On Swiss multi-party political system andpolarization in Politnetz

Adiya Abisheva

with F. Schweitzer and U. Serdult and David GarciaChair of Systems designwww.sg.ethz.ch

Outline

1 The Politnetz Community

2 Network Polarization in Politnetz

3 Social Networks of Parties

Chair of Systems design SKIN3 workshopwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 2 / 17

Chair of Systems Design at ETH ZurichMain Research Areas

Economic Networks & Social Organizationse.g. ownership networks, R&D networks, financial networks, ...e.g. online communities, OSS projects, animal societies, ...

Denmark

Portugal

United Kingdom

Belgium

Netherlands

France

Spain

Ireland

Finland

Greece

Sweden Germany

Methodological Approach: Data Driven Modelingeconomic databases: Bloomberg, patent and ownership databasesdigital traces: user interaction, OSN, activity volumes

Chair of Systems design SKIN3 workshopwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 3 / 17

The Politnetz community

Online politics participatory communityRepresentation of the parliament, with users asaudienceOpen discussions between politicians and usersNetwork of support and party affiliation

Dataset

Jan 2011-Jan 2014More than 3200 politicians, 12000 usersFocus on politician network:

15000 support links, 11000 likes, 38000 commentsMultiplex network with three layers

Chair of Systems design SKIN3 workshop The Politnetz Communitywww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 4 / 17

Social interaction links in Politnetz

Support links: directedunweighted network

Likes and Comments: directedweighted networks

Chair of Systems design SKIN3 workshop The Politnetz Communitywww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 5 / 17

Polarization: definition

Esteban & Ray (1994)

Suppose that the population is grouped into significantly-sized“clusters”, such that each cluster is very “similar” in terms of theattributes of its members, but different clusters have memberswith very “dissimilar” attributes. In that case, society is“polarized”.

Flache & Macy (2011)

Polarization of opinions is characterized by a division of thepopulation into a smaller number of fractions with high internalconsensus and sharp disagreement between them.

Chair of Systems design SKIN3 workshop Network Polarization in Politnetzwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 6 / 17

Network polarization: modularity

Q-Modularity

Probability that an edgeconnects two nodes in thesame community c, comparedto random graph

Q =1

2m

∑i,j

[Ai,j −

ki · kj

2m

]·δ(ci , cj)

Two separations:1 ci = party affiliation of politician i : Modularity of parties Qp

2 Communities found with algorithms: Maximal modularity Qmax

Chair of Systems design SKIN3 workshop Network Polarization in Politnetzwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 7 / 17

Network polarization in Politnetz

SupportsQp = 0.677 (10)Qmax = 0.742 (12)

LikesQp = 0.299 (10)Qmax = 0.448 (14)

CommentsQp = −0.07 (10)Qmax = 0.354 (15)

Qp - modularity score of networks with party labels,Qmax - maximum modularity score without party tagsColor corresponds to politicians’ party

Chair of Systems design SKIN3 workshop Network Polarization in Politnetzwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 8 / 17

Modularities of Politnetz: per-party aggregation

FDP Pirate

GLP

SP Christian

SVP/EDU

AL

Grune

BDP

highly modularnetworksupport link actsas a signal forparty membership

Grune

Christian

SVP/EDU

BDP

SP

Pirate

GLP

FDP

AL

likes links staymostly within partynumber ofcross-party linksare higher than insupport links

Pirate

Grune

GLP

Christian

BDP

AL

FDPSVP/EDU

SP

comments revealtwo attractors ofconversation –major parties inSwiss politics: SPand SVP

Chair of Systems design SKIN3 workshop Network Polarization in Politnetzwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 9 / 17

Modularity of likes peaks at elections

Date

0.20

0.30

0.40

0.50

Net

wor

kP

olar

izat

ion

Jan 2011 Jul 2011 Jan 2012 Jul 2012 Jan 2013

Modularity of likes network over time windows of 2 monthsBefore elections, politicians avoid liking posts from other parties

Chair of Systems design SKIN3 workshop Network Polarization in Politnetzwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 10 / 17

Network polarization: users

Qp = −0.001 (9)Qmax = 0.519 (22)

Two clusters detected in the likesnetwork without party labelsLow modularity of party-labeledlikes network shows lack of politicalmobilization among users

Likes network of usersUsers are not aware of party labelsand like posts based on contentrather than party affiliation unlikeamong politicians

Qp - modularity score of networks with party labels,Qmax - maximum modularity score without party tags

Chair of Systems design SKIN3 workshop Network Polarization in Politnetzwww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 11 / 17

Left vs. right party communities structure

Hypothesis

Compared to left-aligned communities, right-aligned communitiesare less influenced by content created by political campaigns, buthave stronger social interaction regarding political topics.

Conover & Menczer (2011)

In the US, online social networks of right politicians are moredense and clustered than the networks of left politicians

Chair of Systems design SKIN3 workshop Social Networks of Partieswww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 12 / 17

Most frequent YouTube content shared bypolitical user groups in Twitter

RANK LEFT RIGHT APOLITICAL1 Minecraft Barack Obama Minecraft2 Call of Duty II Alex Jones Call of Duty II

. . . . . . . . . . . .6 Film Ron Paul Hip hop music

. . . . . . . . . . . .13 Album Police Call of Duty14 Call Mitt Romney Video blog15 Song Russia Today Episode. . . . . . . . . . . .25 Heavy metal Boston NBA. . . . . . . . . . . .27 Episode US NSA Super Junior. . . . . . . . . . . .29 Justin Bieber Bomb Pokemon30 Barack Obama Train Music

Right-aligned users share more politically charged content.Left-leaning users are similar to apolitical.

Chair of Systems design SKIN3 workshop Social Networks of Partieswww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 13 / 17

Social networks of parties

Conover & Menczer (2011)

In the US, online social networks of right politicians are moredense and clustered than the networks of left politicians

We measured three aspects of the social network of a party1 In-degree centralization: tendency towards a star-like structure,

inequality of individual support in the party2 Average path length: information efficiency, closeness of the

community3 Maximum k-core: presence of a rich-club, resilient core of

leaders

Chair of Systems design SKIN3 workshop Social Networks of Partieswww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 14 / 17

Party political position and social structure

averagepath

lengthmaxim

umk-core

in-degreecentralization

2.8

3.2

3.6

44.4

611

1621

2631

00.1

0.2

3.5 4 4.5 5 5.5 6 6.5

Left/Right position

SP

SP

SP

GLP

GLP

GLP

SVP

SVP

SVP

FDP

FDP

FDP

Christian

Christian

Christian

Grune

Grune

Grune

We focus on the support network of parties withmore than 150 politicians in PolitnetzParty position coded in surveysLeft parties have higher maximum k-core valuesRight parties have longer path lengthsGreen parties have higher centralization

Opposite of Conover & Menczer in Twitter USRight parties are not more closely connectedEffect of status quo rather than political alignment

Chair of Systems design SKIN3 workshop Social Networks of Partieswww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 15 / 17

Summary

We analyzed network polarization (modularity) in PolitnetzSupport network stable and very polarizedPolarization in likes network increases when elections are close

Parties show different network structures

Scientific importance and impact of research1 Contribution towards local research communities of direct democracy

and e-voting,2 General approach can be applied towards international case studies,3 Assessment of impact of political campaigns, effectivity of mobilization,

and the impact of decision policy making.

Chair of Systems design SKIN3 workshop Summarywww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 16 / 17

Future research questions

Detecting alignment on non-aligned political partiesBipolarization of the networkMatching answers of politicians to smart-vote dataPredict configuration of Swiss government and party coalitionPopularity of politicians/career path of politiciansAnalyze users opinions and political activity from onlinenewspapers (20min.ch)

Chair of Systems design SKIN3 workshop Summarywww.sg.ethz.ch Adiya Abisheva SKIN3 workshop Budapest, Hungary 22 May, 2014 17 / 17