1

Does Facebook Increase Political Participation? Evidence from a Field Experiment

Yannis Theocharis1

Mannheim Centre for European Social Research

Will Lowe2 Princeton University

This is the accepted version of the manuscript published by Taylor & Francis in Information, Communication & Society on 17 December. The published manuscript is available online and can

be downloaded at http://www.tandfonline.com/doi/full/10.1080/1369118X.2015.1119871

Abstract Duringthelastdecade,muchofpoliticalbehaviourresearchhascometobeconcerned with the impact of the internet, and more recently socialnetworking sites like Facebook, on political and civic participation.Althoughexistingresearchgenerallyfindsamodestlypositiverelationshipbetweensocialmediauseandofflineandonlineparticipation,themajorityofcontributionsrelyoncross-sectionaldatasothecausal impactofsocialmediauseremainsunclear.ThepresentstudyexamineshowFacebookuseinfluences reported political participation using an experiment. Werecruited young Greek participants without a Facebook account andrandomlyassignedasubsettocreateandmaintainaFacebookaccountforayear.InthispaperweexaminetheeffectofhavingaFacebookaccountondiversemodesofonlineandofflineparticipationaftersixmonths.WefindthatmaintainingaFacebookaccounthadclearlynegativeconsequencesonreportsofofflineandonlineformsofpoliticalandcivicparticipation.

politicalparticipation;experiments;politicalbehaviour;Facebook;socialmedia

1 yannis.theocharis@uni-mannheim.de 2 wlowe@princeton.edu

2

Introduction

During the lastdecade,muchpoliticalbehaviour literaturehasconsidered the impactof

newcommunication technologies,especiallysocialmediasuchasFacebookandTwitter,

onpoliticalparticipation.Manyresearchershavehopedthatasworldwideinternetaccess

broadens these technologies will help re-engage citizens in the democratic processes

(Dalton, 2008; Hay, 2007; Norris, 2002), for example by implementing more direct e-

democracythroughonlineplebiscites to therecreationofavirtualpublicsphere though

online consultations (Coleman and Blumler, 2009). New internet technologies will

complementsocialcapital(Wellmanetal.,2001),affectculturalvalues(Norris,2001),and

enhancecitizenengagementwithpolitics(Barber,1999;Etzioni,1993).

Inparticular, theheavilydigitalmedia-basedpresidential campaignofBarackObama in

2008waspromotedasaremedynotonlyfordecliningengagementwithelectoralpolitics

but also for the participatory decline that characterises direct action groups (Earl and

Kimport, 2011). Social media use would, it was argued, lower transaction costs by

removing obstacles to participation, generate new more efficient forms of bottom-up

organisationandcoordination,andcreatenewformsofparticipationtocomplementthe

existing repertoires (Bimber etal., 2005; Van Laer and Van Aelst, 2010; Bennett and

Segerberg, 2013). This argument gained further support from the importantmobilising

rolethatsocialmediaplayedintheArabSpringandOccupyevents(HowardandHussain,

2013;Lynch,2011;Shirky,2011;TufekciandWilson,2012;Theocharisetal.2015).

Much research has been devoted to investigating the relationship between internet use

and political participation both before and since the arrival of social media. In general

results have been positive; a meta-analysis examining internet’s effects on political

participationhasfoundweakormodestlypositiverelationshipsbetweeninternetuseand

offline political participation under a range of operationalisations in cross-sectional

observational data (Boulianne, 2009). More recently, scholars have placed greater

emphasisonthemobilisingroleofpopularsocialmediaplatforms,suchasFacebook,on

offline political participation, while a number of studies have tried to tackle causal

questions through theuseof panel data (Dimitrova etal., 2014;Holt etal., 2013;Kahne

etal.,2013).Theresultshavegenerallybeenmixedwithsomestudiesfindingnoeffects,

otherstudiesfindingspositiveeffectsandstillothersfindingeffectsonlyoncivicbutnot

onpoliticalparticipation.Asarecentmeta-analysisbyBoulianneconcludes,"themetadata

demonstrate a positive relationship between social media use and participation

3

(Boulianne, 2015, p.524). More than 80% of coefficients are positive". "However", the

author notes, "questions remain about whether the relationship is causal and

transformative. Only half of the coefficients were statistically significant. Studies using

panel data are less likely to report positive and statistically significant coefficients

between social media use and participation, compared to cross-sectional

surveys"(Boulianne, 2015, p.524). Importantly, Boulianne also notes that "In terms of

causal effects, few studies employ panel data and none of the studies employ an

experimentaldesign,whichwouldhelpestablishcausality"(Boulianne,2015,p.534).

Aimingtoprovidefirmerevidenceaboutsocialmedia’seffectsonparticipation,thisstudy

adopts an experimental approach. It tests the relationship between Facebook use and

mobilisation in a survey setting in which possession and maintenance of a Facebook

accountwas randomly assigned, and examined its effect on reported online and offline

political participation after six months. Taking into consideration recent research

suggestingthat,inadditiontoownershipofasocialmediaaccountthetypeofsocialmedia

use may decisively affect subsequent online or offline participation (Gil de Zuniga and

Valenzuela, 2010; Valenzuela, 2012; Bachmann and deZuniga, 2013; Holt etal., 2013;

Towner,2013;TangandLee,2013;Xenosetal.,2014),wealsopresentfindingsshowing

how intensity of Facebook use, embededness in online networks, and civic use of

Facebook affected online and offline participation patterns of the treatment group

participants. From the literature we expected that Facebook use would increase both

onlineandofflineengagement.Wealsoexpected that intensityofFacebookuse, greater

numberoffriends,beingexposedtomobilisinginformationandinteractingwithpolitical

contentpostedby friendsonFacebookwouldbepositively related topolitical and civic

participation. However, we find that, in this population, Facebook use leads to a

substantialdecline inreportedcivicandpoliticalparticipation.Ourpost-hocanalysison

the treatment group shows a similar story with the most common (Facebook-related)

antecedents of participation being negatively correlatedwith all forms examined in the

study. As the studywas conducted in Greece during one of themost turbulent periods

(socially, politically and economically) in its recent history (and thus in a context very

differenttotheonemostotherrelevantstudieshavebeencarriedout),wearguethatthe

politicalcontextmayhaveplayedacriticalroleinthewayFacebookaffectedparticipatory

patterns. To this end, we offer possible explanations, including that, rather than being

usedasatoolformobilisation,Facebookwasmostlyusedasatoolforentertainmentand

distractionfromthediresituationinwhichthevastmajorityofGreeksfoundthemselves

4

in,orasatoolforfurtherdisillusionmentfrompoliticspromptingdisconnectionwiththe

political arena. In what follows we review the literature on social media’s impact on

political participation, outline our expectations and state our hypotheses. We then

elaborateonourresearchdesign,datacollectionandmeasures,andpresentourresults.

Thearticleconcludeswithadiscussionofthefindingsandtheirimplications.

PoliticalParticipation,theInternetandSocialMedia

Much research on political participation in the last decade has shown an increasing

detachment of citizens of Western countries from politics (Putnam, 2000; Dalton and

Wattenberg, 2000). Amidst pessimistic approaches about the future of representative

democracy,theroleoftheinternetinaffectingparticipatorybehaviourhasbecomecentral

in political participation debates. The literature reports mostly positive relationships

between internet use and different forms of civic and political participation (Bimber,

2001; Mossberger etal., 2008; Shah etal., 2005; Xenos and Moy, 2007; Bakker and

deVreese,2011;QuintelierandVissers,2008;Valenzuela,2012;Valenzuelaetal.,2009).

Indicatively, the internet’s positive influence on political participation has been

demonstratedbyameta-analysiswhich,examiningatotalof38studieswith166effects,

found that positive effects significantly outnumbered negative ones (Boulianne, 2009).

However, scholars have been reluctant to draw firm conclusions about the internet’s

mobilisingpotential.Notonlyhavemanyoftherelationshipsthesestudieshaverevealed

beentooweakanddrawnfromconveniencesamples,theyhavealsobeenoverwhelmingly

tackled with cross-sectional survey data which render the causal mechanisms unclear.

Moreovermost of these studies have demonstrated similar participatory inequalities as

those already in place offline (Schlozman etal., 2012; Xenos andMoy, 2007), providing

support to the so-called reinforcement theories that saw the internet as a tool thatwill

mainly empower those already politically engaged (Norris, 2001). Still, the fact that

researchexaminingthe internet’s impactonpoliticalparticipation in the less-interactive

pre-Web2.0erahasproducedagenerallyhopefulpictureaboutthemobilisingeffectsof

general internet use, strengthened hopes about the respective potential of the more

interactivesocialmediaplatforms.

The riseofWeb2.0and specifically socialnetworking sitesbroughtmoreopportunities

formobilisationthroughtheircapacity tosupport thewidespreaddiffusionof(political)

information across diverse networks of individuals. In addition, the radical lowering of

5

opportunity costs for engaging with politics led to the widespread adoption of diverse

onlinerepertoires(Smith,2013).Thishasledtovariousconceptualdifficultiesasthereis

noconsensuswithregardstowhethersuchactscanbelegitimatelyconsideredasmodes

ofpoliticalparticipation.Oneofthemostimportantissuesintheliteraturehasthusbeen

theinvestigationofthemodeofonlineorofflineparticipationinfluencedbytheinternet.

Oser and colleagues (Oser etal., 2013) have found that online activism is itself a

distinctive typeofpoliticalparticipation Inadifferentstudy,Cantijoch(Cantijoch,2012)

found that internet usemay not have an impact on allmodes of political participation,

withuseofinternetmorefrequentlyleadingtoanincreasedlikelihoodtobecomeactivein

politicalprotest.Otherresearchershavealsofoundthattherearedistinctmodesofonline

participationthatarecomparabletoclassictraditionalofflinerepertoires,butalsoother,

more expressive social media-based and online-only activities that lack an offline

equivalent(GibsonandCantijoch,2013;HirzallaandVanZoonen,2010).

Research on social media’s effects has generally moved beyond the examination of the

internet as a uniform platform which can either mobilise or reinforce participation

(Norris,2001),placingmoreemphasis,andindeeddiscovering,thatthetypeofuse(Tang

andLee,2013;Towner,2013;GildeZunigaandValenzuela,2010;Bondetal.,2012),and

eventheplatform(BachmannanddeZuniga,2013),mattersforpoliticalmobilisation.In

anexperimentwith61millionpeople,Bondandcolleagues(Bondetal.,2012)foundthat,

on Facebook, mobilisation messages sent in social contexts (posted, for example, by

friendsor friendsof friends) can influencepolitical self-expression, information seeking

andevenvoting.Rojas andPuig-i-Abril found that informationalusesof ICTs (including

socialmedia) are significantly related to expressive participation onlinewhich “in turn

results to a host of traditional or offline civic and political participatory behaviours

indirectly through mobilization efforts” (Rojas and Puig-I-Abril, 2009, p.902). Gil de

Zuniga and Valenzuela (2010) also showed that network size and conversations with

weak tiesonlinearea strongpredictorof civicbehaviours.Facebookcombinesmanyof

themechanismsbywhichonlineactivityhasbeen theorised toaffectmobilisation, such

as: lowering thecostofparticipationandmakingengagementwith,andcommitment to,

causes easier and more flexible (Bimber etal., 2005; Earl and Kimport, 2011; Shirky,

2008); improving political knowledge by allowing constant access to a wider range of

news media (Chaffee and Frank, 1996; Scheufele, 2002); building and maintaining of

social capital by building new bridges among people, strengthening existing offline

relationships (Ellison etal., 2007; Gil de Zuniga and Valenzuela, 2010) and enabling

6

individualstodevelopthenormsof trustandreciprocitythatarenecessaryforeffective

participationinpubliclife(KittilsonandDalton,2011).

Facebook’sMobilisingPotential

Facebook is predominantly a social networking site aimed at fulfilling social and

entertainment purposes. A number of studies have shown that it can satisfy people’s

entertainment and recreation needs (Ekstrom and Östman, 2013; Kahne etal., 2013),

enabling themto joingroupsandeventsrelated to theirpersonal interestsandhobbies,

upload their photos and videos, comment on, and chat with their friends. However,

Facebook has various embedded features that can make it a suitable venue for the

developmentofpoliticalattitudesandbehaviourseventothosepoliticallyuninterestedor

unaffiliated. Themost important of those is theNewsFeed timeline feature. Facebook’s

timelinestructureexposesuserstoamixedinformationenvironmentrangingfromsports,

movies, gossip,music and so on. This is important for the kind of ‘by-product’ learning

mechanism, which Chadwick, quoting Downs (Downs, 1957), argues can “soften

informationalinequalitiesbetweensocialgroups,ensurebroadpopularawarenessofkey

politicaleventsand,mostcrucially,spurustoactonthatknowledge,comeelectionday”

(Chadwick, 2012, p.41). Facebook is an ideal medium for fulfilling users’ informational

needs inthis fashion. Its integratedNewsFeedexposes,unintentionallyornot3,usersto

news sources, (political) opinions and potentiallymobilising information about political

events posted by friends. Early research has found that opinion expression and

conversation aboutpolitical affairs can lead tomobilisation and engagement inpolitical

activities(KatzandLazarsfeld,1955;Lazarsfeldetal.,1944).Recentstudies(Baum,2003;

Tewksbury etal., 2001; Morris and Morris, 2013) have corroborated the involuntary

learningeffectsofmassmediaandparticularlytheinternet.Moreover,despiterelyingon

aconveniencesampleofuniversitystudents,arecentstudyontheeffectsofFacebookon

participation has also confirmed that offline political participation is prominently

explainedbyexposuretosharedpoliticalinformationonFacebook(TangandLee,2013).

Whatismore,recentstudieshavealsoconfirmedthattheinternet–andespeciallysocial

3AccordingtoarecentstudybythePEWInternetandAmericanLifeProjectonly4percentofsocialmediausersblock,unfriend,orhidesomeonebecausetheydisagreedwithsomethingtheypostedaboutpolitics(http://pewinternet.org/~/media/Files/Reports/2012/PIP_SNS_and_politics.pdf)

7

media – is likely to trigger online political talk which is as conducive to political

engagementas interpersonaldiscussion (GildeZunigaandValenzuela,2010; Jungetal.,

2011;Rojas andPuig-I-Abril, 2009; Shah etal., 2005; Tewksbury etal., 2001). Based on

theseinsights,itbecomesapparentthatuseofFacebookmay,byvirtueoftheplatform’s

ownsocialisationlogicandtheembeddedfunctionspromotingit(e.g.liking,commenting,

sharing, starting a group), allow for the emergence of different uses that can lead to

differentmodesofparticipation.

Examining the impact of Facebook use on participation, several studies have generally

reportedpositiveeffects.ArecentstudybyEnjolrasandcolleagues(2013)foundthat,ina

protest context, links with information structures on social media exert a strong and

independenteffectonpoliticalmobilisation.Recent studiesusingpaneldatahave found

supportthatFacebookcanleadtoincreasedpoliticalactivity(Dimitrovaetal.,2014;Holt

etal.,2013),especiallywhentheyareusedforpoliticalnewsandvisitingpoliticalcontent

such as presidential candidates’ Facebookwebsites. It needs to be noted, however, that

these studies were conducted during election times and thus both interest in, and

attentionto,newsisheightened(DelliCarpiniandKeeter,1997).As isthecasewiththe

literature exploring the internet’s effect during the pre-Web 2.0 era, however, not all

research findings reveal positive effects. Zhang and colleagues (Zhang etal., 2009),who

looked specifically on the impact of socialmedia, found that despite that their usewas

significantly related to increased civic participation, there was no relationship with

political participation, a result corroborated by a recent study using panel data

(Theocharis&Quintelier,2015).Moreover,BaumgartnerandMorris(2010),contraryto

whatotherstudieshaveshown,foundthatthosewhorelyonsocialmediafornewswere

not more likely to participate in politics by traditional means. Extant research, thus,

provides inconclusive findings about the effect of Facebook on various forms of

participation, while notable is the small number of studies using longitudinal or

experimental designs to provide firmer support about the direction of causality

(Boulianne,2015).

FieldExperimentsasaMethodforTacklingFacebook’sEffectson

Participation

There is much ambivalence surrounding the direction of causality when it comes to

Facebook’s effects on different modes of participation. Nearly all studies presenting

8

positiverelationshipsbetweenFacebookuseandparticipationarecross-sectional,soitis

hard to rule out the possibility of confounding factors that affect social media use and

participation.WenotethatpastexperimentstudiesontheeffectsofFacebookonpolitical

participationdonot address this issuedirectly (Bond etal., 2012).Bond and colleagues

looked at mobilisation of existing Facebook users, whereas we follow a different path.

RatherthanstudyingtheeffectofpoliticalinformationseenbyalreadyexistingFacebook

users,weconsidertheeffectsofbringingpeopleontotheFacebookplatformbyrecruiting

people who did not have a Facebook account at all. If people not already using the

platformhaveadistinctprofileofpoliticalparticipationtothosewhoselectontoit, then

we will identify a different causal effect that does not necessarily originate in the

informationexposuremechanismexploredbyBondandcolleagues.

Againstthisbackgroundandespeciallyconsideringtheinteractiveandparticipatory(even

ifonlyexpressive)possibilitiesthatemergethroughsocialconnectionsonceonecreatesa

Facebook account, we expected that creating a Facebook account would lead to an

increase in one’s participation when compared with someone who does not have an

account.Asthisisanexperimentalstudywiththeaimoftestingtheindependenteffectof

havingaFacebookaccountonparticipationwedonotspeculateabouthowthe typesof

uses of Facebook that people may adopt after acquiring a Facebook account affect

participation (although, in our post hoc analysis, we do take into consideration how

different aspects of Facebook use that have proved influential in past research, like

intensity,networkembedednessandpoliticaluse,maybeaffectingthetreatmentgroup’s

participation patterns). Rather, by randomising ownership of Facebook account we are

interestedinwhetherownershipandmaintenanceofsuchanaccount(whichpresumably,

ifmaintained,willinvolvetypicalusesthatotherresearchhasshowntobeinfluential)is

causallyrelatedtomoreparticipation.

Given findings by extant research showing that online mobilisation can be medium-

specific(BestandKrueger,2005;Vissersetal.,2011)andthatonlineparticipationisitself

adistinctformofparticipation(Oseretal.,2013),weexpectthatFacebookusewillleadto

moreonlinepoliticalparticipation.Basedonrecentresearchonsocialmedia’simpacton

extra-institutionalparticipation (Cantijoch,2012;Valenzuela,2012),wealsoexpect that

Facebook’seffectwouldbemorepronouncedon formsofextra-institutionalrather than

traditionalformsofparticipation.Finally,asextantresearchsupportsthatthemoreopen

and collaborative online culture that social media allow for is more conducive to

9

engagement in civic activities (Bennett, 2012;Kahne etal., 2013;Pasek etal., 2009),we

expectFacebookusewillleadtomoretomorecivicengagement.Inhypothesisform:

Hypothesis 1 (H1): Use of Facebook leads to more offline participation

(traditional, civic and extra-institutional) and therefore Facebook users will

reporthigherofflineparticipationthannon-Facebookusers.

Hypothesis 2 (H2): Use of Facebook leads tomore online participation and

therefore Facebook users will report higher online participation than non-

Facebookusers.

MethodandMeasures

Caseselection

ThedatareportedinthisstudywerecollectedinGreecein2011,aperiodofwidespread

social turbulence, as well as unprecedented political instability due to the debt crisis.

Greek citizens’ strong reaction to austerity packages led to the organisation of mass

demonstrations and strikes, occupations of public institutions, widespread civil

disobedience and vibrant internet activism. The chosen period is thus of particular

significance and relevance to the subject of research. This is not only because of the

intense social turbulence in Greece during that period due to the peak of the financial

crisis, but also because it was the period leading up to what ended up to be two

consecutive elections in less than two months. Unsurprisingly, given how unpopular

politicalparties-andvirtuallyanythingthathadtodowithpolitics-wasatthetime,these

elections saw by far the lowest electoral turnout for more than 50 years. Given how

critical these elections were4, this not only implied an increased overall mediation of

politicalissuesandpositionsthrougheverycommunicationchannelavailable–withsocial

media thought to be amajor avenue formassmobilisation (Kathimerini, 2011; Henley,

2012), but also a period in which higher involvement with, and exposure to, political

issueswas to be expected (Delli Carpini and Keeter, 1996). Increased involvement and

4Manyanalystsarguedthattheresultofthatelectionwoulddefinewhetherthecountrywentbankrupt,andconsequentlybeexpelledfromtheEurozone,orremainwithintheEuroeconomicarea.

10

exposuretosuchavibrantandpolariseddiscussion,however,mayalsohaveresultedin

furtherdisillusionmentandevendisconnectwithpolitics.The lowelectoral turnoutand

thestrikingriseofprotestactivityduringtheentiredurationofthestudycertainlypoint

towardsthatdirectionandopenthedoorforcontext-basedinterpretationsofthestudy’s

findings.Despitepresentinganunusualcontext,andgiventheriseofsocialmedia-enabled

politicalactivitysincethen,weperceivethistime-frameasveryimportantfortheaimsof

this study which sought to test Facebook’s impact on political behaviour because the

circumstances and news reports imply intense use of social media platforms for

engagementwithpolitics.

Methodofrecruitmentandsampling

Between October 2011 and 2012 we commissioned a reputable independent Greek

researchagencytorecruit200peopleaged18-35(50percentfemale,50percentmale),

who did not have a Facebook account5. The participants were recruited using random

digitdialingfromPatras—thethirdlargesturbanareainGreece—andwerecontacted

fromSeptember1-6,2011andMarch1-6,2012.

Afteraskingthescreeningquestionsregardingtheparticipant’sageandwhethertheyhad

a Facebook account, the survey operators proceeded with the survey and, at the end,

asked the participants whether they would be interested in participating further.

Specifically, the survey operator asked the participants whether, in exchange for a gift

voucher from a popular electronics shop, they were willing to create and maintain a

Facebookaccountforayear,respondingtoasimilartelephoneinterviewsixmonthsafter

thatcall.Iftheparticipantaccepted,theoperatortooktheparticipant’sdetails,randomly

assigned them into the treatment (n=120)or control (n=80)6group, andasked tobe

notifiedofthetreatmentgroupparticipants’newlycreatedprofileonFacebookthrougha

Facebook message within two weeks. To ensure that participants would create and

5GreeceisamongtheleastsaturatedwithFacebookcountriesinEuropewitharound4millionusersandlessthan37percentpenetrationinthepopulation(http://www.checkfacebook.com).AstheagegroupofthehighestadoptionofFacebookinthecountryisthatof25-34,followedbythose18-24,wechosetofocusontheseagecategoriesaimingtocapturepeoplelikelytoadoptFacebookforthedurationofoursurvey.

6Theunequalsplitbetweentreatmentandcontrolgroupswasdecidedbasedontheexpectationthatpanelattritionwouldbemuchhigherinthetreatmentgroupduetothetypeoftreatment.

11

maintain a profile on Facebook, the agency sent the voucher details to the participants’

newly created Facebook profiles after the participants communicated to the agency

throughtheseprofiles.Thiswasalsothewayinwhichtheinitial(priortothetelephone

interview) subsequent communication with the survey participants took place for the

followupinterviewcall.Inthiswayweensuredthatthetreatmentgroupwereusingtheir

account.IntensityandtypeofusemadeoftheFacebookaccountweremeasuredthrougha

number of questions about intensity, number of friends acquired, andmore (see below

sectiononvariablesusedinposthocanalysis).Duringthecourseoftheexperimenttwo

control group members reported opening an account. They are excluded from this

analysis.

In thisstudywe focusat theeffectofhavingaFacebookaccountonparticipationat the

secondcontactpointwhenourrespondentshavemaintainedanaccount forsixmonths,

usingallrespondentswhocompletedWave2andcompliedwiththeinstructions(197of

200;onedrop-outandtwonon-compliers).Wedonothavestrongexpectationsaboutthe

timecourseofthemobilisationeffectswearestudying—itistheoreticallypossiblethat

effects on mobilisation are cumulative and more time is required to show a complete

effect,orthattheeffectappearsanddisappearsonaveryshorttimescale—however,we

expectthatsixmonthsshouldbealongenoughtoseetheeffectsofinterest.

Politicalparticipationmeasures

Weuse thedefinitionofpoliticalparticipationprovidedbyVerbaandcolleagues (Verba

etal., 1995, p.9), as an ‘activity that is intended to or has the consequence of affecting,

either directly or indirectly, government action’. However because different modes of

politicalparticipationmaynotonlybeusedbydifferentpeoplebutmayconstitutedistinct

types of engagementwith politics altogether (Cantijoch, 2012; Dalton etal., 2009; Oser

etal., 2013; Zukin etal., 2006), we distinguish analytically between online, offline,

traditional,civic,andextra-institutionalmodesofparticipationinadditiontoprovidinga

combinedmeasure.

We conceptualise offline political participation as having three components: traditional,

extra-institutional and civic participation. Traditional participation is participation

directlyattached to theelectoralprocess.Wemeasure thisusing theparticipation items

developedbyVerbaandNie (VerbaandNie,1972) thataskrespondents towhatextent

theyengagedinthelasttwelvemonthsinfivepredefinedpoliticalactivities.Theseinclude

votinginnationalelections,persuadingotherstovoteinelection,workingforapartyor

12

politician, raising funds for a political party and attending political meetings. These

responsesscaletogether(Mokken’sH=0.52(s.e.0.061),Cronbach’salpha=0.52)sowefita

binaryItemResponseTheory(IRT)modelandusestandardisedpointestimatesfromthe

posteriordistributionoffactorscoresasourindex.

WeusedIRTmodelsbecausewewanttomodelourLikertandbinaryitemsinawaythat

does not, like Factor Analysis, assume Normally distributed items, but does maintain

Factor Analysis’ assumption of a continuous underlying factor and local item

independence. Mokken's H statistic allows us to test that levels of the items we think

measure the same construct are monotonically related, i.e. that they do form a scale.

HavingconfirmedouritemsscaletogetherwefitgeneralisedIRTmodelstomaximisethe

information we can extract7. One implication of using IRT models (or indeed Factor

Analysis) is that our underlying constructs have no natural scale so can be arbitrarily

normalised. We use this to present our results in standard deviations of responses,

therebyabstractingawayfromthedetailsofthebinary,Likert,orothertypesofitemswe

used.

Wemeasureextra-institutionalparticipationbyaskingrespondents towhatextent they

engaged in the last twelve months in five pre-defined political activities (Barnes etal.,

1979).Theseincludeworkingforapoliticalactiongroup,signingapetition,deliberately

buyingoravoidingbuyingcertainproducts,participatingindemonstrationsanddonating

money to causes. These items scale together (Mokken’s H=0.44 (0.047), Cronbach’s

alpha=0.64)andweremeasuredusingstandardised factorscores froman IRTmodel,as

describedabove.

Weconceptualisecivicparticipationas‘participationaimedatachievingapublicgood,but

through direct hands-onwork in cooperationwith others, such as volunteerwork in a

community’(Zukinetal.,2006,p.51)andmeasureitbyaskingrespondentstowhatextent

they engaged in the last twelve months in three pre-defined political activities. These

includeworking or volunteering for a community project,working or volunteering in a

non-political group (e.g., hobby club or reforestation team) and raising money for, or

participating in, charity activity (e.g., a charitymarathon). The responses scale together

7 NotethatsimilarresultscanbeobtainedbyapplicationofFactorAnalysis,providedthatweworkwithanappropriatetypeofitemcorrelationmatrix,e.g.polychroriccorrelationsforourLikertitems.Weprefertoworkwiththedatadirectly.)

13

(Mokken’sH=0.52 (0.047), Cronbach’s alpha=0.68) andweremeasuredby standardised

factorsscoresfromanIRTmodel.

There is no standard battery of questions measuring online political participation, but

several scholars have attempted to construct suitable item sets (Bakker and deVreese,

2011) and scales tomeasure online participation (Gibson and Cantijoch, 2013; Hirzalla

and Van Zoonen, 2010). Following this work we operationalise online political

participation using six pre-defined political activities. These include visiting the

government’s or a politician’s website, reacting online to content posted by such a

website,signinganonlinepetition,visitingsitesofpoliticalactiongroups,reactingonline

to content posted by such awebsite and creating a blog or a social network group for

politicalreasons.Theresponseswerescaledasbeforetocreateanormalisedmeasureof

onlineparticipation(Mokken’sH=0.47(0.063),Cronbach’salpha=0.76).

Treatmentandcontrolvariables

The causal variable of interest is having andmaintaining a Facebook account. Although

thisisrandomised,samplingvariationinoursamplecangeneratecovariateimbalance,so

weperformouranalysiscontrollingforanumberofpossiblyconfoundingvariables.These

arethestandardpredictorsofpoliticalparticipationfromtheliterature: levelofpolitical

interest and ideological self-placement on a left-right axis (Norris, 2002; Verba etal.,

1995), but also postmaterialist orientations (Inglehart 1990) and membership in

associations (Pharr and Putnam, 2000). We also include control for age, gender and

highestlevelofcompletededucation.Inanattempttoimproveprecisionwealsoinclude

for each participation measure its pre-treatment value although our results are not

sensitivetothischoice.Mostresponseshadamissingnessrateof6percentorbelow,with

the exceptions of ‘did you vote in the last election’ (16.1 per cent missing), left-right

ideology on a 0-10 point scale (15.8 per cent missing), and income band (24 per cent

missing).Allmissingdataweremultiply imputed(usingtheAmelia IIpackage(Honaker

etal., 2011)).Unlessotherwise stated, allmodel results are constructedby combination

fromimputeddata(LittleandRubin,2002).

14

Variablesusedinthepost-hocanalysis

Taking into consideration findings from previous research, in our post-hoc analysiswe

includedseveralitemsthat,withinthecontextofsocialmedia,havebeenfoundinfluential

forsubsequentengagementindifferenttypesofparticipation.Theseincludedintensityof

Facebookuse(Xenosetal.,2014),measuredthroughthe item‘approximatelyhowmuch

time do you spend on Facebook’ with possible answers ranging from ‘less than 10

minutes’ to ‘more than3hours’; networkembeddedness (Gil deZuniga andValenzuela,

2010), whichwasmeasured through the item approximately howmany friends to you

haveonFacebook,withpossibleanswersrangingfrom‘lessthan10’to ‘morethan300’;

exposuretopotentiallymobilisinginformation(TangandLee,2013)whichwasmeasured

through the item ‘please tell me how much you agree or disagree with the following

statements:duringthelastsixmonthsIhavereadinformationonFacebookaboutsocial

andpoliticalinitiativesinwhichIwouldliketoparticipate’withpossibleanswersranging

from‘stronglyagree’to‘stronglydisagree’;andinteraction,throughdiscussion,ofpolitical

content posted by friends on Facebook (Rojas and Puig-I-Abril, 2009), measured by

throughitem‘Duringthelast6months…Ihavecommentedonfriends’postsofpolitical

natureonFacebook’.

Results

Toestimate theeffectofFacebookwe ran t-tests comparing treatment to control group

outcomesforeacheachparticipationmeasure(traditional,civic,extra-institutional,online

andageneralparticipationmeasure)andlinearregressionsontreatmentconditionwith

controlvariablesincludingtheparticipant’spre-treatmentscoreontheresponsevariable.

Since treatment is randomly assigned we expect, and largely confirm, no significant

relationshipswiththecontrols.

Asacheckontherobustnesstofunctionalformassumptionsintheregressionanalysiswe

additionally estimated an average treatment effect (ATE) viamatchingusingpropensity

scores (Sekhon, 2011). We used one-to-one matching with a 0.1 standard deviation

callipertodiscardhard-to-matchcases.Thisledtoaround15unmatchedrespondentsfor

eachparticipationvariable.Intheresultstablesbelowthematchingestimate‘ATE’should

becomparedtotheregressiontreatmentcoefficient‘Facebook’.

15

We do not report average treatment effects on the treatment (ATT) and control group

(ATC) separately. These are not directly comparable to the regression coefficient on

treatment (but provide substantively the same result) and, predictably, scarcely differ

from the ATE due to the randomised design. It is important to note that our research

design targets an average treatment effect on the control group (ATC) for the Greek

population because our treatment assignment gives Facebook accounts to a random

subset of subjects that do not already have them, but does not remove accounts from

those that do, so it cannot provide information about ATE or ATT effects of Facebook

directly.Nevertheless,inouranalysisanATEisappropriate.

We show matching results for a one imputation rather than average models that are

constructed frompotentially different cases in each imputation. As far aswe are aware

thereiscurrentlynostandardmethodologicalguidanceforperformingmatchinganalyses

onmultiplyimputeddata.Standarderrorsandconfidenceintervalscomputationusethe

Abadie-Imbensmethodwhich takes intoaccountvariationdue tomatching (Abadieand

Imbens,2006).

TheregressionandmatchingmodelsresultsareshowngraphicallyinFigure1.Table1,2

and 3 show the effects of treatment onmeasures of reported participation. Dependent

variables are unit-normalised so the treatment coefficients should be interpreted as

effectsinstandarddeviationsofreportedparticipation8.

[Figure1andTables1,2,and3abouthere]

The overall picture is straightforward: all point estimates of treatment effects are

negative.Inmoredetail,startingwithofflineformsofparticipation,thepointestimateof

8For each reported participation type we construct what is effectively a weighted

aggregateofpossiblydifferentnumbersandtypesofitemssoitmakessensetonormalisethefactorscorestohavefixedvariance.Thisallowsustointerpretcoefficientsintermsofstandard deviations of response, abstracting from the details of our survey items. Byaggregatinginanyway,fromadditiveindexconstructiontotheIRTmodelswefithere,weshould,provided the itemsdoall tap the sameparticipation type, improve theaccuracyand efficiency of measurement. The price paid is that it is not easy to map back toparticular items. It is also not possible to directly compare levels of different types ofreported participation. To offset this disadvantage we also construct a generalparticipationmeasure.

16

the effect of Facebook is negative on both traditional and extra-institutional forms of

politicalparticipation,thoughnotsignificantlyso,andwefindnosupport forH2.This is

surprising considering the large amount of studies that have demonstrated positive

effects, even weak or modest ones. Civic engagement is the only offline participation

variableuponwhichFacebookeffects are statistically significant.This stands inmarked

contrasttoKahneandcolleagues’(Kahneetal.,2013)andHoltandcolleagues’(Holtetal.,

2013) positive findings which were also derived from panel data. Moving to online

participation, our findings show the treatment has, again, a negative effect on online

politicalparticipation9.

The last Table (3) shows the results for a combined, ‘general political participation’

variableconstructedfromscaledfactorscoresusingbothofflineandonlineparticipation

items.Thefindingsareconsistentwiththepreviousmodelsandpredictablymoreprecise.

Theoverallmessage fromtheanalysis is thatacquisitionanduseofaFacebookaccount

leadstolowerlevelsofreportedpoliticalandcivicengagement.

What about the type of Facebook use? Previous empirical work has shown that it is

actually theways inwhichoneusesFacebook thatmay result tomoreparticipation.As

alreadydiscussed,fourofthemostprominentmobilisationmechanismswereexaminedin

thetreatmentgroup.Table4offersacorrelationbetweenourmaindependentvariables

and four independent variables used to measure intensity of use, size of Facebook

network,exposuretomobilisinginformationandusingFacebookforpoliticalpurposes.As

canbeseenintheTable,withtheexceptionofusingFacebookforpolitical/civicpurposes,

all other variables are weakly negatively correlated with all forms of participation.

Facebook network size and receiving mobilising information on Facebook are both

negatively and statistically significantly correlated with engaging in traditional acts.

Facebookuseforpolitical/civicpurposesistheonlyvariablethatispositivelycorrelated

with diverse forms of engagement, and is statistically significantly correlated with

traditional and online engagement. Although obviously no causal claims can be made

basedonthisanalysis,theeffectsgenerallycomplementtheoverallpicturethatacquiring

aFacebookaccountofferslittleconcreteparticipatorybenefits.

9Theeducationvariablealsogeneratesstatisticallysignificanteffects.Therearetoofewcasestoconfidentlydistinguishsubgroupeffects,althoughhigherlevelsofeducationcorrelateweaklywithlargerdropsinreportedparticipation.

17

[Table4abouthere]

Discussion

“Our sense is that the hyperbole surrounding newWeb developments (Web2.0) as they relate to citizenship may be just that—hype. In fact, it isreminiscentofearlierhopesfor‘‘Web1.0,’’andbeforethat,television,andevenearlierdevelopmentsinmasscommunicationsTechnology”

(BaumgartnerandMorris,2010,p.38)

Inoneof thefewstudies findingnosignificanteffectsofsocialmediauseonpolitical

participationandpoliticalattributessuchaspoliticalknowledge,BaumgartnerandMorris

urgedforcautionwheninterpretingthemobilisingpotentialofsocialmediaplatformslike

Facebook. Our study intended to contribute to a better understanding of the causal

relationship between digital media and participation, a question that has been at the

forefrontof thedebateon the internet’spotential since themediumwaspopularised in

thelate1990s.Muchofinternetandpoliticsliteraturehasbeenconcernedwiththeeffect

of digital media use on participation during the last decade yielding altogether more

positive than negative relationships (Boulianne, 2009) despite the lack of consistent

measures (Gibson and Cantijoch, 2013). Yet, as many scholars have noted (Boulianne,

2009, 2015;Dimitrova etal., 2014;Kahne etal., 2013; Shah etal., 2005), panel data are

bettersuitedfortacklingthiscausalquestion,andthesehavebeeninshortsupply.

OurstudycontributestothedebateprovidingevidenceshowingthatuseofFacebook,by

far themost popular socialmedia platformwithmore than 1 billion users, has a clear

negativeeffectonparticipation.Thefindingsaddscepticismaboutthepotentialofsocial

networking sites to affect participation; Facebook use not only fails to influence

engagement in a number of classic and widely used political and civic participation

repertoires,butaffectsthemnegatively.Perhapsthemosttroublingresultinourstudyis

thedemonstrationofanegativerelationshipinallformsofparticipation,ratherthan–as

isusuallyargued—inofflinehigh-costformsofengagement.Thisshouldraisetheoretical

concerns that go beyond the saturated ‘slacktivism’ and ‘clicktivism’ debates abound in

mainstreammedia(Gladwell,2010;Morozov,2009).

The starkness of our study’s counter-intuitive main finding invited us to search for

explanations through more subtle paths that could convincingly explain it. What

differentiates this study from studies that use panel data and find positive effects on

18

political and civic engagement is not only that itwasnot conductedduring an electoral

campaign(Dimitrovaetal.,2014;Holtetal.,2013),orwithasampleofschoolchildrenor

adolescents(Kahneetal.,2013;Theocharis&Quintelier,2015),butthatthisstudymakes

useof a randomisedexperimental designwhich requires frompeoplewhodon’t have a

Facebook profile to create andmake use of one. An important distinctionwas also the

contextinwhichthestudytookplace.Indeed,asourfindingsechoearlystudiesreporting

theinternetasatoolthatcontributestosocialisolationandfurtherdisengagement(Kraut

etal.,1998;NieandErbring,2000)duetotheentertainmentfunctionsofthemedium,we

find that a convincing explanationmay lie on Facebook’s recreation and entertainment

capacities.

ArivaltheorytotheonethatseesFacebookasamobilisingfactorforengagementisone

that sees it as a de-mobilizing force whose functions distract the users’ attention from

politics.Althoughourdatasetdidnot includevariables for testing thispossibility, this is

farfromanexoticspeculation.FromPostman’sdystopianwarningsinAmusingOurselves

toDeath (Postman,1985) toNorrisand Jones’ (1998)concerns thatrecreationalusesof

theinternetmayerodepublicinvolvement,thepossibilityofusingtheinternetpurelyasa

methodofdistractingoneself fromhasalwaysbeenvisible.That the situation inGreece

duringthetime inwhichthestudywasconductedwaschallengingformostGreeks,and

downright dire for many, reinforces the idea that Facebook may have been a way of

escaping everyday struggles by disassociating oneself from, rather than engaging with,

politics. Political communication literature has long emphasised that media can be a

sourceofentertainment thatoffersmomentsofescapismfromthestrugglesofdaily life

(LazarsfeldandMerton,1948).Empiricalresearchhasalsosupportedthisidea.Inastudy

ontheeffectsofEastGermans’exposuretoWestGermantelevisionontheirsupportofthe

EastGermancommunistregime,KernandHainmueller(2009)foundevidenceconsistent

with the theory that television was primarily used as a source for entertainment and

distractionthatmadelifeundercommunismmorebearable.Itisthusplausibletosuggest

that, intheworstdaysofthecrisisinGreece,Facebookplayedasimilarlyentertainment

roleindemobilisingratherthanamobilisingthepopulation.

There is also a different, and almost contrary, interpretation. One of the major

characteristics of the time period in which the study was conducted was the strong

polarisationamongthepublicopinionregardingthepaththecountryshouldtake.Atthe

time ofwriting, Greece is undergoing a very similar and probably evenmore turbulent

periodduringwhichthepopulation, facingbankclosuresandcapitalcontrols, is fiercely

19

dividedoverareferendumtoapproveordisapprovetheinternationalcreditors’package

ofausteritymeasures.Oneofthemostmajorconcernsvoicedinthemainstreammediaat

the time – and consistently since - is the effect of polarisation on social media which

createsanincisive,uncivil,andoftendeleteriousatmosphere,likenedbythemajorGreek

newspaper Kathimerini to a “civil war” (Kathimerini, 2015). A number of studies has

shownthatincisiveanduncivildiscussionsonsocialmediahaveanegativeeffectnotonly

ontheattitudesandbehavioursofthosewhoengageinthediscussionitself(Hwangetal.,

2008),butalsoadisillusioningeffecttothosewhoattend(Andersonetal.,2013).Withthe

discussionragingonFacebook,withaccusations frombothsupportersandcriticsof the

government,andwithrowsbetweenfriendsandacquaintancesattheirpeak,itisfarfrom

implausible that using the platformmay have led to disaffection with politics. Such an

explanationisalsofirmlygroundedinexistingliterature. Inarandomisedexperiment in

which a subset of “internet-less” people in Tanzania, prior to the presidential election

were assigned to the “internet use” treatment group which was given access to the

internet,Bailard(2012)foundthattheinternetinfluencedindividuals’perceptionforthe

fairnessoftheelectionnegatively.WeagreewithBailard’sstatementthattheinternet’s--

andinourcaseFacebook’s--effectsmaybeheavilycontextuallydependent.Thisstudy’s

surprising results do not, thus, call only for a replication of our design in a different

context,butalsoopensupanewandexcitingavenueforfurtherresearchonsocialmedia

thattakescontextualfactorsintogreaterconsideration.

PuttingasidetheexplanationsofFacebook’snegativeeffects,themainimplicationofthis

study is that increased caution should be exercised when it comes to statements and

assumptions about the direction of causality that underlies the relationship between

socialmedia use and participation. This research underlines the need formore studies

with longitudinal and experimental designs exploring the relationship between digital

mediaandparticipationandthemechanismsofinfluence.

Finally, there are two limitations to the study: first, the sample size is relatively small.

Thus,whilewe canbe fairly confidentofourgeneral findingswedonothave sufficient

precision tocompareeffect sizes fordifferentparticipationmeasures.Second,whileour

itemsdoscaletogetheraspredictedbytheoryandconfirmedbyMokken’smeasures,they

are not particularly strong. In particular the weaker scales are thosemeasuring offline

participation,whichmayexplain the imprecise results. If this is the case,weshouldnot

readtoomuch intothenon-significanceof therelevanttreatmentcoefficients; forallwe

knowtheeffectsmaybeasrobustlynegativeastheonlinemeasuresorunaffectedbythe

20

manipulation. Another limitation of the experimental design is that we do not get

informationabouttheparticularmechanismsthatmaybecausingnegativeeffects.Thisis

astraightforwardconsequenceoftherandomisedexperimentaldesign.

References

Abadie,A.andImbens,G.(2006).LargeSamplePropertiesofMatchingEstimatorsfor

AverageTreatmentEffects.Econometrica,74:235–267.

Anderson,A.A.,Brossard,D.,Scheufele,D.,Xenos,M.,andLadwig,P.(2013).The“Nasty

Effect”:OnlineIncivilityandRiskPerceptionsofEmergingTechnologies.Journalof

Computer-MediatedCommunication,19(3):373-387.

Bachmann,I.anddeZuniga,H.(2013).NewsPlatformPreferenceasapredictorof

politicalandcivicparticipation.Convergence:TheInternationalJournalofResearchinto

NewMediaTechnologies.

Bailard,C.S.(2012).AFieldExperimentontheInternet’sEffectinanAfricanElection:

SavvierCitizens,DisaffectedVoters,ofBoth?.JournalofCommunication,62(2):330-344.

Bakker,T.P.anddeVreese,C.H.(2011).GoodNewsfortheFuture?YoungPeople,

InternetUse,andPoliticalParticipation.CommunicationResearch,38(4):451–470.

Barber,B.(1999).ThreeScenariosfortheFutureofTechnologyandStrongDemocracy.

PoliticalScienceQuarterly,113(4):573–589.

Barnes,S.,Kaase,M.,Allerbeck,K.,Heunks,F.,Inglehart,R.,Jennings,M.,Klingemann,H.,

Marsh,A.,andRosenmayr,L.(1979).PoliticalAction:MassParticipationinFiveWestern

Democracies.Sage,BeverlyHills.

Baum,M.(2003).Softnewsgoestowar:PublicopinionandAmericanforeignpolicy.

PrincetonUniversityPress,Princeton.

Baumgartner,J.andMorris,S.(2010).MyFaceTubePolitics:SocialNetworkingWebSites

andPoliticalEngagementofYoungAdults.SocialScienceComputerReview,28(1):24–44.

Bennett,L.(2012).ThePersonalizationofPolitics:PoliticalIdentity,SocialMedia,and

ChangingPatternsofParticipation.TheANNALSoftheAmericanAcademyofPoliticaland

SocialScience,644(1):20–39.

Bennett,L.andSegerberg,A.(2013).TheLogicofConnectiveAction:DigitalMediaandthe

PersonalizationofContentiousPolitics.CambridgeUniversityPress,Cambridge.

21

Best,S.andKrueger,B.(2005).AnalyzingtheRepresentativenessofInternetPolitical

Participation.PoliticalBehaviour,27(2):183–216.

Bimber,B.(2001).InformationandPoliticalEngagementinAmerica:TheSearchfor

EffectsofInformationTechnologyattheIndividualLevel.PoliticalResearchQuarterly,

54(1):53–67.

Bimber,B.,Flanagin,A.,andStohl,C.(2005).Reconceptualizingcollectiveactioninthe

contemporarymediaenvironment.CommunicationTheory,15(4):365–388.

Bond,R.,Fariss,C.,Jones,J.,Kramer,A.,Marlow,J.,andFowler,J.(2012).A61-million-

personexperimentinsocialinfluenceandpoliticalmobilization.Nature,489:295–298.

Boulianne,S.(2009).DoesInternetUseAffectEngagement?AMeta-AnalysisofResearch.

PoliticalCommunication,26(2):193–211.

Boulianne,S.(2015).SocialMediauseandParticipation:aMeta-analysisofCurrent

research.Information,Communication&Society,18:524–538.

Cantijoch,M.(2012).DigitalMediaandOfflinePoliticalParticipationinSpain.InDigital

MediaandPoliticalEngagementWorldwide:AComparativeStudy,pages118–137.

CambridgeUniversityPress,Cambridge.

Chadwick,A.(2012).RecentShiftsintheRelationshipBetweentheInternetand

DemocraticEngagementinBritainandtheUnitedStates:Granularity,Informational

Exuberance,andPoliticalLearning.InAnduiza,E.,Jensen,J.,andJorba,L.,editors,Digital

MediaandPoliticalEngagementWorldwide:AComparativeStudy,pages39–55.Cambridge

UniversityPress,Cambridge.

Chaffee,E.andFrank,S.(1996).HowAmericansgetpoliticalinformation:Printversus

broadcastnews.AnnalsoftheAmericanAcademyofPoliticalandSocialScience,546:48–58.

Coleman,S.andBlumler,J.(2009).TheInternetandDemocraticCitizenship:Theory,

PracticeandPolicy.CambridgeUniversityPress,Cambridge.

Dalton,R.(2008).CitizenshipNormsandtheExpansionofPoliticalParticipation.Political

Studies,56(1):76–98.

Dalton,R.,VanSickle,A.,andWeldon,S.(2009).TheIndividual–InstitutionalNexusof

ProtestBehaviour.BritishJournalofPoliticalScience,40(01):51.

Dalton,R.andWattenberg,M.(2000).PartiesWithoutPartisans:PoliticalChangein

AdvancedIndustrialDemocracies.OxfordUniversityPress,Oxford.

22

DelliCarpini,M.andKeeter,S.(1996).WhatAmericansKnowAboutPoliticsandWhyIt

Matters.YaleUniversityPress,NewHaven,CT.

DelliCarpini,M.andKeeter,S.(1997).WhatAmericansKnowAboutPoliticsandWhyit

Matters.YaleUniversityPress,NewHaven.

Dimitrova,D.V.,Shehata,A.,Stromback,J.,andNord,L.W.(2014).TheEffectsofDigital

MediaonPoliticalKnowledgeandParticipationinElectionCampaigns:EvidenceFrom

PanelData.CommunicationResearch,41(1):95–118.

Downs,A.(1957).AnEconomicTheoryofDemocracy.HarperandRow,NewYork.

Earl,J.andKimport,K.(2011).DigitallyEnabledSocialChange:ActivismintheInternet

Age.MITPress,Cambridge.

Ekstrom,M.andÖstman,J.(2013).Information,Interaction,andCreativeProduction:The

EffectsofThreeFormsofInternetUseonYouthDemocraticEngagement.Communication

Research.

Ellison,N.,Steinfield,C.,andLampe,C.(2007).TheBenefitsofFacebook“Friends:”Social

CapitalandCollegeStudents’UseofOnlineSocialNetworkSites.JournalofComputer-

MediatedCommunication,12(4):1143–1168.

Enjolras,B.,Steen-Johnsen,K.,andWollebaek,D.(2013).Socialmediaandmobilizationto

offlinedemonstrations:Transcendingparticipatorydivides?NewMedia&Society,

15(6):890–908.

Etzioni,A.(1993).TheSpiritofCommunity:Rights,Responsibilities,andtheCommunitarian

Agenda.CrownPublishers,NewYork.

Gibson,R.andCantijoch,M.(2013).ConceptualizingandMeasuringParticipationinthe

AgeoftheInternet:IsOnlinePoliticalEngagementReallyDifferenttoOffline?TheJournal

ofPolitics,75(3):701–716.

GildeZuniga,H.andValenzuela,S.(2010).TheMediatingPathtoaStrongerCitizenship:

OnlineandOfflineNetworks,WeakTies,andCivicEngagement.CommunicationResearch,

38(3):397–421.

Gladwell,M.(2010).WhytheRevolutionWillnotbeTweeted.TheNewYorker.

Hay,C.(2007).WhyWeHatePolitics.PolityPress,Cambridge.

Henley,J.(2012).Greeceonthebreadline:poolingresourcestoprovideaneducation.

23

Hirzalla,F.andVanZoonen,L.(2010).BeyondtheOnline/OfflineDivide:HowYouth’s

OnlineandOfflineCivicActivitiesConverge.SocialScienceComputerReview,29(4):481–

498.

Holt,K.,Shehata,A.,Stromback,J.,andLjungberg,E.(2013).Ageandtheeffectsofnews

mediaattentionandsocialmediauseonpoliticalinterestandparticipation:Dosocial

mediafunctionasleveller?EuropeanJournalofCommunication,28(1):19–34.

Honaker,J.,King,G.,andBlackwell,M.(2011).AmeliaII:AProgramforMissingData.

JournalofStatisticalSoftware,45(7):1–47.

Howard,P.N.andHussain,M.M.(2013).Democracy’sFourthWave?DigitalMediaandthe

ArabSpring.OxfordUniversityPress,Oxford.

Hwang,H.Borah,P.Namkoong,K.andVeenstra,A.(2008).DoesCivilityMatterinthe

Blogosphere?ExaminingtheInteractionEffectsofIncivilityandDisagreementoncitizen

Attitudes.Paperpresentedatthe58thAnnualConferenceoftheInternational

CommunicationAssociation,Montreal.

Jung,N.,Kim,Y.,andGildeZúñiga,H.(2011).TheMediatingRoleofKnowledgeand

EfficacyintheEffectsofCommunicationonPoliticalParticipation.MassCommunication

andSociety,14(4):407–430.

Kathimerini(2015).«Εμφύλιος»σταμέσακοινωνικήςδικτύωσης”[“Civilwar”insocial

networkingsites”.Kathimerini,June6,Availableonlineat:

http://www.kathimerini.gr/821393/article/epikairothta/politikh/emfylios-sta-mesa-

koinwnikhs-diktywshs.

Kahne,J.,Lee,N.J.,andFreezel,J.(2013).TheCivicandPoliticalSignificanceofOnline

ParticipatoryCulturesamongYouthTransitioningtoAdulthood.JournalofInformation

Technology&Politics,10(1):1–20.

Kathimerini(2011).’Indignant’Greekstoreturnforseconddayofprotest.

Katz,E.andLazarsfeld,P.(1955).PersonalInfluence:ThePartPlayedbyPeopleintheFlow

ofMassCommunication.FreePress,NewYork.

Kern,H.andHaimueller,J.(2009).OpiumfortheMasses:HowForeignMediaCanStabilize

AuthoritarianRegimes.PoliticalAnalysis,17(4):377–399.

24

Kittilson,M.andDalton,R.(2011).VirtualCivilSociety:TheNewFrontierofSocial

Capital?PoliticalBehaviour,33:625–644.

Kraut,R.,Patterson,M.,Lundmark,V.,Kiesler,S.,Mukopadhyay,T.,andScherlis,W.

(1998).Internetparadox.Asocialtechnologythatreducessocialinvolvementand

psychologicalwell-being?AmericanPsychologist,53(9):1017–31.

Lazarsfeld,P.,Berelson,B.,andGaudet,H.(1944).ThePeople’sChoice:HowtheVoter

MakesuphisMindinaPresidentialCampaign.Duel,Sloan&Pearce,NewYork.

Lazarsfeld,P.andMerton,R.(1948).MassCommunication,PopularTaste,andOrganized

SocialAction.InBryson,L.,editor,TheCommunicationofIdeas,pages95–118.Harper,

NewYork.

Little,R.andRubin,D.(2002).StatisticalAnalysiswithMissingData.Wiley,London.

Lynch,M.(2011).AfterEgypt:TheLimitsandPromiseofOnlineChallengestothe

AuthoritarianArabState.PerspectivesonPolitics,9(2):301–310.

Morozov,E.(2009).Thebravenewworldofslacktivism.ForeignPolicy.

Morris,S.andMorris,S.(2013).DigitalInequalityandParticipationinthePolitical

Process:RealorImagined?SocialScienceComputerReview.

Mossberger,K.,Tolbert,C.J.,andMcNeal,R.S.(2008).DigitalCitizenship:TheInternet,

SocietyandParticipation.MITPress,Cambridge.

Nie,N.andErbring,A.(2000).Internetandsociety:Apreliminaryreport.Technicalreport,

StanfordInstitutefortheQuantitativeStudyofSociety,Stanford.

Norris,P.(2001).DigitalDivide:CivicEngagement,InformationProverty,andtheInternet

Worldwide.CambridgeUniversityPress,Cambridge.

Norris,P.(2002).DemocraticPhoenix:ReinventingPoliticalActivism.CambridgeUniversity

Press,Cambridge.

Norris,P.andJones,D.(1998).VirtualDemocracy.TheHarvardInternationalJournalof

Press/Politics,3(1):1–4.

Oser,J.,Hooghe,M.,andMarien,S.(2013).IsOnlineParticipationDistinctfromOffline

Participation?ALatentClassAnalysisofParticipationTypesandTheirStratification.

PoliticalResearchQuarterly,66(1):91–101.

25

Pasek,J.,More,E.,andRomer,D.(2009).RealizingtheSocialInternet?OnlineSocial

NetworkingMeetsOfflineCivicEngagement.JournalofInformationTechnology&Politics,

6(3):197–215.

Pharr,S.andPutnam,R.(2000).DisaffectedDemocracies:What’sTroublingtheTrilateral

Countries?PrincetonUniversityPress,Princeton.

Postman,N.(1985).AmusingOurselvestoDeath:PublicDiscourseintheAgeofShow

Business.Penguin,NewYork.

Putnam,R.(2000).Bowlingalone:ThecollapseandrevivalofAmericancommunity.Simon

&Schuster,NewYork.

Quintelier,E.andVissers,S.(2008).TheEffectofInternetUseonPoliticalParticipation:

AnAnalysisofSurveyResultsfor16-Year-OldsinBelgium.SocialScienceComputer

Review,26(4):411–427.

Rojas,H.andPuig-I-Abril,E.(2009).MobilizersMobilized:Information,Expression,

MobilizationandParticipationintheDigitalAge.JournalofComputer-Mediated

Communication,14(4):902–927.

Scheufele,D.A.(2002).ExaminingDifferentialGainsfromMassMediaandtheir

ImplicationsforParticipatoryBehaviour.CommunicationResearch,29(1):46–65.

Schlozman,K.,Verba,S.,andBrady,H.(2012).TheUnheavenlyChorus:UnequalPolitical

VoiceandtheBrokenPromiseofAmericanDemocracy.PrincetonUniversityPress,New

Jersey.

Sekhon,J.(2011).MultivariateandPropensityScoreMatchingSoftwarewithAutomated

BalanceOptimization.JournalofStatisticalSoftware,42(7):1–52.

Shah,D.,Cho,J.,Eveland,W.,andKwak,N.(2005).InformationandExpressioninaDigital

Age:ModelingInternetEffectsonCivicParticipation.CommunicationResearch,32(5):531–

565.

Shirky,C.(2008).HereComesEverybody:HowChangeHappensWhenPeopleCome

Together.Penguin,London.

Shirky,C.(2011).ThePoliticalPowerofSocialMedia.Foreignaffairs,90(1):28–41.

Smith,A.(2013).CivicEngagementintheDigitalAge.Technicalreport,PewResearch

Center.

26

Tang,G.andLee,F.(2013).FacebookUseandPoliticalParticipationTheImpactof

ExposuretoSharedPoliticalInformation,ConnectionsWithPublicPoliticalActors,and

NetworkStructuralHeterogeneity.SocialScienceComputerReview,31(6):763–773.

Tewksbury,D.,Weaver,A.,andMaddex,B.(2001).AccidentallyInformed:IncidentalNews

ExposureontheWorldWideWeb.JournalismandMassCommunicationQuarterly,

78(3):533–544.

Theocharis,Y.&Quintelier,E.(2015)StimulatingCitizenshiporExpanding

Entertainment?TheeffectofFacebookonadolescentparticipation.NewMedia&Society,

availableonlinebeforeprint:

http://nms.sagepub.com/content/early/2014/08/27/1461444814549006.abstract

Theocharis,Y.,Lowe,W.,vanDeth,J.W.&García-Albacete,G.(2015)“UsingTwitterto

MobilizeProtestAction:OnlineMobilizationPatternsandActionRepertoiresinthe

OccupyWallStreet,Indignados,andAganaktismenoiMovements.Information,

Communication&Society,18(2):202-220.

Towner,T.(2013).AllPoliticalParticipationisSociallyNetworked?NewMediaandthe

2012Election.SocialScienceComputerReview,31(5):527–541.

Tufekci,Z.andWilson,C.(2012).SocialMediaandtheDecisiontoParticipateinPolitical

Protest:ObservationsFromTahrirSquare.JournalofCommunication,62(2):363–379.

Valenzuela,S.(2012).ProtestingintheAgeofSocialMedia:Information,Opinion

ExpressionandActivisminOnlineNetworks.InThe5thLatinAmericanPublicOpinion

Congress,Bogota,Colombia.

Valenzuela,S.,Park,N.,andKee,K.F.(2009).IsThereSocialCapitalinaSocialNetwork

Site?:FacebookUseandCollegeStudents’LifeSatisfaction,Trust,andParticipation.

JournalofComputer-MediatedCommunication,14(4):875–901.

VanLaer,J.andVanAelst,P.(2010).InternetandSocialMovementActionRepertoires.

Information,Communication&Society,13(8):1146–1171.

Verba,S.andNie,N.(1972).ParticipationinAmerica:PoliticalDemocracyandSocial

Equality.UniversityofChicagoPress,Chicago.

Verba,S.,Schlozman,K.,andBrady,H.(1995).VoiceandEquality:CivicVoluntarismin

AmericanPolitics.HarvardUniversityPress,Cambridge.

27

Vissers,S.,Hooghe,M.,Stolle,D.,andMaheo,V.-A.(2011).TheImpactofMobilization

MediaonOff-LineandOnlineParticipation:AreMobilizationEffectsMedium-Specific?

SocialScienceComputerReview,30(2):152–169.

Wellman,B.,Haase,A.Q.,Witte,J.,andHampton,K.(2001).DoestheInternetIncrease,

Decrease,orSupplementSocialCapital?:SocialNetworks,Participation,andCommunity

Commitment.AmericanBehaviouralScientist,45(3):436–455.

Xenos,M.andMoy,P.(2007).DirectandDifferentialEffectsoftheInternetonPoliticaland

CivicEngagement.JournalofCommunication,57(4):704–718.

Xenos,M.,Vromen,A.,andLoader,B.(2014).Thegreatequalizer?Patternsofsocialmedia

useandyouthpoliticalengagementinthreeadvanceddemocracies.Information,

Communication&Society,17(2):151–167.

Zhang,T.,Johnson,T.,Seltzer,T.,andBichard,S.L.(2009).TheRevolutionWillbe

Networked:TheInfluenceofSocialNetworkingSitesonPoliticalAttitudesandBehaviour.

SocialScienceComputerReview,28(1):75–92.

Zukin,C.,Keeter,S.,Andolina,M.,Jenkins,K.,andDelliCarpini,M.(2006).ANew

Engagement?PoliticalParticipation,CivicLife,andtheChangingAmericanCitizen.Oxford

UniversityPress,Oxford.

Acknowledgment: The authors would like to thank the participants of the "Social Media and Political Participation" panel of the 2014 EPSA General Conference, as well as the participants of the "Experimental Research" panel of the 110th APSA Annual Meeting for commenting on earlier versions of this paper. Yannis Theocharis would especially like to thank Jan W. van Deth for his advice, ideas about the study’s design and his support, as well as to acknowledge the financial contribution of the Alexander von Humboldt Foundation and the Mannheim Centre for European Social Research. Yannis Theocharis is Research Fellow at the Mannheim Centre for European Social Research (MZES), University of Mannheim. His research interests are in political behaviour, new media, social networks, and social capital. Will Lowe is Senior Research Specialist in the Department of Politics at Princeton University. He is a political methodologist focusing on quantitative political text analysis.

28

Table 1: Models of offline participation modes. For each mode, model 0 is equivalent to a t-test of difference of means, model 1 is a linear regression, and model 2 is the ATE from a matching analysis using propensity scores.

29

Table 2: Models for offline and online modes. For each mode, model 0 is equivalent to a t-test of difference of means, model 1 is a linear regression, and model 2 is the ATE from a matching analysis using propensity scores. Standard errors are in parentheses.

30

Table 3: Models for all measures combined. For each mode, model 0 is equivalent to a t-test of difference of means, model 1 is a linear regression, and model 2 is the ATE from a matching analysis using propensity scores. Standard errors are in parentheses.

31

Table 4: Correlation table of modes with attributes of the treated group.

32

Figure 1: Estimated treatment effects by dependent variable for regression and matching analyses.