 % 2.7 %
FDP 5.9 % 5.8 % 5.9 % 2.6 % 2.4 % 2.8 %
SPD 6.3 % 5.7 % 6.7 % 2.0 % 2.0 % 1.7 %
Positive Word Percentage (%)
Party Oct-12-2018 Dec-19-2018 Feb-20-2019 June 2019 July 2019 August 2019
AfD 5.1 % 4.5 % 5.0 % 4.4 % 4.3 % 4.5 %
B90/Die Grünen 5.4 % 6.2 % 5.9 % 5.4 % 5.5 % 6.3 %
CDU/CSU 6.6 % 5.8 % 6.6 % 6.2 % 6.4 % 6.2 %
Die Linke 5.2 % 4.8 % 4.9 % 4.6 % 5.2 % 4.5 %
FDP 5.1 % 5.0 % 6.2 % 5.9 % 5.8 % 5.9 %
SPD 6.4 % 6.5 % 6.8 % 6.3 % 5.7 % 6.7 %

38. INTERNAL NETWORK INFLUENCERS – OCT-12-2018
38
Rank Twitter Handle Party In-Degree
Out-
Degree
Betweenness
Centrality
Eigenvector
Centrality
1 c_lindner FDP 59 18 25740.531 0.018
2 kahrs SPD 25 32 15147.732 0.014
3 katarinabarley SPD 31 2 8493.334 0.010
4 mgrossebroemer CDU/CSU 21 16 8234.326 0.008
5 larsklingbeil SPD 25 9 8033.471 0.008
6 stbrandner AfD 10 26 7976.457 0.010
7 peteraltmaier CDU/CSU 32 1 7956.670 0.010
8 sven_kindler B90/Die Grünen 21 18 6686.655 0.011
9 udohemmelgarn AfD 7 36 6365.859 0.011
10 rbrinkhaus CDU/CSU 31 1 6311.749 0.005
11 dorisachelwilm Die Linke 12 21 5432.233 0.007
12 johannesvogel FDP 7 19 5137.178 0.005
13 rkiesewetter CDU/CSU 4 25 4855.530 0.005
14 alice_weidel AfD 26 1 4689.686 0.009
15 dietmarbartsch Die Linke 14 6 4584.795 0.005
Betweenness
Centrality
Eigenvector
Centrality

39. FULL NETWORK INFLUENCERS – OCT-12-2018
39
Rank Twitter Handle Party In-Degree
Out-
Degree
Betweenness
Centrality
Eigenvector
Centrality
1 kahrs SPD 25 515 3954118.61 0.009
2 eskensaskia SPD 7 240 1619942.49 0.003
3 anked Die Linke 12 176 1219345.88 0.002
4 udohemmelgarn AfD 7 296 1169617.79 0.009
5 djanecek B90/Die Grünen 13 142 827590.17 0.002
6 renatekuenast B90/Die Grünen 11 145 805045.48 0.003
7 c_lindner FDP 61 74 735614.75 0.005
8 rkiesewetter CDU/CSU 4 135 726132.40 0.002
9 schneider_afd AfD 7 234 658452.66 0.008
10 drandreasnick CDU/CSU 4 132 633946.34 0.002
11 lieblingxhain B90/Die Grünen 7 127 604051.51 0.002
12 frankschwabe SPD 8 101 578840.63 0.002
13 lisapaus B90/Die Grünen 7 128 542398.86 0.002
14 olliluksic FDP 11 130 507873.19 0.003
15 miro_spd SPD 10 80 497234.59 0.002
Betweenness
Centrality
Eigenvector
Centrality

40. INTERNAL NETWORK INFLUENCERS – AUGUST 2019
40
Rank Twitter Handle Party In-Degree
Out-
Degree
Betweenness
Centrality
Eigenvector
Centrality
1 peteraltmaier CDU/CSU 52 6 13315.978 0.017
2 heikomaas SPD 45 1 11125.246 0.009
3 udohemmelgarn AfD 5 40 10782.212 0.007
4 c_lindner FDP 49 19 10681.718 0.018
5 kahrs SPD 32 44 9697.306 0.014
6 abaerbock B90/Die Grünen 39 10 8240.197 0.012
7 katjakipping Die Linke 28 10 8047.848 0.005
8 paulziemiak CDU/CSU 38 8 7991.705 0.010
9 baerbelbas SPD 12 31 7364.514 0.006
10 olliluksic FDP 17 39 6588.962 0.016
11 tschipanski CDU/CSU 9 36 6326.758 0.010
12 cem_oezdemir B90/Die Grünen 34 16 6053.835 0.011
13 andischeuer CDU/CSU 38 3 6043.266 0.013
14 eskensaskia SPD 14 35 5943.170 0.012
15 th_sattelberger FDP 10 33 5113.055 0.011
Betweenness
Centrality
Eigenvector
Centrality

41. EXTERNAL NETWORK INFLUENCERS – AUGUST 2019
41
Rank Twitter Handle Category Indegree
Betweenness
Centrality
1 welt News/Media 115 1988243.028
2 spdde Party 144 1970892.003
3 spdbt Party 128 1915653.728
4 die_gruenen Party 109 1723811.824
5 akk Politician 124 1556913.568
6 dielinke Party 85 1466186.103
7 cdu Party 114 1266732.958
8 spiegelonline News/Media 81 1258353.580
9 cducsubt Party 100 1202502.309
10 gruenebundestag Party 84 1081743.754
11 olafscholz Politician 101 1039363.609
12 tagesspiegel News/Media 64 851519.692
13 linksfraktion Party 64 779983.049
14 tagesschau News/Media 65 754499.804
15 faznet News/Media 73 748936.554
16 fdp Party 78 661145.616
17 afd Party 63 639764.922
18 sz News/Media 60 614854.062
19 mpkretschmer Politician 75 596165.392
20 tazgezwitscher News/Media 49 548448.097
The most influential Twitter users
outside the Bundestag are related
national party accounts, large news
media outlets and politicians without a
seat in the Bundestag.

42. PARTY NETWORK: CDU/CSU
42http://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208616

43. PARTY NETWORK: SPD
43http://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208678

44. PARTY NETWORK: B90/DIE GRÜNEN
44https://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208679

45. PARTY NETWORK: FDP
45https://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208695

46. PARTY NETWORK: DIE LINKE
46https://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208696

47. PARTY NETWORK: AFD
47https://nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208587

48. Summary
▪ Clusters
▪ Changing coalition clusters
▪ Ambiguity between internal and full network data
▪ Party Interaction
▪ A lot of talk about the AfD, very little talk with the AfD
▪ B90/Die Grünen is most unified party
▪ Influencers
▪ Internal: High ranking party officials and top tweeters
▪ External: Media outlets, Party accounts, (regional) politicians
Further research
▪ We need more network maps!
▪ Compare to other platforms
▪ Compare to other parliaments
48

49. MEMBERS OF THE EUROPEAN PARLIAMENT (USER LIST)
list:Europarl_EN/all-meps-on-twitter, 20 May 2019: https://nodexlgraphgallery.org/Pages/Graph.aspx?graphID=197547 49

50. LANDTAGSWAHLEN BRANDENBURG (SEARCH API)
ltwbb OR ltwbb19 until:2019-08-31: http://www.nodexlgraphgallery.org/Pages/Graph.aspx?graphID=208480 50

51. USER ACCOUNT INVESTIGATION
51https://www.vanityfair.com/news/2017/06/trump-says-he-wont-stop-tweeting
https://theguardiansofdemocracy.com/trump-schedules-last-minute-meetings-based-whatever-saw-fox-friends-report/

52. USER ACCOUNT INVESTIGATION 52
Based on Twitter users followed by @realdonaldtrump
https://nodexlgraphgallery.org/Pages/Graph.aspx?graphID=174922

53. Questions? Please send an email to harald@smrfoundation.org
All network maps related to the Bundestag can be found here:
https://nodexlgraphgallery.org/Pages/Default.aspx?search=%23nxlbundestag
Please visit the following website for more information:
https://www.smrfoundation.org/2018/09/14/research-project-mapping-political-networks/
This case study is part of the research project
Mapping Political Networks
at the Social Media Research Foundation

54. KEY FEATURES OF NODEXL PRO
54
Network Analysis Content AnalysisData Import Data ExportVisualization
Network Overview
Network size and composition
Graph density, modularity
Group Analysis
Group by cluster
e.g. Clauset-Newman-Moore
Group metrics
Vertex metrics
Degree/In-/OutDegree
Betweenness/Closeness/
Eigenvector/ PageRank
Path Analysis
Text Analysis
Words and word pairs from
Tweets, Posts, Replies, …
Sentiment Analysis
Positive/Negative Sentiment
Your list of Keywords
Top Content Summary
By entire network / by group
Top hashtags, URLs, domains
Top words and word pairs
Time Series Analysis
By minute/hour/day/…
By hashtag/word/language/…
Data formats
Excel/UCINET/GraphML/
Pajek/GEFX/GDF
Social media data
3rd Party importers
Data formats
Excel/UCINET/GraphML/
Pajek/GEFX/GDF
Publish to the web
NodeXL Graph Gallery
Export to Powerpoint
Export to Polinode
Customize
Shape, size, color, label of
vertices, edges and groups
Autofill Columns
Graph Layout
Various layout algorithms
e.g. Harel-Koren Fast
Multiscale
Group-In-a-Box Layout
Treemap
Force-directed
Packed rectangles
Automate Key Features with NodeXL Data Recipes