Political stability and the fragmentation of online publics in multilingual states

Francesco Bailo The University of Sydney

Internet, Politics, and Policy conference Oxford Internet Institute, University of Oxford

23 September 2016 @FrBailo Factionalisation, group grievance and political stability

CHE CHE 10.0 CHE CHECHECHECHECHECHE 1 BELBEL 1 BELBEL BIHBIH BEL BELBELBEL BELBELBELBELBEL BIHBIHBIHBIHBIH BEL BEL BIHBIH UKRUKRUKRUKRUKRUKRUKR UKR UKR 7.5 UKR 0 UKRUKRUKRBIH 0 UKRUKRUKRUKRBIH UKRBIH UKRBIHBIH BIHBIHBIH BIHBIHBIHBIH UKR BIH UKR BIHBIH BIH −1 −1 5.0 BELBELBEL political stability political stability −2 UKR −2 UKR BEL factionalized elites BELBEL 2.5 BELBEL −3 −3 CHECHECHECHECHECHECHE

2.5 5.0 7.5 10.0 2.5 5.0 7.5 10.0 2.5 5.0 7.5 10.0 factionalised elites (R2 = 0.57) group grievance (R2 = 0.659) group grievance (R2 = 0.701)

Source: Worldwide governance indicators, World Bank (2014) and Fragile States Index, The Fund for Peace (2014)

[email protected] 2/20 @FrBailo Segregation along ethnolinguistic lines and quality of government

According to Alesina and Zhuravskaya (2011)

I ‘higher ethnic and linguistic segregation is associated with significantly lower government quality’ and

I ‘generalized trust is lower in more segregated countries and higher in countries with good government’

[email protected] 3/20 @FrBailo Political and ethnolinguistic fragmentation

Political dimension As position on a 1 to 5 (left to right) continuous scale

For parties based on Wikipedia data (Political position) mapped to continuous scale For Facebook users based on the average value of their likes (each likes assume the value of the party to which the page is linked)

Ethnolinguistic dimension As relative frequency of use of a particular language

For parties based on the proportion of comments in a specific language For Facebook users based on the proportion of postings (posts and comment) in a specific language

Fragmentation As ratio in the volume of interactions between and within politically and ethnolinguistically homogeneous groups

[email protected] 4/20 @FrBailo Data

4 countries Belgium (BEL), Bosnia and Herzegovina (BIH), (UKR), Switzerland (CHE) 93 parties running in the general elections held in 2014 (BEL, BIH, UKR) and 2015 (CHE) 1423 Facebook pages each linked to a party Identification of Facebook pages was a two-step process

1. Identification of Facebook pages by scraping parties’ homepages and through Facebook searches; 2. Collection of Facebook pages liked by pages identified in step 1

[email protected] 5/20 @FrBailo Data

country pages posts comments likes profiles BEL 864 44,739 116,368 1,254,577 203,263 BIH 206 19,219 42,001 1,011,952 113,231 CHE 214 9,934 25,114 241,327 72,100 UKR 137 16,843 69,325 968,112 121,446

[email protected] 6/20 @FrBailo Country demographics Belgium Ukraine

Dutch Ukrainian French Russian

Switzerland Bosnia and Herzegovina German French Italian

Bosnian Serbian [email protected] 7/20

Source: Language distribution based on last Census available for each country with the exception of Belgium where language use is not a Census question @FrBailo Language use in parties’ pages BEL (Concetration = 0.94) BIH (Concetration = 0.69) Vox Populi Belgica Democratic People's Union Parti libertarien Nouvelle Wallonie Alternative Serb Democratic Party Mouvement de Gauche Socialist Party Islam Social Democratic Party of Gauches communes UnionBosnia for aand Better Herzegovina Future of Faire place Nette BiH DemocraticÉgalitaires Federalist Bosnian−HerzegovinianOur Party People'sIndependent Party Patriotic Party−Sefer LA DROITE DemocraticHalilovi.. Front VALEURS LIBÉRALES CITOYENNES Croatian Democratic Union Mouvement Réformateur Party for Bosnia1990 and Centre démocrate humaniste Ecolo People's Party forHerzegovina Work and Parti Socialiste Croatian DemocraticBetterment Union of Rassemblement Wallonie France Workers' Party of Belgium LabourBosnia Party and of Herzegovina Bosnia and Belgische Unie ... Union Belge Herzegovina Party of Democratic Action Open Vlaamse LiberalenPirate Party en Social Democratic Union of Christen−DemocratischDemocraten en Bosnia and Herzegovina VlaamsSD&P HSP−AS BiH New Flemish Alliance Party of Democratic Progress Socialistische Partij Anders Bosnian Party Vlaams Belang Groen Party of Democratic Activity 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00

Dutch French Bosnian Serbian

CHE (Concetration = 0.76) UKR (Concetration = 0.6)

Solidarity Communist Party of Ukraine Ticino League 5.10 FDP.The Liberals Christian Democratic People's Liberal Party of Ukraine Party of Switzerland Ukraine is United Green Party of Switzerland Strength and Honour Conservative Democratic Party People's Front of Switzerland National Democratic Party of Ukraine Swiss People's Party Ukraine of the Future Social Democratic Party of United Country Switzerland All−Ukrainian Union Evangelical People's Party of "Fatherland" Switzerland Green Liberal Party of People's Power Switzerland Bloc of Ukrainian Left Forces Federal Democratic Union of Petro Poroshenko Bloc Switzerland Congress of"Solidarity" Ukrainian Nationalists Alternative Left Svoboda 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00

[email protected] French German Italian Russian Ukrainian 8/20 @FrBailo Network analysis Two types of network: Network of parties connections are drawn based on the number of comments received by each pair of party from the same user Network of users connections represent the number of direct replies exchanged among users Two types of Exponential Random Graph Models (ERGM), estimating the significance of node’s characteristics (political position and language) in edge formation: Temporal Exponential Random Graph Model (TERGM) Longitudinal dynamic network, same number of nodes (parties), changing edges (cross-party posting) Exponential Random Graph Models Series of networks, one network for each day in the period, changing nodes (users), changing edges (direct replies) [email protected] 9/20 @FrBailo Network of parties (cross-posting): Belgium

Gauches Gauches communes communes

Rassemblement Democratic Rassemblement Democratic Wallonie Federalist Wallonie Federalist France Independent France Independent

Ecolo Ecolo Workers' Workers' Centre Centre People's Party People's Party démocrate démocrate Party of Party of humaniste humaniste Belgium Christen−Democratisch Belgium Christen−Democratisch Mouvement en Mouvement en Réformateur Vlaams Réformateur Vlaams Parti Pirate Parti Pirate Socialiste Belgische Party Socialiste Belgische Party Groen Groen LA Unie ... Open LA Unie ... Open DROITE Union Vlaamse DROITE Union Vlaamse Belge Liberalen Socialistische Belge Liberalen Socialistische Parti en Partij Parti en Partij New New libertarien Democraten Anders libertarien Democraten Anders Vlaams Flemish Vlaams Flemish Belang Alliance Belang Alliance

Nouvelle Nouvelle Dutch Wallonie Wallonie French Alternative VALEURS Alternative VALEURS LIBÉRALES SD&P LIBÉRALES SD&P CITOYENNES CITOYENNES

Faire Faire place place Nette Nette

Left Center Right Vox Vox Populi Populi Belgica Belgica

[email protected] 10/20 @FrBailo Network of parties (direct reply): Ukraine

Left Center Right

[email protected] 11/20 @FrBailo TERGM on networks of parties

edges edges edges

Political position (absdiff) Political position (absdiff) Political position (absdiff)

Dutch (absdiff, %) Dutch (absdiff, %) Dutch (absdiff, %)

−2.0 −1.5 −1.0 −0.5 0.0 −3.0 −2.0 −1.0 0.0 −2.0 −1.5 −1.0 −0.5 0.0

Bars denote CIs. Bars denote CIs. Bars denote CIs.

(a) Before election: Belgium (b) After election: Belgium (c) Entire period: Belgium

edges edges edges

Political position (absdiff) Political position (absdiff) Political position (absdiff)

German (absdiff, %) German (absdiff, %) German (absdiff, %)

−4 −3 −2 −1 0 −15 −10 −5 0 −4 −3 −2 −1 0

Bars denote CIs. Bars denote CIs. Bars denote CIs.

(a) Before election: Switzerland (b) After election: Switzerland (c) Entire period: Switzerland

[email protected] 12/20 @FrBailo TERGM on networks of parties

edges edges edges

Political position (absdiff) Political position (absdiff) Political position (absdiff)

Bosnian (absdiff, %) Bosnian (absdiff, %) Bosnian (absdiff, %)

−4 −3 −2 −1 0 1 −6 −4 −2 0 2 −4 −3 −2 −1 0 1

Bars denote CIs. Bars denote CIs. Bars denote CIs.

(a) Before election: Bosnia (b) After election: Bosnia (c) Entire period: Bosnia

edges edges edges

Political position (absdiff) Political position (absdiff) Political position (absdiff)

Ukrainian (absdiff, %) Ukrainian (absdiff, %) Ukrainian (absdiff, %)

−2.5 −2.0 −1.5 −1.0 −0.5 0.0 −3.0 −2.0 −1.0 0.0 −2.5 −1.5 −0.5 0.0

Bars denote CIs. Bars denote CIs. Bars denote CIs.

(a) Before election: Ukraine (b) After election: Ukraine (c) Entire period: Ukraine [email protected] 13/20 @FrBailo Segregation of linguistically homogeneous areas in Belgium

(a) FR ↔ FR: (15,556 edges) (b) NL ↔ NL: (1,622 edges) (c) NL ↔ FR: (297 edges) Figure: Cross-party users’ posting within and among linguistic regions (Brussels excluded)

[email protected] 14/20 @FrBailo ERGMs on networks of users (direct reply)

0 −1 −2 (coef)

Political −3

0 −2 −4 (coef) Dutch −6 −8

2000 1500 1000 500 Vertices 0 Belgium

0 −2

(coef) −4 Political

0 −5

(coef) −10 German

500 400 300 200

Vertices 100 0 Switzerland

[email protected] 15/20 @FrBailo ERGM on users’ direct reply network

0 −1 −2 (coef)

Political −3

2 1 0

(coef) −1

Bosnian −2

600 400 200 Vertices 0 Bosnia and Herzegovina

0 −1

(coef) −2 Political

0 −1 −2 (coef) −3 Ukrainian

600 400 200 Vertices Ukraine

[email protected] 16/20 @FrBailo Ukrainian crisis 2014

1000

russian

ukrainian

comments/day 0

1000

Dec 2013 Jan 2014 Feb 2014 Mar 2014 Apr 2014 May 2014 Jun 2014 Figure: Fequency of Facebook comments around the 2014 Ukrainian crisis

[email protected] 17/20 @FrBailo Ukrainian crisis 2014

0

−1 (coef) Ukrainian

−2

1.00

0.75

0.50

Language 0.25 significance

0.00

2000

1500

1000 Vertices 500

0

Figure: ERGM on users’ direct reply network: 2014 Ukrainian crisis [email protected] 18/20 @FrBailo Conclusions

I The distance in political position is a significant and (as expected) negative predictor for engagement among users, independently from the overall stability of the country.

I The distance in language background is a significant and negative predictor for engagement in more stable countries but less so in fragile countries, which experience more inter-ethnolinguistic engagement.

I Group grievance is thus not necessarily associated with less exchanges among groups but it might result in more exchanges.

I After the military escalation of the 2014 Ukrainian crisis, exchanged across ethnoliguistic lines increased.

[email protected] 19/20 Political stability and the fragmentation of online publics in multilingual states

Francesco Bailo The University of Sydney

Internet, Politics, and Policy conference Oxford Internet Institute, University of Oxford

23 September 2016