《CEPR:社交媒体的政治经济学(英文版)(213页).pdf》由会员分享,可在线阅读,更多相关《CEPR:社交媒体的政治经济学(英文版)(213页).pdf(213页珍藏版)》请在三个皮匠报告上搜索。
1、Edited by Filipe R.Campante,Ruben Durante and Andrea TeseiThe Political Economy of Social Media 9781912179800ISBN 978-1-912179-80-0187 boulevard Saint-Germain|75007 Paris|France33 Great Sutton Street|LONDON EC1V 0DX|UK TEL:+44(0)20 7183 8801|EMAIL:CEPRCEPR.ORGWWW.CEPR.ORGThe emergence of social medi
2、a has reshaped the way humans communicate,interact and coordinate with each other.Assessing the impact of that transformation on politics has been one of the great social science questions of the last or decade or so,and will continue to occupy researchers for a long time to come.This book provides
3、a snapshot of how economists in particular have been trying to answer this question.It contains 18 chapters,written by some of the leading scholars working on the topic,summarising empirical evidence on different dimensions of the political impact of social media.The book starts by considering how s
4、ocial media platforms have affected the overall wellbeing of their users.It then goes over how they have changed the behaviour of voters,particularly through news consumption,and whether it can be linked to phenomena such as increased polarisation or the rise of populism.The next section looks at ho
5、w politicians have responded to the new environment,and how that in turn has affected elections.The following two sections address the coordination role of social media,asking how it has affected political mobilisation and,on the negative side,the spread of political hatred.Another section focuses o
6、n how social media has changed politics in the autocratic context of China.Lastly,the final chapters shed light on how the political role of other,so-called legacy media has been impacted by the new technologies.Put together,the contributions described in this book showcase how the ubiquity of socia
7、l media,the nature of the networks that emerge through it,and the absence of barriers to entry in producing and broadcasting content all converge to make this technology a uniquely consequential transformation in the media environment.The Political Economy of Social Media The Political Economy of So
8、cial Media WITH SUPPORT OF CEPR PARIS FOUNDING PARTNERSCEPR PRESSCentre for Economic Policy Research187 boulevard Saint-Germain75007,Paris,France 33 Great Sutton StreetLondon,EC1V 0DX,UKTel:+44(0)20 7183 8801Email:ceprcepr.orgWeb:www.cepr.org ISBN:978-1-912179-80-0Copyright CEPR PressThe Political E
9、conomy of Social Media Edited by Filipe R.Campante,Ruben Durante and Andrea TeseiCENTRE FOR ECONOMIC POLICY RESEARCH(CEPR)The Centre for Economic Policy Research(CEPR)is a network of over 1,700 research economists based mostly in European universities.The Centres goal is twofold:to promote world-cla
10、ss research,and to get the policy-relevant results into the hands of key decision-makers.CEPRs guiding principle is Research excellence with policy relevance.It was founded in the UK in 1983,where it is a Charity,and in November 2019 CEPR initiated the creation of an Association under French law,in
11、order to provide a vehicle for an expansion in France.The members of the Conseil dAdministration of the Association are identical to the UK Board of Trustees.CEPR is independent of all public and private interest groups.It takes no institutional stand on economic policy matters and its core funding
12、comes from its Institutional Members,projects that it runs and sales of publications.Because it draws on such a large network of researchers,its output reflects a broad spectrum of individual viewpoints as well as perspectives drawn from civil society.CEPR research may include views on policy,but th
13、e Trustees/members of the Conseil dAdministration of the Association do not give prior review to its publications.The opinions expressed in this report are those of the authors and not those of CEPR.Chair of the Board Sir Charlie BeanFounder and Honorary President Richard PortesPresident Beatrice We
14、der di MauroVice Presidents Maristella Botticini Philippe Martin Ugo Panizza Mar Reguant Hlne ReyChief Executive Officer Tessa OgdenContentsForeword viiIntroduction 9Filipe R.Campante,Ruben Durante and Andrea TeseiSection 1 Welfare effects of social media1 The welfare effects of social media 25Hunt
15、Allcott,Luca Braghieri,Sarah Eichmeyer and Matthew Gentzkow2 Social media and mental health 31Luca Braghieri,Roee Levy and Alexey MakarinSection 2 Social media and voters3 Social media,news consumption and polarisation 47Roee Levy4 Homophily,group size and the diffusion of political information in s
16、ocial networks 59Yosh Halberstam and Brian Knight5 Political implications of the rise of mobile broadband internet 67Sergei Guriev,Nikita Melnikov and Ekaterina Zhuravskaya6 Mobile internet and the rise of communitarian politics 77Marco Manacorda,Guido Tabellini and Andrea Tesei7 The effect of socia
17、l media on elections:Evidence from the United States 89Thomas Fujiwara,Karsten Mller and Carlo SchwarzSection 3 Social media and politicians8 Politics 2.0:The multifaceted effect of broadband internet on political participation 101Filipe Campante,Ruben Durante and Francesco Sobbrio9 New technologies
18、 and political competition:The impact of social media communication on political contributions 111Maria Petrova,Ananya Sen and Pinar YildirimSection 4 Social media and mobilisation10 Social media and protest participation:Evidence from Russia 121Ruben Enikolopov,Alexey Makarin and Maria Petrova11 So
19、cial media and mobilisation 131Leopoldo Fergusson and Carlos Molina12 Liberation technology:Mobile phones and political mobilisation in Africa 143Marco Manacorda and Andrea TeseiSection 5 Social media and hatred13 Social media and Xenophobia:Evidence from Russia 153Leonardo Bursztyn,Georgy Egorov,Ru
20、ben Enikolopov and Maria Petrova14 Can social media spur offline hatred?163Karsten Mller and Carlo SchwarzSection 6 Social media in autocracies15 The political economy of social media in China 175Bei Qin,David Strmberg and Yanhui Wu 16 Social media in autocracies 185David Y.Yang Section 7 Social med
21、ia and legacy media17 Social media and legacy media 195Sophie Hatte and Ekaterina Zhuravskaya 18 Contagion from social media to mainstream media 203Julia Cag,Nicolas Herv and Batrice Mazoyer vIIForewordThe rise of social media has profoundly transformed society,reshaping communication and informatio
22、n consumption.Its widespread use has placed unprecedented pressure on norms and institutions,challenging them to adapt to a rapidly changing social and political landscape.Its legacy,meanwhile,is already being fiercely contested by academics.This eBook brings together a diverse array of contribution
23、s focusing on the political economics of social media,providing a comprehensive exploration of the impact of the internet and social media on the global political landscape.The chapters contribute valuable insights into the welfare effects of social media,highlighting its addictive nature and negati
24、ve ramifications on mental health.The research also explores the influence of social media on voter behaviour,including its role in exacerbating recent polarisation trends through the creation of like-minded echo chambers and the selective dissemination of political information,which can both enhanc
25、e transparency and contribute to political discontent.Other chapters explore the dynamic relationship between politicians and social media,revealing how these platforms enable politicians to effectively raise resources and mobilise voters,at a fraction of the cost of traditional methods.The authors
26、uncover the influential role of social media in shaping elections,but not always in a consistent direction.The research also addresses the complex interplay between social media and protests,tracing its emergence as a highly effective tool for coordination and collective action.Finally,the authors e
27、xplore the unique impact of social media in autocratic regimes and investigates their impact on the global legacy media.Overall,this eBook provides compelling insights into social medias nuanced and complicated impact on the political and social landscape.Rather than providing judgements on whether
28、social media is good or bad,the chapters underscore the variability of consequences based on diverse outcomes and contexts.These findings will help inform policymakers on the future direction of social media regulation and moderation,recognising its enduring significance in society,particularly in t
29、he realm of politics.CEPR is grateful to Filipe R.Campante,Ruban Durante,and Andrea Tesei for their expert editorship of the eBook.Our thanks also go to Anil Shamdasani for his skilled handling of its production.CEPR,which takes no institutional positions on economic policy matters,is delighted to p
30、rovide a platform for an exchange of views on this important topic.Tessa OgdenChief Executive Officer,CEPRNovember 20239INTRODUCTION|CAMPANTE,DURANTE AND TESEIIntroductionFilipe R.Campante,a Ruben Durantebce and Andrea TeseideaJohns Hopkins University;bNational University of Singapore;cICREA-Univers
31、itat Pompeu Fabra;dQueen Mary University of London;eCEPRThe advent of the internet and social media represents one of the most important social transformations of our time.Their ubiquitous presence in our daily lives has reshaped the way humans communicate,interact,and coordinate with each other.Thi
32、s has far-reaching consequences for norms and institutions,which are subject to unprecedented pressures to cope with a rapidly changing social and political environment.A well-established body of evidence has shown that traditional media,such as radio and TV,have had a substantial impact on politica
33、l behaviour and outcomes.1 Due to its two-way nature and low barriers to entry into the production and dissemination of content,social media has the potential to stir politics even further.Looking at the public debate on the impact of technology,there seems to be a consensus that the political impli
34、cations of social media are huge.And yet,there is substantial disagreement as to the nature of those implications.Some have argued that platforms such as Facebook helped make a massive blow for democracy and citizen participation,and against the power of autocrats(Ghonim 2012).More recently,however,
35、the view has turned decidedly more negative,with some laying the blame on social media for a host of negative political developments,including the rise of fake news and hate speech(Haidt 2022).This is a call for empirical research,if there has ever been one.Fortunately,social scientists have answere
36、d that call.At this point,in 2023,we can confidently state that a body of literature has emerged that provides credible evidence on many aspects of the impact of social media on politics.While we are far from having all the answers we would like,we now know way more than we did many years ago.This v
37、olume is an attempt to capture a snapshot of that effort.It is by necessity a partial one,given how voluminous the literature has become.It focuses on the political economics of social media,meaning that it compiles contributions from the field of economics.Even within that field,it is meant to be m
38、ore illustrative than comprehensive there is simply too much quality research for any effort of this kind to aspire to be encyclopaedic.Yet it aims to paint a coherent picture of what the internet and social media have wrought to our political landscape and beyond.1 For instance,see Stromberg(2004),
39、Yanagizawa-Drott(2014)and Adena et al(2015)on the impact of the radio,Gentzkow(2006),Prior(2007),Campante and Hojman(2013),and Durante et al.(2019)on broadcast TV,Della et al.(2007)and Martin and Yurukoglu(2017)on cable TV,among others.For a survey,see Prat and Stromberg(2015).10THE POLITICAL ECONOM
40、Y OF SOCIAL MEDIAThat picture shows that social media has had a unique kind of impact,because of specific features that set it apart from previous media technologies.First,while some of these technologies were portable(e.g.radio)and arguably had addictive properties(e.g.television),social media can
41、be carried around and exert its pull on consumers literally everywhere and at all times,via mobile phones.This ubiquity entails a broader scope for impact on everyday life and human welfare.Second,social media is prone to forming homophilic networks through which like-minded,pro-attitudinal content
42、is more likely to spread.This gives rise to a specific type of political impact,with social media being linked to increased polarisation and the diffusion of political content that capitalises on distrust of others,strengthening in-group biases and animosity toward outsiders,as in the case of the re
43、cent rise of populism.Third,and closely related to the previous feature,social media is characterised by uniquely low barriers to entry into the production and dissemination of content political or otherwise.While previous technologies such as radio or TV were largely one-way avenues,with a(relative
44、ly)small number of outlets broadcasting content for a mass audience,social media allows everyone to be a content provider.This makes a huge difference to the possibilities they offer to citizens in terms of organising and coordinating for collective action but also to political entrepreneurs,who can
45、 use social media to spread their own messages,for their own strategic purposes.THE WELFARE EFFECTS OF SOCIAL MEDIAIn the broadest sense,and as with every technology,we care about the impact of social media on wellbeing.That is where this volume starts,in Part 1.As challenging as measuring wellbeing
46、 can be,the literature has come up with ingenious approaches.In Chapter 1,Hunt Allcott,Luca Braghieri,Sarah Eichmeyer and Matthew Gentzkow implement a randomised evaluation incentivising Facebook users to drop off the platform for one month.The outcome is greater offline interaction and,perhaps rela
47、tedly,a small but significant increase in self-reported subjective wellbeing.In fact,the treatment also increased the likelihood of experimental subjects being interested in tools to limit social media usage after the experiment.This highlights the idea that social media has addictive features(Allco
48、tt et al.2022),suggesting the limitations of revealed preference-type arguments whereby the widespread use of social media implies that it has ipso facto net positive effects on users.The evidence of negative effects of social media on individual wellbeing is reinforced by the observational evidence
49、 in Chapter 2,in which Luca Braghieri,Roee Levy and Alexey Makarin exploit the variation induced by the initial spread of Facebook through different college campuses,using a differences-in-differences approach.They show that the introduction of Facebook at a college had a negative impact on students
50、 self-reported mental health,and especially for individuals already susceptible to mental illness.This underscores that the effects of social media may differ across different groups.11INTRODUCTION|CAMPANTE,DURANTE AND TESEIIn sum,social media can have negative effects on subjective wellbeing,and th
51、e fact that people use it cannot be taken by itself as evidence that it increases their welfare.This suggests that there is scope for considering regulation of social media usage and for encouraging content moderation in online platforms.This conclusion,based on the individual consequences of social
52、 media usage,is further strengthened when their aggregate implications are considered.In particular,as with other media technologies,social media could have important implications in the political arena,which obviously distinguishes the media industry from other markets.That is where we turn to next
53、.SOCIAL MEDIA AND vOTERSWe start in the specific realm of individual voter behaviour,in Part 2.In fact,the same work that identifies the negative effects of social media on individual wellbeing already points to possible trade-offs,from a societal perspective,when it comes to political engagement.Af
54、ter all,Chapter 1 indicates that individuals who dropped off Facebook end up less informed about politics,though the experimental variation does not seem enough to affect turnout.It is enough,however,to reduce political polarisation,suggesting that social media may have indeed played a part in incre
55、asing polarisation in recent years,as many have accused it of doing(Sunstein 2017,Haidt 2022).The polarising effect of social media is explored further in Chapter 3.In it,Roee Levy presents another experimental intervention,manipulating not presence on Facebook but rather the content users are expos
56、ed to.The key result is that exposure to social media content indeed increases affective polarisation,that is,the extent to which partisans dislike the other party.Importantly,the experimental intervention randomly assigns pro-versus counter-attitudinal content,and in doing so and by following socia
57、l media usage behaviour by the experimental subjects it is able to adjudicate between different possible mechanisms.In particular,Levy shows that the effect on polarisation works through the fact that social media exposes users to disproportionately pro-attitudinal content,and is able to attribute a
58、 substantial part of that to the algorithms they employ.This substantiates the common intuition that the business models of existing social media platforms may have consequences in terms of the political environment.The interplay between social media and polarisation is also the theme of Chapter 4,i
59、n which Yosh Halberstam and Brian Knight turn to another major social media platform that has been widely accused of fostering polarisation Twitter(now rebranded as X).The authors start with a model that predicts that social media users will tend to be disproportionately exposed to like-minded conte
60、nt,through endogenous homophily in their networks.They then look at the Twitter presence of candidates from the two US major political parties,and the network formed by their followers,which allows them to proxy for the ideology of a subset of Twitter users based on the partisan affiliation of the p
61、oliticians they follow.They find strong homophily:liberals(resp.conservatives)12THE POLITICAL ECONOMY OF SOCIAL MEDIAare disproportionately likely to have liberal(resp.conservative)followers themselves,and much more so than with other types of social interactions.This in turn translates into disprop
62、ortionate consumption of pro-attitudinal information,especially in the political realm.The evidence that social media can indeed create echo chambers that amplify polarisation feeds naturally into the possibility that the content flowing through those chambers could have implications for political o
63、utcomes.This is the theme explored in Chapter 5,in which Sergei Guriev,Nikita Melnikov and Ekaterina Zhuravskaya take a global look at the impact of the spread of social media.They do so by taking advantage of the fact that,at the global scale,the spread was intimately connected to the expansion of
64、3G mobile networks the first technology that made mobile internet access a reality.Using a differences-in-differences approach,their key finding is that increased mobile internet coverage led to reduced levels of trust in government.This seems to be driven by countries where traditional media were n
65、ot free,and where the arrival of social media thereby facilitates access to information on government misconduct.While this is arguably a welcome development of increased transparency,the authors also find that things are different in the European context:there,it is populist politicians who seem to
66、 have benefited from the new technology.This is precisely the theme developed further by Marco Manacorda,Guido Tabellini and Andrea Tesei in Chapter 6.Their key finding is that,in Europe,the expansion of 3G and 4G mobile technology was associated with increased support for communitarian parties,that
67、 is,those advocating for the interests of insiders(e.g.native-born populations,majoritarian groups)against outsiders(e.g.immigrants,minorities).They exploit granular geographical variation in the expansion of the mobile networks,showing that their arrival was followed by increased voting for communi
68、tarian parties.To establish a causal impact,they use the variation in network access induced by proximity to the birthplace of a manager in charge of decisions by telecommunications firms.They argue that the effect is being driven by online content favouring messages that strengthen bias towards in-
69、groups and against out-groups,and provide survey evidence showing voters opinions shifting in that direction.This benefited communitarian parties in general,and the subgroup classified as populist as part of that broader pattern.While the conclusion that social media works to the electoral advantage
70、 of populist politicians seems tempting,a word of caution comes from Chapter 7.In it,Thomas Fujiwara,Karsten Muller and Carlo Schwarz look at the 2016 US election,where many observers partly attributed the unexpected victory of Donald Trump to his social media presence,especially on Twitter.They exp
71、loit the geographical variation in early Twitter adoption generated by attendance at the 2007 South by Southwest(SXSW)festival,which had a strong impact on the platforms spread.They find that exposure to Twitter induced by that variation was negatively associated with Trumps vote share in 2016 and 1
72、3INTRODUCTION|CAMPANTE,DURANTE AND TESEI2020(but not with the Republican congressional vote share).This underscores the power of social media to affect elections,but not always in a consistent direction instead,it can be contingent on specific candidate characteristics.In sum,the evidence strongly p
73、oints in the direction of social media having significant effects on the behaviour of voters.This starts from polarisation,reinforced by the tendency of social media to create homophilic echo chambers disseminating pro-attitudinal content.Social media also spreads political information to voters in
74、general,which can increase transparency,but the selective nature of the content that is spread potentially opens the way for political discontent.2SOCIAL MEDIA AND POLITICIANSHaving gone over the impact of social media on voters,Part 3 then turns to what it means for the behaviour of the supply side
75、 of the political market,namely,politicians.While there is less written on that topic,some evidence has emerged that the ability to strategically use these technologies,given the low barriers to entry,may well advantage politicians,possibly at the expense of voters.Chapter 8,by Filipe Campante,Ruben
76、 Durante and Francesco Sobbrio,illustrates how the introduction of social media changes the possibilities facing political entrepreneurs,in ways that can undo the initial impact of the pre-social media internet.They exploit variation in the expansion of broadband technology in Italy and show that th
77、is expansion initially translated into reduced voter turnout,as supporters of more extreme parties became less likely to vote.However,the arrival of a new political movement,created around the early social media platform M,helped revert that trend:the Five Star Movement expanded more rapidly in plac
78、es with greater broadband access,and attracted more voters in those places when it eventually started competing in elections.In short,the ability to mobilise voters online,at very low cost,offers a possibility for new political actors to enter the fray and reverse the initial demobilising effect of
79、the internet.But what do incumbent politicians do,as they can also take advantage of the new technologies?In Chapter 9,Maria Petrova,Ananya Sen and Pinar Yildirim study the role of social media in helping politicians raise resources.They analyse how contributions to US congressional campaigns change
80、 in response to the candidates opening Twitter accounts,exploiting variation in Twitter penetration across different states.They show that,in the month where a politician opens their account,there is a positive jump in the individual contributions they receive,but only in states with high Twitter pe
81、netration.2 On the importance of the content of the information spread by social media,a particularly prominent instance is that of disinformation,as illustrated by the phenomenon of fake news.Allcott and Gentzkow(2017)and Lazer et al(2018)discuss the reach of this particular problem.14THE POLITICAL
82、 ECONOMY OF SOCIAL MEDIATheir evidence suggests that this happens mostly because of the platforms impact in helping politicians become better known by a previously untapped set of individuals,as opposed to reminding prior contributors to give more.In sum,the early evidence from this literature sugge
83、sts that social media empowers politicians,enabling them to mobilise their supporters in more effective ways,which are available at a very low cost and hence to all politicians rather than just a few.This offers new strategic possibilities,which can empower new actors.Much remains to be learned on w
84、hat these new actors bring to the picture.SOCIAL MEDIA AND PROTESTThe literature has also addressed the impact of social media beyond electoral politics.The role of social media as an effective vehicle for coordination and collective action(Campante et al.2022)naturally makes it a potential facilita
85、tor of costly political activities,perhaps best exemplified by political protest.This is the topic of Part 4.In Chapter 10,Ruben Enikolopov,Alexey Makarin and Maria Petrova take advantage of the idiosyncratic role played by one specific university St Petersburg State University(SPSU)in the creation
86、and expansion of VKontakte(VK),a prominent Russian social media platform.They show that cities with a large number of SPSU students in the cohort to which VKs founder belonged,relative to the numbers in other cohorts,had stronger VK penetration.They were also more likely to have demonstrations(and l
87、arger ones at that)against the government following the 2011 elections,which were widely perceived as manipulated.In contrast,the effect in terms of electoral results(in less disputed elections)seemed to be in favour of the governments party.This indicates that social media played an important role
88、both through the content shared on it and by facilitating coordination for collective action.In Chapter 11,Leopoldo Fergusson and Carlos Molina look at the problem through a broader lens.They study the impact of Facebook across different countries,taking advantage of the fact that versions of the pl
89、atform in different languages were introduced at different points in time.As it turns out,the incidence of protest increased after the introduction of a Facebook platform in a given language in countries with a greater presence of speakers of that language relative to those with fewer speakers.Given
90、 that no similar pattern occurs for the prevalence of different political views,the authors conclude that the mechanism is via the facilitation of coordination towards collective action.The role of coordination is underscored,in a different context,in Chapter 12.Marco Manacorda and Andrea Tesei focu
91、s on 2G mobile phone technology,which does not lend itself to web use but facilitates communication via voice and text.Using detailed georeferenced data on the rollout of 2G technology and on the incidence of protests across Africa over 15 years,they show that mobile phones were instrumental to mass
92、 political mobilisation across local areas within countries,but only during periods of economic 15INTRODUCTION|CAMPANTE,DURANTE AND TESEIdownturn,when reasons for grievance emerge and the cost of participation falls.Their results highlight the interplay between the online and offline worlds,suggesti
93、ng that communication technologies can act as a tool for coordinating pre-existing offline grievances,leading individuals to mobilise and amplifying their discontent.The evidence thus suggests that the low barriers to entry and the effectiveness of social media in allowing for communication do trans
94、late into greater mobilisation for collective action by citizens.In conjunction with its effect on electoral politics,it becomes clear that the political impact of social media is deep and wide-ranging:it can empower all sorts of different political actors to pursue many sorts of political activitie
95、s.SOCIAL MEDIA AND HATREDOne particularly important kind of political action that one may be concerned about due to its potential negative impact is what we may call hatred:collective action directed against specific groups.This is the focus of Part 5.Chapter 13,by Leonardo Bursztyn,Georgy Egorov,Ru
96、ben Enikolopov and Maria Petrova,looks once again at the Russian context and investigates the impact of social media on the prevalence of xenophobic attitudes and ethnic hate crimes.Using the same variation exploited in Chapter 10,they show that the expansion of the VK social media platform increase
97、d the share of individuals holding extreme xenophobic views(based on an online survey experiment)and the occurrence of hate crimes against ethnic minorities.Moreover,they show that the two are connected:the increase in crimes is especially strong in cities with higher pre-existing levels of social m
98、edia and for crimes with multiple perpetrators.This underscores the dark side of social medias ability to facilitate collective action,by allowing like-minded individuals to find one another and to coordinate towards action.Important recent episodes in Western Europe and in the United States are the
99、 focus of Chapter 14,by Karsten Muller and Carlo Schwarz.They show first that the large influx of refugees into Germany in 2015,as a result of the civil war in Syria,was associated with an increase in anti-refugee rhetoric in social media by the far-right Alternative fur Deutschland(AfD)party and al
100、so in violent attacks against refugees.Using variation in internet and Facebook outages over time and across German towns,they show that the correlation between online hatred and offline hate crimes weakens when the internet or Facebook goes offline.Similarly,anti-Muslim sentiment and hate crimes in
101、creased in the United States,especially after Donald Trump initiated his presidential campaign,and those increases were stronger in places with greater Twitter penetration.Using the variation induced by the SXSW festival,discussed in Chapter 7,they argue that Twitter had a causal impact,mediated by
102、Trumps tweeting.Importantly,the effect was concentrated in places with substantial pre-existing activity by hate groups.16THE POLITICAL ECONOMY OF SOCIAL MEDIAThe evidence thus suggests that offline and online activities are also inextricably linked by social media when it comes to hatred against sp
103、ecific groups.These technologies facilitate the spread of hateful content as well as coordination around translating animosity into action.SOCIAL MEDIA IN AUTOCRACIESThe role of social media in autocracies has been another prominent topic of study.The case of China has been of special interest,in li
104、ght not only of the countrys size and importance,but also of the well-documented efforts by the Chinese regime in trying to harness new media and communications technologies towards its own goals.This is the focus of Part 6.In Chapter 15,Bei Qin,David Strmberg and Yanhui Wu sketch out the social med
105、ia landscape in China.They start by estimating the Chinese governments social media presence,looking at the Sina Weibo platform,and show that it is much more extensive than commonly assessed,with some 600,000 government-affiliated Weibo accounts,with posting particularly concentrated on politically
106、sensitive topics.In addition,they find evidence that the government uses social media to monitor local officials,as well as to exert surveillance over collective action by citizens.Yet the spread of information on such action can take place over social media,which highlights a key tension facing an
107、autocratic government such as that of China:the value of information versus the risk of collective action against the regime.This tension is further exemplified in Chapter 16,by David Yang.The chapter reports on a field experiment conducted in China randomly assigning a group of university students
108、to free access to a virtual proxy network(VPN)tool,which allows access to internet content from outside China,thereby bypassing censorship.The key finding is that access to uncensored content has little impact on actual acquisition of politically sensitive information.On the other hand,additional in
109、centives to subscribe to a Western news outlet did increase that acquisition,and furthermore,this brought important and persistent changes to knowledge,beliefs and(intended)behaviour,in ways that run against the Chinese government.This pattern suggests that,while censorship may have managed to creat
110、e an apathetic environment low demand for politically sensitive information,this does not necessarily imply fear.Sufficient incentives towards the acquisition of information could actually lead to an increased propensity for political action.In sum,the role of social media in an autocratic regime on
111、ce again highlights how its effects can go in different directions,as it can empower citizens but also governments.17INTRODUCTION|CAMPANTE,DURANTE AND TESEISOCIAL MEDIA AND LEGACY MEDIALast but not least,understanding the broader impact of social media requires not only considering their direct effe
112、cts,but also how they shape and affect other media.After all,social media never exist in isolation,but as a part of a broader ecosystem.As such,Part 7 looks at the impact on so-called“legacy”media.Chapter 17,by Sophie Hatte and Ekaterina Zhuravskaya,focuses on the interplay between social media and
113、TV news coverage in the context of the Israeli-Palestinian conflict.Once again,the low barriers to entry in the dissemination of content comes to the forefront,in the example of citizen-journalists:social media content documenting citizens experience of conflict can affect how the conflict gets cove
114、red on TV.Using variation in social media access driven by internet outages in Israel and Palestine,they show that more citizen-generated content leads to more TV coverage of the conflict.What is more,it changes the tone of that coverage,bringing greater emotional intensity and focus on the impact o
115、n civilians.As a result of that influence,coverage becomes more similar across different TV news channels.The direct influence of social media content on news coverage is also the topic of Chapter 18,by Julia Cag,Nicolas Herv and Batrice Mazoyer.Using a large sample of French-language tweets,they id
116、entify instances of events covered in both social and legacy media.They then rely on variation in the impact of Twitter content that is driven by the network centrality of users,as opposed to the intrinsic newsworthiness of the event they address,interacted with variation in the presence of competin
117、g stories at the time of the event.They find that an increase in the number of tweets about an event has a causal impact on the news coverage by legacy media,which is particularly strong for legacy media outlets that rely more on advertising revenues.The findings also suggest that this may bias trad
118、itional news coverage,since Twitter users are not a representative sample of the broader audience,and Twitter-pushed stories do not seem to generate greater interest among readers.The evidence clearly indicates that social media influences the behaviour of legacy media,both in terms of the issues th
119、ey cover and how they cover them.This suggests that the impact of social media can go well beyond the direct effect that is often more easily measured and assessed.WHAT TO MAKE OF THIS?The unique features of social media,compared to pre-existing media technologies its pervasiveness,the kind of conte
120、nt it favours,the low barriers to entry,and the multi-way mass communication it facilitates imply that its effects on political life can be complex and nuanced.Social media can empower voters and politicians,citizens and autocrats,increase participation or decrease it,bring people together or push t
121、hem apart.Far from 18THE POLITICAL ECONOMY OF SOCIAL MEDIAsimple answers regarding whether social media is good or bad,the work compiled in this volume establishes that the implications can push in different directions,depending on different outcomes and contexts.One thing that remains a crucial,ope
122、n topic for future research is what to do about social media from a policy point of view.How can we harness its potential and minimise its harm?In contrast with many addictive goods that bring little social upside beyond individual enjoyment and whose consumption,for the most part,we would therefore
123、 like to discourage social media can also have positive effects that are not fully internalised by individuals.This makes it difficult to figure out what optimal regulatory policy should be,and how to achieve it.Some possible directions are already apparent.To give but one example,the evidence in Ch
124、apter 3 indicates that the algorithms employed by social media companies play an important role in the impact of social media usage on polarisation.Regulating them,or changing the incentives that underpin their design,seems worthy of consideration by policymakers.By the same token,the chapters in Pa
125、rt 7 underscore that social media strongly conditions what happens on traditional media platforms,and this interplay should affect how policy approaches the regulation of those platforms.In any case,more research is needed to guide policy.Yet there are substantial challenges to pushing research in t
126、his direction.For one,social media has become ubiquitous,and the kind of exogenous variation in access exploited by some of the contributions covered here the spread of broadband or 3G access,or idiosyncrasies in the initial expansion of certain platforms is,by now,hard to come by.The experimental a
127、pproach that other contributions have used is a natural alternative,but they generally preclude the kind of large-scale,general-equilibrium impact that is often the object of policy interest.Yet the importance of the questions associated with these issues is bound to keep drawing intense attention f
128、rom researchers going forward.One recent example,published too late to be included in this book,is the series of collaborations between teams of social scientists and social media companies for large-scale experiments(Gonzlez-Bailn et al.2023,Guess et al.2023a,2023b,Nyhan et al.2023).The results she
129、d light on the power and limitations of algorithms and user features,the degree of segregation in news consumption,and the limits of its effects on polarisation.At the same time,they are bound to raise further questions that will continue to stimulate research going forward.Another important avenue
130、to explore,which much of the work here exemplifies,is considering social media in its role as an arena for political action,which means that social media provides a rich source of data for answering political economy questions beyond those related to the impact of the technology itself.The continual
131、 improvement of language processing tools possibly enhanced by artificial intelligence will ensure that this will be a vibrant area of research going forward.19INTRODUCTION|CAMPANTE,DURANTE AND TESEITowards that goal,one possible type of policy intervention would be to facilitate access to data for
132、research purposes.Data are both a key by-product of social media usage and a resource that is crucial to the business models of the companies operating in this space.While companies have an interest in allowing access to their data for research whose results might be of use to them,they also have a
133、strong incentive to try and control the dissemination of those results.There seems to be scope for policy to induce companies to increase no strings attached opportunities for data access.What seems certain,however,is that social media will endure as a key force in society,and in politics in particu
134、lar.We will keep studying it also,to paraphrase Sir Edmund Hillary,because its there.REFERENCESAdena,M,R Enikolopov,M Petrova,V Santarosa and E Zhuravskaya(2015),“Radio and the Rise of the Nazis in Pre-war Germany”,The Quarterly Journal of Economics 130(4):1885-1939.Allcott,H and M Gentzkow(2017),“S
135、ocial Media and Fake News in the 2016 election”,Journal of Economic Perspectives 31(2):211-236.Allcott,H,M Gentzkow,and L Song(2022),“Digital Addiction”,American Economic Review 112(7):2424-2463.Campante,F,R Durante,and A Tesei(2022),“Media and Social Capital”,Annual Review of Economics 14:69-91.Cam
136、pante,F and D A Hojman(2013),“Media and Polarization:Evidence from the Introduction of Broadcast TV in the United States”,Journal of Public Economics 100:79-92.Della Vigna,S and E Kaplan(2007),“The Political Impact of Media Bias”,mimeo,UC Berkeley and Stokholm University.Durante,R,P Pinotti,and A Te
137、sei(2019),“The Political Legacy of Entertainment TV”,American Economic Review 109(7):2497-2530.Gentzkow,M(2006),“Television and Voter Turnout”,The Quarterly Journal of Economics 121(3):931-972.Ghonim,W(2012),Revolution 2.0:The Power of the People is Greater than the People in Power:A Memoir,Houghton
138、 Mifflin Harcourt.Gonzlez-Bailn,D,D Lazer,P Barber et al.(2023),“Asymmetric Ideological Segregation in Exposure to Political News on Facebook”,Science 381.6656,392-398.Guess,A M,N Malhotra,J Pan et al.(2023),“How do Social Media Feed Algorithms affect Attitudes and Behavior in an Election Campaign?”
139、,Science 381(6656):398-404.20THE POLITICAL ECONOMY OF SOCIAL MEDIAGuess,A M,N Malhotra,J Pan et al.(2023),“Reshares on Social Media amplify Political News but do not detectably affect Beliefs or Opinions”,Science 381(6656):404-408.Haidt,J(2022),“Why the Past 10 years of American life have been uniqu
140、ely stupid”,The Atlantic,11 April.Lazer,D M J,M A Baum,Y Benkler et al.(2018),“The Science of Fake News”,Science 359(6380):1094-1096.Martin,G J and A Yurukoglu(2017),“Bias in Cable News:Persuasion and Polarization”,American Economic Review 107(9):2565-2599.Nyhan,Brendan,J Settle,E Thorson et al.(202
141、3),“Like-minded Sources on Facebook are Prevalent but not Polarizing”,Nature 620(7972):137-144.Prat,A,and D Strmberg(2013),“The Political Economy of Mass Media”,Advances in Economics and Econometrics 2:135-187.Prior,M(2007),Post-broadcast Democracy:How Media Choice Increases Inequality in Political
142、Involvement and Polarizes Elections,Cambridge University Press.Strmberg,D(2004),“Mass Media Competition,Political Competition,and Public Policy”,The Review of Economic Studies 71(1):265-284.Sunstein,C R,(2017),“A Prison of our own Design:Divided Democracy in the Age of Social Media”,Democratic Audit
143、 UK.Yanagizawa-Drott,D(2014),“Propaganda and Conflict:Evidence from the Rwandan Genocide”,The Quarterly Journal of Economics 129(4):1947-1994.ABOUT THE AUTHORSFilipe Campante is a Bloomberg Distinguished Professor of International Economics at Johns Hopkins University,and a Research Associate at the
144、 National Bureau of Economic Research.He is interested in political economy,development economics,and urban/regional issues.His research looks at what constrains politicians and policy makers beyond formal checks and balances:media,cultural norms,institutions,political protest.His work has appeared
145、in leading academic journals,and received numerous mentions in media outlets all over the world.Born and raised in Rio de Janeiro,Brazil,he holds a PhD from Harvard University,an MA from the Pontifical Catholic University of Rio de Janeiro,and a BA from the Federal University of Rio de Janeiro,all i
146、n economics.Ruben Durante is Professor of Economics at the National University of Singapore and ICREA researcher at Pompeu Fabra University.He is also affiliated faculty to the Barcelona School of Economics and IPEG,Research Advisor at Gallup US,and Research Fellow of CEPR,CESifo,and IZA.He has held
147、 visiting positions at Yale University,PSE,and INSEAD.He studies political economy with a focus on the functioning and impact 21INTRODUCTION|CAMPANTE,DURANTE AND TESEIof traditional and new media in democratic societies.His work has been published in top journals in economics,political science,and m
148、anagement.His research has been supported by several funding agencies including the European Research Council through a Starting Grant,a Consolidator Grant,and a Proof of Concept Grant.He holds a BA in Economics from the University of Messina,a Masters in Political Economy from Sorbonne University,a
149、nd a Masters and a PhD in Economics from Brown University.Andrea Tesei is Associate Professor of Economics at Queen Mary University of London.He is also a Research Affiliate of CEPR,a Research Associate of the CEP at the London School of Economics and a Research Fellow of CESifo.He has been a visiti
150、ng scholar at Northwestern University.His main research interests are in political economy and development economics,with a focus on the political,economic and social impact of traditional and new media in both developed and developing countries.His research has been published in leading scientific
151、journals,including the American Economic Review,Econometrica and the Review of Economics and Statistics.Andrea obtained his PhD from Universitat Pompeu Fabra in 2012.SECTION 1 WELFARE EFFECTS OF SOCIAL MEDIA25THE WELFARE EFFECTS OF SOCIAL MEDIA|ALLCOTT,BRAGHIERI,EICHMEYER AND GENTZKOWCHAPTER 1The we
152、lfare effects of social mediaHunt Allcott,a Luca Braghieri,b Sarah Eichmeyerb and Matthew GentzkowaaStanford University;bBocconi UniversityIn the last decade,social media has woven its way deep into our lives.Facebook has 2.3 billion monthly active users,and by 2016 the average user was spending nea
153、rly an hour per day on it and its sister platforms.There may be no technology since television that has so dramatically reshaped the way we get information and spend our time.Early on,platforms like Facebook,Twitter and Instagram were hailed for their potential to make communication and the sharing
154、of information easier.Now,the conversation is dominated by potential harms,from addiction to depression to political polarisation.Despite the abundance of speculation about the potential effects of social media,hard evidence remains scarce.In a recent paper,we provide a large-scale randomised evalua
155、tion of the welfare impacts of Facebook,the largest social media platform(Allcott et al.2020).This provided the largest-scale experimental evidence to date on Facebooks impact on a range of outcomes.We find that deactivating Facebook for one month leads people to spend more time with friends and fam
156、ily.It also leaves them less informed about the news,less polarised in their political opinions,and a little happier and more satisfied with their lives.We find that after the time off Facebook,users want it back,but they use it significantly less than before.Our findings are in line with other impo
157、rtant work on the same topic(e.g.Mosquera et al.2020,Muller and Schwarz 2021,Braghieri et al.2022).STUDY DESIGNWe recruited 1,600 US Facebook users online and randomised them into a deactivation(or treatment)group and a control group.The deactivation group received US$102 in exchange for staying off
158、 Facebook for the four weeks leading up to the US midterm election in November 2018;the control group kept using Facebook as usual.We measured a suite of outcomes using text messages,surveys,emails and administrative voting records.We recorded key measures twice once in October,before the beginning
159、of the deactivation period(baseline);and once in November,after the deactivation period had concluded(endline).We then compared the changes in those outcomes in the deactivation group to those in the control group.Our surveys had very high response 26THE POLITICAL ECONOMY OF SOCIAL MEDIArates:of the
160、 580 people in the deactivation group,only seven failed to complete the endline survey.Of the 1,081 people in the control group,only 17 failed to complete the endline.To verify deactivation,we repeatedly pinged the URLs of participants public Facebook profiles.While a user can limit how much content
161、 other people can see in their profiles,they cannot hide their public profile page.The public profile page returns a valid page when an account is active but returns an error message when an account is deactivated.Overall,90%of users in the deactivation group followed our instructions and deactivate
162、d their accounts.For our impact evaluation,we estimate the local average treatment effect of deactivation.That is,we use the treatment indicator to instrument for the percentage of deactivation checks in which a person is observed to be deactivated.KEY FINDINGSBeing off Facebook freed up an average
163、of one hour to spend on other activities.How people use this extra time helps us understand which activities Facebook is crowding out,and this in turn tells us something about Facebooks effects.If Facebook time just replaces other social media or similar digital activities,the effects of deactivatio
164、n might be small.If it replaces high-quality social interactions with family and friends,we might worry more about outcomes like(un)happiness,loneliness and depression.If it replaces consumption of high-quality news,we might worry more about impacts on political knowledge and polarisation.Our survey
165、s show that Facebook does not substitute for other digital activities if anything,people reported spending less time on other social media and digital platforms while their Facebook accounts were deactivated.The deactivation group reported spending more time on offline activities,including face-to-f
166、ace socialising and solitary activities like watching TV.Our next set of findings focuses on news knowledge and political outcomes.Deactivating Facebook caused a significant reduction in total news consumption and news knowledge.Among other things,we find that those in the deactivation group were si
167、gnificantly worse at answering quiz questions about current issues in the news.At the same time,the deactivation group ended up significantly less polarised in a range of measures,including their views on policy issues such as immigration and policing(Figure 1).Our overall index of political polaris
168、ation fell by 0.16 standard deviations.As a point of comparison,prior work has found that a different index of political polarisation rose by 0.38 standard deviations between 1996 and 2018(Boxell 2020).There is no detectable effect on political engagement,as measured by voter turnout in the midterm
169、election.27THE WELFARE EFFECTS OF SOCIAL MEDIA|ALLCOTT,BRAGHIERI,EICHMEYER AND GENTZKOWFIGURE 1 EFFECTS ON POLARISATION0.2.4.6Density-4-2024Issue opinions(in units of Control group standard deviations)Treatment DemocratTreatment RepublicanControl DemocratControl Republicankernel=epanechnikov,bandwid
170、th=0.2231Note:The dashed lines show the distribution of these views among control-group Democrats(blue)and Republicans(red).The solid lines represent the deactivation group.In both groups Democrats views are well to the left of Republicans views,but the inter-party differences are visibly smaller in
171、 the deactivation group,suggesting that deactivation moderated views in both parties.In terms of wellbeing,we find that Facebook deactivation causes small but significant increases in self-reported individual life satisfaction and happiness,and significant decreases in self-reported levels of anxiet
172、y.We also elicited self-reported wellbeing using daily text messages,and find positive but statistically insignificant effects of Facebook deactivation on this outcome.As shown in Figure 2,an index of all measures together shows that deactivation caused significant improvements in overall wellbeing,
173、with the overall index improving by 0.09 standard deviations.As a point of comparison,this is about 2540%of the effect of psychological interventions including self-help therapy,group training and individual therapy,as reported in a meta-analysis by Bolier et al.(2013).These results are consistent w
174、ith a recent quasi-experimental study finding that Facebook may have adverse effects on mental health(Braghieri et al.2022).We find little evidence to support the hypothesis suggested by prior work that Facebook might be more beneficial for active users for example,users who regularly comment on pic
175、tures and posts from friends and family instead of just scrolling through their news feeds.28THE POLITICAL ECONOMY OF SOCIAL MEDIAFIGURE 2 EFFECTS ON SUBJECTIvE WELLBEINGHappinessLife satisfactionLoneliness (-1)Depressed (-1)Anxious (-1)AbsorbedBored (-1)SMS happinessSMS positive emotionSMS not lone
176、lySubjective well-being index-.10.1.2Treatment effect(standard deviations)Note:Each point in the figure measures the effect of Facebook deactivation on wellbeing outcomes,measured in standard deviations.The lines to the left and right of each point indicate the 95%confidence interval.All outcomes ar
177、e scaled so that the right of the figure indicates more positive outcomes.(Thus,measures of loneliness,depression,anxiety,and boredom are inverted.)The final row shows an index of all measures together,showing that deactivation caused significant improvements in overall well-being.Finally,we measure
178、d whether deactivation affected peoples demand for Facebook after the study was over,as well as their opinions about Facebooks role in society.As the experiment ended,participants assigned to the deactivation group reported planning to use Facebook much less in the future.Several weeks later,the dea
179、ctivation groups reported usage of the Facebook mobile app was about 11 minutes(or 22%)lower than in control.In line with these self-reported measures,we found that 5%of the deactivation group still had their accounts deactivated nine weeks after the experiment ended.The deactivation group was also
180、more likely to click on a post-experiment email providing information about tools to limit social media usage.Reduced post-experiment use aligns with our finding that deactivation improved subjective well-being,and it is also consistent with the hypotheses that Facebook is habit forming in the sense
181、 of Becker and Murphy(1988)or that people learned that they enjoy life without Facebook more than they had anticipated.A recent field experimental study on digital addiction(Allcott et al.2022)supports this notion,suggesting that self-control problems may cause 31%of social media use.29THE WELFARE E
182、FFECTS OF SOCIAL MEDIA|ALLCOTT,BRAGHIERI,EICHMEYER AND GENTZKOWBIG PICTUREThere is no doubt that many users perceive large benefits from Facebook.A majority of participants would require a payment of$100 or more to deactivate Facebook for a month.Even after four weeks of deactivation,these valuation
183、s remained high and our participants continued to spend substantial time on Facebook every day.The results on news consumption and knowledge suggest that Facebook is an important source of news and information.Our participants answers in free-response questions and follow-up interviews make clear th
184、e diverse ways in which Facebook can improve peoples lives,whether as a source of entertainment,a way to organise a charity or an activist group,or a vital social lifeline for those who are otherwise isolated.Any discussion of social medias downsides should not obscure the basic fact that it fulfill
185、s deep and widespread needs.At the same time,our results also make clear that the downsides are real.We find that four weeks without Facebook improves subjective wellbeing and substantially reduces post-experiment demand,suggesting that forces such as addiction may cause people to use Facebook more
186、than they otherwise would.We find that while deactivation makes people less informed,it also makes them less polarised,consistent with the concern that social media have played some role in the recent rise of polarisation in the US.The trajectory of views on social media with early optimism about gr
187、eat benefits giving way to alarm about possible harms is a familiar one.Innovations from novels to TV to nuclear energy have had similar trajectories.Along with the important existing work by other researchers,we hope that our analysis can help move the discussion from simplistic caricatures to hard
188、 evidence,and provide a sober assessment of the ways a new technology affects both individual people and larger social institutions.REFERENCESAllcott,H,L Braghieri,S Eichmeyer and M Gentzkow(2020),“The Welfare Effects of Social Media”,American Economic Review 110(3):629-676.Allcott,H,M Gentzkow and
189、L Song(2022),“Digital Addiction”,American Economic Review 112(7):2424-2463.Becker,G and K Murphy(1988),“A Theory of Rational Addiction”,Journal of Political Economy 96(4):675-700.Bolier,L,M Haverman,G Westerhof,H Riper,F Smit and E Bohlmeijer(2013),“Positive Psychology Interventions:A Meta-Analysis
190、of Randomized Controlled Studies”,BMC Public Health 13(119).Boxell,L(2020),“Demographic change and political polarization in the United States”,Economics Letters 192.30THE POLITICAL ECONOMY OF SOCIAL MEDIABraghieri,L,R Levy and A Makarin(2022),“Social Media and Mental Health”,American Economic Revie
191、w 112(11):3660-3693.Mosquera,R,M Odunowo,T McNamara,X Guo and R Petrie(2020),“The economic effects of Facebook”,Experimental Economics 23:575-602.Muller,K and C Schwarz(2021),“Fanning the Flames of Hate:Social Media and Hate Crime”,Journal of the European Economic Association 19(4):21312167.ABOUT TH
192、E AUTHORSHunt Allcott is a Professor of Global Environmental Policy at Stanford University.He is the co-director of the Stanford Environmental and Energy Policy Analysis Center,a research associate at the National Bureau of Economic Research,an affiliate of ideas42 and Poverty Action Lab,and a membe
193、r of the board of editors of the American Economic Journal:Economic Policy.Luca Braghieri is an Assistant Professor in the Department of Decision Sciences at Bocconi University.He holds a PhD in economics from Stanford University and a bachelors degree from Harvard College.Most of his work is on top
194、ics related to applied microeconomics,behavioral economics,and political economy.Sarah Eichmeyer is an Assistant Professor of Economics at Bocconi University.Her research spans the areas of public economics and political economy.She holds a PhD in economics from Stanford University,and undergraduate
195、 and masters degrees from the University of Heidelberg and the University of Zurich,respectively.Matthew Gentzkow is the Landau Professor of Technology and the Economy at Stanford University.He studies applied microeconomics with a focus on media industries.He received the 2014 John Bates Clark Meda
196、l,given by the American Economic Association to the American economist under the age of forty who has made the most significant contribution to economic thought and knowledge.He is a member of the National Academy of Sciences,a fellow of the American Academy of Arts and Sciences and the Econometric
197、Society,and a senior fellow at the Stanford Institute for Economic Policy Research.31SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINCHAPTER 2Social media and mental healthLuca Braghieri,Roee Levy and Alexey MakarinBocconi University;Tel Aviv University and CEPR;MIT Sloan School of Manageme
198、ntOver the last two decades,the mental health of adolescents and young adults in many countries has worsened considerably(Twenge et al.2019).Data from the US shows that the fraction of individuals aged 1823 who reported experiencing a major depressive episode in the past year almost doubled between
199、2008 and 2018(NSDUH 2019).Similarly,over the same time period,suicides became more prevalent in the US and are now the second leading cause of death for individuals aged 1524 years old(National Center for Health Statistics 2021).Since the increased prevalence of mental illness among adolescents and
200、young adults coincided with the diffusion of social media,researchers,journalists and policymakers alike began to wonder whether the two phenomena might be related(Twenge and Campbell 2019).In the autumn of 2021,a series of articles in the Wall Street Journal alleging that Meta(previously Facebook)w
201、as aware that Instagram had a negative effect on teenage girls body image brought the relationship between social media and mental health to the forefront of public debate(Wells et al.2021).Soon after,the US Congress held a committee hearing on the topic.Despite the urgent need for studies on whethe
202、r social media is detrimental to mental health,causal evidence remains scarce.Most existing papers estimate correlations between social media use and mental health(Bekalu et al.2019,Berryman et al.2018,Dienlin et al.2017,Kelly et al.2018;Lin et al.2016,Twenge and Campbell 2019).A few experiments inc
203、entivise randomly selected participants to reduce their social media use and hence do estimate causal effects,but they do not concentrate primarily on mental health(Allcott et al.2020,2021,Mosquera et al.2020).1 In a recent paper(Braghieri et al.2022),we provide the most comprehensive causal evidenc
204、e to date on the effects of social media on mental health by leveraging a unique natural experiment:the staggered roll-out of Facebook across US college campuses.Our empirical strategy allows us to estimate the short-to medium-run effects of Facebook on a rich set of mental health outcomes ranging f
205、rom depression,to generalised anxiety 1 A recent paper focuses on the causal effect of the internet rather than social media on mental health,and finds that access to high-speed internet increased incidence of mental disorders among young adults in Italy(Donati et al.2022).32THE POLITICAL ECONOMY OF
206、 SOCIAL MEDIAdisorder,to anorexia.2 Overall,we find that the introduction of Facebook at a college had a negative effect on student mental health,especially as far as depression and generalized anxiety disorder are concerned.ESTIMATING THE CAUSAL EFFECT OF FACEBOOKOur research design leverages Faceb
207、ooks gradual expansion across US colleges as a natural experiment.Facebook was created by Harvard undergraduate Mark Zuckerberg in February 2004.Initially,access to the platform was limited to Harvard students.Over the subsequent two and a half years,Facebook gradually expanded to other colleges in
208、the US and abroad until eventually,in September 2006,it opened its doors to everyone in the world above the age of 13.The staggered nature of Facebooks roll-out allows us to compare changes in student mental health in colleges that just received Facebook access to changes in student mental health in
209、 colleges still without Facebook access in a difference-in-differences approach.Although we study the expansion of a new technology,our study is not limited to a small subset of early adopters.When Facebook became available,colleges witnessed rapid and widespread adoption among students.Based on dat
210、a provided by Facebook,we estimate that,in September 2005,approximately 85%of undergraduate students in colleges with access to Facebook had an account.Not only did Facebook spread rapidly and widely in the student population;usage was also intense.In early 2006,close to three-quarters of users logg
211、ed into the site at least once a day,and the average user logged in six times a day(Hass 2006).To estimate Facebooks effects on mental health,we rely on two datasets:one contains the dates in which Facebook was introduced at 775 US colleges;the other contains individual-level survey data about stude
212、nt mental health from the National College Health Assessment(NCHA).The database containing the Facebook introduction dates was constructed as follows.For the first 100 colleges that received access to Facebook,we relied on the introduction dates collected and made public in previous studies(Jacobs e
213、t al.2015,Traud et al.2012).For the remaining colleges,we collected introduction dates using the Wayback Machine,an online archive that contains snapshots of various websites at different points in time.3 Our outcome variables come from the National College Health Assessment(NCHA),the largest and mo
214、st comprehensive dataset on 2 The paper relates to an emerging literature,some of it featured in this CEPR eBook,exploiting the expansion of social media platforms to study the effects of social media on a variety of outcomes.The empirical strategy adopted in this paper is closely related to the one
215、 in Armona(2019),who leverages the staggered introduction of Facebook across U.S.colleges to study labor market outcomes more than a decade later.Enikolopov et al.(2020)and Fergusson and Molina(2020)exploit the expansion of the social media platform VK in Russia and of Facebook worldwide,respectivel
216、y,to show that social media use increases protest participation.Bursztyn et al.(2019)and Mller and Schwarz(2020)exploit the expansion of VK and Twitter,respectively,and find that social media use increases the prevalence of hate crimes.Additional research on social media and political outcomes inclu
217、des Enikolopov et al.(2018),Fujiwara et al.(2021),and Levy(2021).For a detailed overview,see Zhuravskaya et al.(2020).A unique feature of our setting is that it allows us to measure the effects of the sharp roll-out of the biggest social media platform in the world at a time in which very few close
218、substitutes were available.3 We thank Luis Armona for his collaboration in putting together the dataset containing Facebook expansion dates.33SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINthe mental health of US college students available at the time of Facebooks expansion(Leshner and Sch
219、erer 2021).We have access to the universe of responses to all NCHA survey waves administered between the spring of 2000 and the spring of 2008,the longest stretch of time around Facebooks early expansion in which the survey questionnaire did not vary.In order to allow us to carry out our analysis,th
220、e organisation administering the NCHA the American College Health Association generously provided us with a customised dataset that contained,together with the students answers to the NCHA survey,a variable specifying the semester in which Facebook became available at the college attended by the sur
221、vey respondent.4 Our main outcome variable is an index of poor mental health constructed by taking an equally weighted average of all the mental health questions in the NCHA survey inquiring about a respondents recent past.We also analyse the effect of Facebook on three sub-indices:an index of quest
222、ions about depression-related symptoms;an index of questions about other mental health conditions;and an index of questions about depression-related services,such as taking anti-depressants.Mental health is a domain where self-reported outcomes are especially useful and self-reported symptoms are pa
223、rt of standard medical practice(Chan 2010).Still,the survey questions we analyse are not necessarily the questions medical professionals use in practice.Therefore,to validate that the NCHA survey questions measure mental health accurately,we conducted an original survey among more than 500 college s
224、tudents containing both the NCHA questions and the questions from canonical depression and generalised anxiety disorder screeners the PHQ-9 and GAD-7,respectively known to be highly predictive of actual medical mental illness diagnoses(Kroenke et al.2001,Spitzer et al.2006).We find that our index of
225、 poor mental health based on the NCHA questions is strongly correlated with the PHQ-9 and GAD-7 scores(correlation coefficients of 0.66 and 0.61 respectively),increasing our confidence that the NCHA survey is picking up the elements that feature into mental illness diagnoses.4 For privacy reasons,ou
226、r dataset does not contain college identifiers.34THE POLITICAL ECONOMY OF SOCIAL MEDIATHE EFFECT OF FACEBOOK ON MENTAL HEALTHFigure 1 presents the causal difference-in-differences estimates of the impact of Facebook on mental health outcomes.5FIGURE 1 EFFECTS OF THE INTRODUCTION OF FACEBOOK ON STUDE
227、NT MENTAL HEALTHDepressionSymptomsOtherSymptomsDepressionServices Last year felt hopelessLast year felt overwhelmedLast year felt exhaustedLast year felt very sadLast year severely depressedLast year seriously considered suicideLast year attempted suicideLast year depressionIndex Symptoms Depression
228、Last year anorexiaLast year anxiety disorderLast year bulimiaLast year seasonal affect disorderIndex Symptoms Other ConditionsLast year depression diagnosisTherapy depressionCurrent medication depressionIndex Depression ServicesIndex Poor Mental Health-.15-.1-.050.05.1.15Treatment effect(standard de
229、viations)The results show that the introduction of Facebook at a college had a negative impact on student mental health.The effect size on the index of poor mental health is 0.085 standard deviation units.This corresponds to approximately 84%of the difference in the index of poor mental health betwe
230、en students in our sample with and without credit card debt.As an alternative point of comparison,the impact of introducing Facebook at a college is around 22%of the causal effect of a sudden unemployment spell(Paul and Moser 2009).5 Specifically,we estimate the following two-way fixed effects(TWFE)
231、model:Yicgt=c+t+Facebookgt+Xi +Xc +icgt,(1)where Yicgt represents an outcome for individual i who participated in survey wave t and attends college c that belongs to expansion group g;c indicates college fixed effects;t indicates survey-wave fixed effects;Facebookgt is an indicator for whether,in su
232、rvey wave t,Facebook was available at colleges in expansion group g;Xi and Xc are vectors of individual-level and college-level controls,respectively.We estimate Equation(1)using OLS and cluster standard errors at the college level.We also address recent econometric concerns with staggered differenc
233、e-in-differences research designs by showing robustness to the use of a variety of alternative estimators(Borusyak et al.2021,Callaway and SantAnna 2021,De Chaisemartin and dHaultfoeuille 2020,Sun and Abraham 2021).35SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINThe effects we find are st
234、rongest for depression and anxiety disorder.College-wide access to Facebook increased the number of students who reported experiencing severe depression or generalised anxiety disorder in the last year by 7%and 20%,respectively.Figure 1 shows that these estimates correspond to effects of around 0.07
235、0.08 in standard deviation units.This effect on severe depression is similar in magnitude to the effect observed in an experiment conducted by Allcott et al.(2020).Such similarity is striking,especially in light of the fact that the time period,survey question,target population,and empirical strateg
236、y in Allcott et al.(2020)are different from the ones in our paper.In contrast to depression and anxiety,we do not find significant effects on self-reports of anorexia and bulimia.When estimating the effect of Facebook on mental health over time using an event-study regression,we find evidence that t
237、he effect increases as colleges are exposed to Facebook for more semesters(see Figure 2).Importantly for our empirical strategy,we do not find any significant effects or trends before Facebook is introduced at a college.The lack of pre-trends assuages potential concerns about our effect being driven
238、 by differential trends in mental health between colleges that received access to Facebook relatively early and colleges that received access relatively late.FIGURE 2 EFFECTS OF FACEBOOK ON THE INDEX OF POOR MENTAL HEALTH BASED ON DISTANCE TO/FROM FACEBOOK INTRODUCTION-.20.2.4.6Coefficient-8-7-6-5-4
239、-3-2-1012Semester to/from FB IntroductionNote:Constructed using Sun and Abrahams(2021)estimator.36THE POLITICAL ECONOMY OF SOCIAL MEDIAIn order to study whether the effects of Facebook are concentrated among individuals who are particularly vulnerable to mental illness or whether they impact all stu
240、dents,we created a measure of predicted susceptibility to mental illness using a LASSO regression and studied heterogeneous treatment effects along that measure.The LASSO regression leverages a set of individual-level immutable characteristics such as gender and age to predict whether a student repo
241、rted having ever received a mental illness diagnosis.As shown in Figure 3,the effects of Facebook on the index of poor mental health impact all students,but they are especially strong among students who are predicted to be most susceptible to mental illness.Furthermore,among those students,Facebook
242、access significantly increased the take-up of psychotherapy and anti-depressants.FIGURE 3 HETEROGENEOUS EFFECTS BY PREDICTED SUSCEPTIBILITY TO MENTAL ILLNESS-.050.05.1.15.2Coefficient12345Quintile of Predicted Susceptibility to Mental IllnessIndex Symptoms Poor Mental Health-.050.05.1.15.2Coefficien
243、t12345Quintile of Predicted Susceptibility to Mental IllnessIndex Depression ServicesDoes the effect of Facebook on mental health have negative downstream repercussions on academic performance?According to the students reports,the answer is yes.The NCHA survey includes a host of questions asking stu
244、dents whether various conditions negatively affected their academic performance.We analyse all conditions related to mental health symptoms,along with an index summarising those symptoms.As shown in Figure 4,students were more likely to say that mental health issues negatively affected their academi
245、c performance after Facebook was introduced at their college.Consistent with our evidence suggesting that depression and anxiety-related disorders are the conditions most severely affected by the introduction of Facebook,we find the largest effect on a question asking about depression,anxiety and se
246、asonal affect disorder.The number of students who reported that those conditions impaired their academic performance increased by three percentage points over a baseline of 13%as a result of the introduction of Facebook.37SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINRobustness Our result
247、s pass numerous robustness checks.First,we observe null effects in placebo tests on variables that in principle should not be affected by the introduction of Facebook,such as our LASSO-predicted susceptibility to mental illness variable that is based on baseline immutable characteristics.Second,the
248、results remain similar in modified versions of our main specifications that take into account possible concerns related to(i)the construction of our index of poor mental health,(ii)the construction of our treatment variable,(iii)particular Facebook expansion groups driving the effects,(iv)other vari
249、ables unrelated to Facebook driving the effects,and(v)possible violations of the parallel trends assumption(see the online appendix in Braghieri et al.2022 for further details).FIGURE 4 DOWNSTREAM EFFECTS ON ACADEMIC PERFORMANCEAcademic perform attention deficitAcademic perform depression/anxiety/se
250、asonal affect disorderAcademic perform eating disorderAcademic perform sleep difficultyAcademic perform stressIndex Downstream Effects-.15-.1-.050.05.1.15Treatment effect(standard deviations)HOW DID FACEBOOK AFFECT MENTAL HEALTH?So far,we have documented that Facebook access negatively affected stud
251、ent mental health.But what was the mechanism behind this effect?Recent scholarship identified two main channels whereby Facebook might directly affect mental health:unfavourable social comparisons(Appel et al.2016)and disruptive internet use(Griffiths et al.2014).Unfavourable social comparisons refe
252、rs to the idea that users might use social media to compare themselves to others.To the extent that those comparisons are unfavourable,they might be detrimental to the users self-esteem and mental health.Disruptive internet use refers to the idea that social media might disrupt students ability to c
253、oncentrate and to carry out their daily tasks,and lead to anxiety.Another possibility is that the 38THE POLITICAL ECONOMY OF SOCIAL MEDIAintroduction of Facebook might lead to behavioural changes that,in turn,affect mental health.Overall,our evidence is most consistent with the unfavourable social c
254、omparisons channel.Unfavourable social comparisons We find two pieces of evidence that suggest that Facebooks effect operated through social comparison.First,we focus on students who are more likely to be affected by unfavourable social comparisons:(i)students who live off-campus and are therefore l
255、ess likely to participate in on-campus social life;(ii)students who have weaker offline social networks as measured by not belonging to a fraternity or sorority;(iii)students who have lower socioeconomic status as measured by carrying credit card debt or working part-time alongside studying;and(iv)s
256、tudents who are overweight.We aggregate these questions into an index of social comparison where respondents are considered to be at higher risk of unfavourable social comparisons if they have an above median number of the characteristics described above(e.g.they live off-campus,are overweight,and h
257、ave credit card debt).Figure 5 shows that Facebook access had a more negative effect on students more likely to suffer from negative social comparisons.All the point estimates are positive and the estimates for off-campus living,credit card debt,and the index are statistically significant.FIGURE 5 H
258、ETEROGENEOUS EFFECTS AS EvIDENCE OF UNFAvOURABLE SOCIAL COMPARISONSPost Facebook Introduction xOff-campus LivingPost Facebook Introduction xNot in Fraternity/SororityPost Facebook Introduction xCredit-card DebtPost Facebook Introduction xWorkPost Facebook Introduction xOverweightPost Facebook Introd
259、uction xIndex Social Comparisons-.08-.040.04.08Interaction Coefficient(standard deviations)39SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINSecond,we test directly whether Facebook affected peoples perceptions of their peers social lives by estimating the impact of the roll-out of Facebook
260、 on students perceptions of their peers drinking behaviours.6 Figure 6 shows that the introduction of Facebook increased the perceived prevalence of alcohol consumption among college students.Based on the questions asking students about their own alcohol consumption,the figure also shows that the in
261、crease in perceived alcohol consumption does not reflect an actual increase in consumption.FIGURE 6 EFFECTS ON ALCOHOL USE AND PERCEPTIONS AS EvIDENCE OF UNFAvORABLE SOCIAL COMPARISONS Typical student drink countDrink countShare of students who drink,30 daysDrink,30 daysDoes typical student drinks d
262、ailyDrink dailyAlcohol perceptions indexAlcohol use index-.050.05.1.15Post Facebook Introduction(standard deviations)UsagePerceptionsOne explanation for a discrepancy between perceptions and reality in regards to alcohol is that students might have a hard time interpreting the content they observe o
263、n social media.In particular,they might forget that what they see on social media is a curated rather than representative version of their peers lives.Indeed,we find an even stronger effect on perceptions among students living off-campus who have to rely more heavily on social media for information
264、about their peers behaviours.The changing perceptions could explain the negative effect on mental health,as inflated perceptions about others social lives might make students feel worse about their own.6 At the time,content related to alcohol featured prominently on Facebook.40THE POLITICAL ECONOMY
265、OF SOCIAL MEDIAAlternative channels We do not find evidence that the effect of Facebook on mental health operated through disruptive internet use.Facebook does not affect the share of students who report that the internet or video games affected their academic performance,as one would expect if Face
266、book were a distracting force.We also do not find evidence that Facebook affected mental health indirectly by affecting other behaviours.Using a battery of questions in the NCHA survey,we find that Facebook did not affect drug use,assaults,sexual assaults or the answers to various questions related
267、to relationships.CONCLUSIONIn 2021,4.3 billion individuals had a social media account,accounting for over half the world population and over 90%of internet users(We Are Social 2021).The repercussions of the rise of social media are thus likely to be far-reaching.We leverage the staggered introductio
268、n of Facebook across US colleges and find that the introduction of Facebook at a college had a negative effect on student mental health.Evidence points to unfavourable social comparisons as the leading mechanism.Since our identification strategy delivers estimates of the effect of Facebook in the mi
269、d-2000s,one might wonder about the extent to which our estimates speak to the effects of social media today.Over the last fifteen years,Facebook introduced a host of new features,including the newsfeed algorithm,business pages,and videos.Although we cannot estimate the effect of these new features o
270、n mental health,we believe our estimates are still relevant because the main force driving our results unfavourable social comparisons is still a common feature of many social media platforms today.In fact,some of the new features introduced by Facebook might have exacerbated the effects of social c
271、omparisons:the information users receive on their peers is now richer(e.g.it includes videos),it is personalised by an algorithm,and content can be accessed at any time or place using a smartphone.We emphasise that our analysis does not aim to estimate the overall welfare effects of social media;rat
272、her,it aims to shed light on a very important component of such a welfare calculation,namely,mental health.Clearly,social media might have positive effects on other outcomes affecting welfare.Indeed,the fact that individuals keep using social media despite the negative effects on subjective wellbein
273、g and mental health suggests that social media platforms might have benefits that compensate for such costs.Ideally,future iterations of these platforms will be able to preserve the benefits while mitigating the mental health costs.41SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINIn terms
274、of policy implications,our evidence on mechanisms suggests that regulators can consider interventions reminding the public that social media posts are not representative of peoples real lives.Such intervention could include behaviour nudges on social media platforms or be part of digital education p
275、rograms.Overall,our results are consistent with the hypothesis that social media might be partly responsible for the recent deterioration in mental health among young adults.It is up to social media platforms,regulators and future research to determine whether and how these effects can be alleviated
276、.REFERENCESAllcott,H,L Braghieri,S Eichmeyer,and M Gentzkow(2020),“The welfare effects of social media”,American Economic Review 110(3):629676.Allcott,H,M Gentzkow and L Song(2021),“Digital addiction”,NBER Working Paper No.28936.Appel,H,A L Gerlach,and J Crusius(2016),“The interplay between Facebook
277、 use,social comparison,envy,and depression”,Current Opinion in Psychology 9:4449.Armona,L(2019),“Online Social Network Effects in Labor Markets:Evidence From Face-books Entry into College Campuses”,available at SSRN 3381938.Bekalu,M A,R F McCloud,and K Viswanath(2019),“Association of social media us
278、e with social well-being,positive mental health,and self-rated health:disentangling routine use from emotional connection to use”,Health Education&Behavior 46(2S):69S80S.Berryman,C,C J Ferguson,and C Negy(2018),“Social media use and mental health among young adults”,Psychiatric Quarterly 89(2):30731
279、4.Borusyak,K,X Jaravel,and J Spiess(2021),“Revisiting event study designs:Robust and efficient estimation”,arXiv preprint arXiv:2108.12419.Braghieri,L,R Levy,and A Makarin(2022),“Social media and mental health”,American Economic Review 112(11):3660-3693.Bursztyn,L,G Egorov,R Enikolopov,and M Petrova
280、(2019),“Social media and xenophobia:evidence from Russia”,NBER Working Paper No.26567Callaway,B and P H SantAnna(2021),“Difference-in-differences with multiple time periods”,Journal of Econometrics 225(2):200230.Chan,D(2010),“So why ask me?Are self-report data really that bad?”,in Statistical and me
281、thodological myths and urban legends,Routledge,pp.329356.De Chaisemartin,C and X dHaultfoeuille(2020),“Two-way fixed effects estimators with heterogeneous treatment effects”,American Economic Review 110(9):296496.42THE POLITICAL ECONOMY OF SOCIAL MEDIADienlin,T,P K Masur,and S Trepte(2017),“Reinforc
282、ement or displacement?The reciprocity of FtF,IM,and SNS communication and their effects on loneliness and life satisfaction”,Journal of Computer-Mediated Communication 22(2):7187.Donati,D,R Durante,F Sobbrio,and D Zejcirovic(2022),“Lost in the net?Broadband internet and youth mental health”,IZA Disc
283、ussion Paper No.15202.Enikolopov,R,A Makarin,and M Petrova(2020),“Social media and protest participation:Evidence from Russia”,Econometrica 88(4):14791514.Enikolopov,R,M Petrova,and K Sonin(2018),“Social media and corruption”,American Economic Journal:Applied Economics 10(1):15074.Fergusson,L and C
284、Molina(2020),“Facebook causes protests”,Documentos CEDE 18002,Universidad de los Andes.Fujiwara,T,K Muller,and C Schwarz(2021),“The effect of social media on elections:Evidence from the United States”,NBER Working Paper No.28849.Griffiths,M D,D J Kuss,and Z Demetrovics(2014),“Social networking addic
285、tion:An overview of preliminary findings”,Chapter 6 in K P Rosenberg and L C Feder(eds),Behavioral Addictions,Academic Press.Hass,N(2006),“In your F”,The New York Times.Jacobs,A Z,S F Way,J Ugander,and A Clauset(2015),“Assembling the Facebook:Using heterogeneity to understand online social network a
286、ssembly”,WebSci 15:Proceedings of the ACM Web Science Conference.Kelly,Y,A Zilanawala,C Booker,and A Sacker(2018),“Social media use and adolescent mental health:Findings from the UK Millennium Cohort Study”,EClinicalMedicine 6:5968.Kroenke,K,R L Spitzer,and J B Williams(2001),“The PHQ-9:Validity of
287、a brief depression severity measure”,Journal of General Internal Medicine 16(9):606613.Leshner,A I and L A Scherer(2021),Mental health,substance use,and wellbeing in higher education:supporting the whole student,National Academies Press.Levy,R(2021),“Social media,news consumption,and polarization:Ev
288、idence from a field experiment”,American Economic Review 111(3):83170.Lin,L Y,J E Sidani,A Shensa,A Radovic,E Miller,J B Colditz,B L Hoffman,L M Giles,and B A Primack(2016),“Association between social media use and depression among us young adults”,Depression and Anxiety 33(4):323331.Mosquera,R,M Od
289、unowo,T McNamara,X Guo,and R Petrie(2020),“The economic effects of Facebook”,Experimental Economics 23(2):575602.43SOCIAL MEDIA AND MENTAL HEALTH|BRAGHIERI,LEVY AND MAKARINMuller,K and C Schwarz(2020),“From hashtag to hate crime:Twitter and anti-minority sentiment”,available at SSRN 3149103.National
290、 Center for Health Statistics(2021),“Health,United States,2019:Table 007”.NSDUH National Survey on Drug Use and Health(2019),“20022019 National Survey on Drug Use and Health Final Analytic File(Codebook)”.Paul,K I and K Moser(2009),“Unemployment Impairs Mental Health:Meta-Analyses”,Journal of Vocati
291、onal Behavior 74(3):264282.Spitzer,R L,K Kroenke,J B Williams,and B Lwe(2006),“A brief measure for assessing generalized anxiety disorder:The GAD-7”,Archives of Internal Medicine 166(10):1092 1097.Sun,L and S Abraham(2021),“Estimating dynamic treatment effects in event studies with heterogeneous tre
292、atment effects”,Journal of Econometrics 225(2),175199.Traud,A L,P J Mucha,and M A Porter(2012),“Social structure of Facebook networks”,Physica A:Statistical Mechanics and its Applications 391(16),41654180.Twenge,J M and W K Campbell(2019),“Media use is linked to lower psychological well-being:eviden
293、ce from three datasets”,Psychiatric Quarterly 90(2):311331.Twenge,J M,A B Cooper,T E Joiner,M E Duffy,and S G Binau(2019),“Age,period,and cohort trends in mood disorder indicators and suicide-related outcomes in a nationally representative dataset,20052017”,Journal of Abnormal Psychology 128(3):185.
294、We Are Social(2021),Digital 2021 April Global Statshot Report.Wells,G,J Horwitz,and D Seetharaman(2021),“Facebook knows Instagram is toxic for teen girls,company documents show”,Wall Street Journal.Zhuravskaya,E,M Petrova,and R Enikolopov(2020),“Political effects of the internet and social media”,An
295、nual Review of Economics 12:415438.ABOUT THE AUTHORSLuca Braghieri is an Assistant Professor in the Department of Decision Sciences at Bocconi University.He holds a PhD in Economics from Stanford University and a bachelors degree from Harvard College.Most of his work is on topics related to applied
296、microeconomics,behavioral economics,and political economy.44THE POLITICAL ECONOMY OF SOCIAL MEDIARoee Levy is a Senior Lecturer(Assistant Professor)at the Tel Aviv University School of Economics and a CEPR Research Affiliate.His research focuses on political economy and social media.He studies the f
297、orces shaping social norms,political attitudes,and policy preferences,and their subsequent influence on individual behavior.Before joining Tel Aviv University,Roee visited MIT as a post-doc and received his PhD from Yale University.Alexey Makarin is an Assistant Professor of Applied Economics and a
298、Mitsubishi Career Development Assistant Professor in International Management at the MIT Sloan School of Management.He earned his PhD in Economics from Northwestern University in June 2019.Alexeys research centres around political economy and the economics of digitization,with a special focus on the
299、 causal effects of social media platforms.SECTION 2 SOCIAL MEDIA AND VOTERS47SOCIAL MEDIA,NEWS CONSUMPTION AND POLARISATION|LEVYCHAPTER 3Social media,news consumption and polarisationRoee LevyTel Aviv University and CEPRThe share of Americans getting news on social media has been steadily increasing
300、 over the recent past.In 2022,70%of Americans consumed news on social media compared to fewer than one in eight Americans in 2008.When asked on which platforms they often get news,1829 year-olds report getting news on social media more than on any other medium,including television and news websites(
301、Shearer 2018).As social media becomes a major source of news,there is growing apprehension over its effects on public opinion.A primary concern is that individuals are exposed to more news matching their ideology on social media.More pro-attitudinal news exposure could occur due to friends sharing l
302、ike-minded content(echo chambers)or as a result of algorithms prioritising content the user agrees with(filter bubbles)(Pariser 2011).More exposure to like-minded news with a clear ideological slant may increase polarisation and threaten democracy(Sunstein 2017).In a recent paper(Levy 2021),I collec
303、ted novel data and conducted an experiment on Facebook to analyse news consumption on social media and its effect on polarisation.The paper has three main findings.First,individuals are exposed to more pro-attitudinal content and visit more pro-attitudinal websites through social media.Second,some o
304、f this behaviour can be explained by the algorithms governing social media.When individuals follow a Facebook page that matches their political opinions,they are more likely to see posts from that page,compared to a followed page that does not share the individuals opinions.Third,exposure to content
305、 on social media increases affective polarisation.Researchers tend to separate between affective polarisation,defined as the extent to which people of different parties dislike each other,and ideological polarisation,defined as the growing gap in opinions between the parties(Tucker et al.2018).While
306、 I do not find evidence that exposure to ideological content increases ideological polarisation,I find that exposure to pro-attitudinal content increases affective polarisation,compared to counter-attitudinal content.Together,the results suggest that social media platforms may be increasing polarisa
307、tion by exposing individuals to more pro-attitudinal content.48THE POLITICAL ECONOMY OF SOCIAL MEDIAIn this chapter,I discuss these three results.The first section analyses the news individuals are exposed to on social media.The second section explains the field experiment I conducted and presents r
308、esults on how news exposure affects attitudes.The third section investigates why individuals are exposed to more pro-attitudinal content on social media.IS NEWS CONSUMED THROUGH SOCIAL MEDIA MORE SEGREGATED?I analyse whether participants tend to consume like-minded news by merging the 2017 and 2018
309、Comscore WRDS Web Behavior Database Panels,from the Wharton Research Data Services at the University of Pennsylvania,with a dataset by Bakshy et al.(2015),defining the slant(ideological leaning)of 500 news domains.For each individual in the panel,I calculate the average slant of news sites visited t
310、hrough Facebook(i.e.by clicking a link in a Facebook post)and the slant of all other news sites visited.I then focus on the sample of participants who consumed news both through Facebook and through other websites and test whether news consumed through Facebook are more extreme and like-minded than
311、other news.Figure 1 shows that news consumed through Facebook are more extreme and pro-attitudinal.Figure 1a presents the distribution of the mean slant of news consumption and finds that news sites visited through Facebook are more extreme.For example,when visiting news sites through Facebook,57%of
312、 individuals consume news that is on average more conservative than the Wall Street Journal or more liberal than the Washington Post,and when visiting news sites through other sources,only 39%of individuals consume such partisan news.However,this figure does not provide information on who consumes e
313、xtreme news.Figure 1b shows that the most ideological individuals consume extreme news.The figure shows a clear correlation between the consumers ideology,proxied based on their zip code,and the slant of their news consumption.The slope for news consumed through Facebook is steeper than the slope fo
314、r news consumed through other sources,indicating that sites visited through Facebook tend to better match the consumers ideology(for example,Republicans visit even more conservative sites through Facebook).To quantify the difference between news consumed through social media and other news,I calcula
315、te segregation in news consumption,defined as the scaled standard deviation of the mean slant of news sites visited by individuals(Flaxman et al.2016).I find that the segregation increases by over 50%when consuming news through Facebook compared to other news consumed.In other words,there is more va
316、riation in the average slant of news sites individuals visit when they click on links appearing on Facebook.I complement this result with data from participants who installed a Google Chrome extension as part of the experiment,which is discussed in the next section.This additional dataset confirms t
317、hat news consumed through social media is more segregated,indicating that this result does not stem from the characteristics of the Comscore 49SOCIAL MEDIA,NEWS CONSUMPTION AND POLARISATION|LEVYsample.I use the richer extension data to better understand the mechanisms leading to segregation.I find t
318、hat the increase in segregation is mostly due to individuals clicking on posts posted by pages on Facebook and not posts by friends.This suggests that we should be more concerned with the personalisation of news outlets in social media feeds and less concerned with homophily.Finally,the extension da
319、ta can be matched with self-reported data on party affiliation.This allows me to calculate isolation in online news consumption,defined as whether conservatives and liberals visit different websites(Gentzkow and Shapiro 2011).I find that isolation is much greater when visiting sites through social m
320、edia.FIGURE 1 NEWS CONSUMPTION IN THE COMSCORE PANELa)Distribution of mean news slantBoston GlobeBreitbartFox NewsHuffPostYahoo NewsNYTSalonUSA TodayWashington PostWSJ0.00.51.01.52.00.50.00.51.0SlantDensityFacebook ReferencesOther Newsb)Ideology and slant of news consumption0.60.30.00.30.60.10.20.30
321、.40.50.60.70.80.9Republican donations shareMean news slant,std.dev.(Higher=more conservative)Facebook ReferencesOther News50THE POLITICAL ECONOMY OF SOCIAL MEDIAThe results in this section are different from previous literature,which has often argued that concerns about echo chambers are overstated(
322、Guess et al.2018).There are at least three explanations for this discrepancy.First,there is disagreement about the definitions of echo chambers and filter bubbles.While I provide evidence that news consumption is more segregated on social media,individuals are not in complete echo chambers as they a
323、re still exposed to moderate and counter-attitudinal opinions in their feeds(Bakshy et al.2015).Second,most news is still consumed through sources other than social media.Hence,even if news sites visited through social media are much more segregated,their effect on aggregate news consumption is limi
324、ted(Flaxman et al.2016).Third,some of the previous studies were conducted before social media became a popular source for news consumption.For example,Gentzkow and Shapiro(2011)find limited isolation in online news consumption,compared to offline sources,using data from 2009,while Peterson et al.(20
325、19)find much greater isolation in 2016.To conclude,the vast majority of Americans think that one-sided news is a very big,or at least a moderately big,problem on social media.1 This section provides evidence that this concern is warranted.News accessed through Facebook is indeed more segregated and
326、extreme than other online news,and although its current impact on total news consumption is limited,segregation may grow as more news is consumed through social media.THE EFFECT OF NEWSA field experiment to estimate the effect of social media news exposureIn February to March 2018,I conducted an exp
327、eriment where participants were randomly assigned to three groups:a group offered the opportunity to like Facebook pages of four liberal news outlets(e.g.,MSNBC),a group offered the opportunity to like pages of four conservative news outlets(e.g.,Fox News)and a control group that was not offered any
328、 outlets.When Facebook users like an outlets page on the platform,posts from the outlet may start appearing naturally in their feed,among many other posts they are exposed to(liking a post is similar to subscribing to specific content from an outlet,and I use the terms like and subscribe interchange
329、ably in this chapter).I designed the experiment to have high external validity.In contrast to lab studies,behaviour in the experiment occurs just as it does in the real world.Besides nudging individuals to like Facebook pages,the experiment did not directly intervene in any behaviour.The news suppli
330、ed to participants was the actual news provided by leading media outlets.Facebooks algorithm determined which of the posts shared by the outlets appeared in the participants Facebook feeds.Most importantly,participants decided whether to skip,read,click,or share posts.As a result,the effect of the i
331、ntervention 1 Pew Research Center American Trends Panel Wave 51,July 2019.51SOCIAL MEDIA,NEWS CONSUMPTION AND POLARISATION|LEVYis almost identical to the experience of millions of Americans who like popular news outlets on Facebook.I analyse three main datasets.First,attitudes were measured using a
332、follow-up survey conducted two months after the experiment.Second,to measure compliance with the intervention and to test whether the treatment affected the posts people shared,I used Facebooks API to collect data,with participants permission,on the pages people like on Facebook and the posts they s
333、hare.Third,to analyse the effects of the experiment on participants Facebook feeds and the news sites they visit,I asked a subset of participants who took the survey on Google Chrome to install an extension collecting this data.Figure 2 summarises the design of the experiment.FIGURE 2 EXPERIMENTAL DESIGN Recruitment using Facebook Ads(978,628 individuals exposed to the ads)Baseline survey,Feb-Marc