The Welfare E ects of Social MediaHunt Allcott, Luca Braghieri, Sarah Eichmeyer, and Matthew Gentzkow April 1, 2019AbstractThe rise of social media has provoked both optimism about potential societal bene ts andconcern about harms such as addiction, depression, and political polarization. We present arandomized evaluation of the welfare e ects of Facebook, focusing on US users in the run-up to the 2018 midterm election. We measured the willingness-to-accept of 2,743 Facebookusers to deactivate their Facebook accounts for four weeks, then randomly assigned a subsetto actually do so in a way that we veri ed. Using a suite of outcomes from both surveys anddirect measurement, we show that Facebook deactivation (i) reduced online activity, includingother social media, while increasing o ine activities such as watching TV alone and socializingwith family and friends; (ii) reduced both factual news knowledge and political polarization;(iii) increased subjective well-being; and (iv) caused a large persistent reduction in Facebookuse after the experiment. Deactivation reduced post-experiment valuations of Facebook, butvaluations still imply that Facebook generates substantial consumer surplus.JEL Codes: D12, D90, I31, L86, O33.Keywords: Social media, political polarization, subjective well-being, consumer surplusfrom digital technologies.|Allcott: New York University and NBER. hunt.allcottnyu.edu. Braghieri: Stanford University. lu-cabragstanford.edu. Eichmeyer: Stanford University. saraeichstanford.edu. Gentzkow: Stanford Universityand NBER. gentzkowstanford.edu. We thank Nancy Baym, Moira Burke, Annie Franco, Alex Leavitt, ToddRogers, and Joel Waldfogel for helpful comments. We thank Raj Bhargava, Zong Huang, and Kelly B. Wag-man for exceptional research assistance. We are grateful to the Sloan Foundation and the Knight Foundation forgenerous support. The study was approved by Institutional Review Boards at Stanford (eProtocol #45403) andNYU (IRB-FY2018-2139). This RCT was registered in the American Economic Association Registry for random-ized control trials under trial number AEARCTR-0003409. Replication les and survey instruments are availablefrom sites.google/site/allcott/research. Disclosures: Allcott is a paid employee of Microsoft Research.Gentzkow does paid consulting work for Amazon and is a member of the Toulouse Network for Information Technol-ogy, a research group funded by Microsoft. Braghieri and Eichmeyer have no relevant or material disclosures.11 IntroductionSocial media have had profound impacts on the modern world. Facebook, which remains by farthe largest social media company, has 2.3 billion monthly active users worldwide (Facebook 2018).As of 2016, the average user was spending 50 minutes per day on Facebook and its sister platformsInstagram and Messenger (Facebook 2016). There may be no technology since television that hasso dramatically reshaped the way people get information and spend their time.Speculation about social medias welfare impact has followed a familiar trajectory, with earlyoptimism about potential bene ts giving way to widespread concern about possible harms. At abasic level, social media dramatically reduce the cost of connecting, communicating, and sharinginformation with others. Given that interpersonal connections are among the most importantdrivers of happiness and well-being (Myers 2000; Reis, Collins, and Berscheid 2000; Argyle 2001;Chopik 2017), this could be expected to bring widespread improvements to individual welfare.Many have also pointed to wider social bene ts, from facilitating protest and resistance in autocraticcountries, to encouraging activism and political participation in established democracies (Howardet al. 2011; Kirkpatrick 2011).More recent discussion has focused on an array of possible negative impacts. At the individuallevel, many have pointed to negative correlations between intensive social media use and bothsubjective well-being and mental health.1 Adverse outcomes such as suicide and depression appearto have risen sharply over the same period that the use of smartphones and social media hasexpanded.2 Alter (2018) and Newport (2019), along with other academics and prominent SiliconValley executives in the time well-spent“ movement, argue that digital media devices and socialmedia apps are harmful and addictive. At the broader social level, concern has focused particularlyon a range of negative political externalities. Social media may create ideological echo chambers“among like-minded friend groups, thereby increasing political polarization (Sunstein 2001, 2017;Settle 2018). Furthermore, social media are the primary channel through which fake news andother types of misinformation are spread online (Allcott and Gentzkow 2017), and there is concernthat coordinated disinformation campaigns can a ect elections in the US and abroad.In this paper, we report on a large-scale randomized evaluation of the welfare impacts of Face-book, focusing on US users in the run-up to the November 2018 midterm elections. We recruited asample of 2,743 users through Facebook display ads, and elicited their willingness-to-accept (WTA)to deactivate their Facebook accounts for a period of four weeks ending just after the election. Wethen randomly assigned the 61 percent of these subjects with WTA less than $102 to either a Treat-ment group that was paid to deactivate, or a Control group that was not. We veri ed compliance1See, for example, Abeele et al. (2018), Burke and Kraut (2016), Ellison, Stein eld, and Lampe (2007), Frison andEggermont (2015), Kross et al. (2013), Satici and Uysal (2015), Shakya and Christakis (2017), and Tandoc, Ferrucci,and Du y (2015). See Appel, Gerlach, and Crusius (2016) and Baker and Algorta (2016) for reviews.2See, for example, Twenge, Sherman, and Lyubomirsky (2016), Twenge and Park (2017), Twenge, Martin, andCampbell (2018), and Twenge et al. (2018).2with deactivation by regularly checking participants public pro le pages. We measured a suite ofoutcomes using text messages, surveys, emails, and direct measurement of activity on Facebookand Twitter. Less than two percent of the sample failed to complete the endline survey, and theTreatment groups compliance with deactivation exceeded 90 percent.Our study o ers the largest-scale experimental evidence available to date on the way Facebooka ects a range of individual and social welfare measures. We evaluate the extent to which time onFacebook substitutes for alternative online and o ine activities, with particular attention to crowdout of news consumption and face-to-face social interactions. We study Facebooks broader politicalexternalities via measures of news knowledge, awareness of misinformation, political engagement,and political polarization. We study the impact on individual utility via measures of subjective well-being, captured through both surveys and text messages. Finally, we analyze the extent to whichforces like addiction, learning, and projection bias may cause sub-optimal consumption choices, bylooking at how usage and valuation of Facebook change after the experiment.Our rst set of results focuses on substitution patterns. A key mechanism for e ects on indi-vidual well-being would be if social media use crowds out face-to-face social interactions and thusdeepens loneliness and depression (Twenge 2017). A key mechanism for political externalities wouldbe if social media crowds out consumption of higher-quality news and information sources. We ndevidence consistent with the rst of these but not the second. Deactivating Facebook freed up 60minutes per day for the average person in our Treatment group. The Treatment group actuallyspent less time on both non-Facebook social media and other online activities, while devoting moretime to a range of o ine activities such as watching television alone and spending time with friendsand family. The Treatment group did not change its consumption of any other online or o inenews sources and reported spending 15 percent less time consuming news.Our second set of results focuses on political externalities, proxied by news knowledge, politicalengagement, and political polarization. Consistent with the reported reduction in news consump-tion, we nd that Facebook deactivation signi cantly reduced news knowledge and attention topolitics. The Treatment group was less likely to say they follow news about politics or the Pres-ident, and less able to correctly answer factual questions about recent news events. Our overallindex of news knowledge fell by 0.19 standard deviations. There is no detectable e ect on politicalengagement, as measured by voter turnout in the midterm election and the likelihood of clickingon email links to support political causes. Deactivation signi cantly reduced polarization of viewson policy issues and a measure of exposure to polarizing news. Deactivation did not statisticallysigni cantly reduce a ective polarization (i.e. negative feelings about the other political party) orpolarization in factual beliefs about current events, although the coe cient estimates also point inthat direction. Our overall index of political polarization fell by 0.16 standard deviations. As apoint of comparison, prior work has found that a di erent index of political polarization rose by0.38 standard deviations between 1996 and 2018 (Boxell 2018).3Our third set of results looks at subjective well-being. Deactivation caused small but signi cantimprovements in well-being, and in particular in self-reported happiness, life satisfaction, depres-sion, and anxiety. E ects on subjective well-being as measured by responses to brief daily textmessages are positive but not signi cant. Our overall index of subjective well-being improved by0.09 standard deviations. As a point of comparison, this is about 25-40 percent of the e ect ofpsychological interventions including self-help therapy, group training, and individual therapy, asreported in a meta-analysis by Bolier et al. (2013). These results are consistent with prior studiessuggesting that Facebook may have adverse e ects on mental health. However, we also show thatthe magnitudes of our causal e ects are far smaller than those we would have estimated using thecorrelational approach of much prior literature. We nd little evidence to support the hypothesissuggested by prior work that Facebook might be more bene cial for active“ users|for example,users who regularly comment on pictures and posts from friends and family instead of just scrollingthrough their news feeds.3Our fourth set of results considers whether deactivation a ected peoples demand for Facebookafter the study was over, as well as their opinions about Facebooks role in society. As the experi-ment ended, participants reported planning to use Facebook much less in the future. Several weekslater, the Treatment groups reported usage of the Facebook mobile app was about 11 minutes (22percent) lower than in Control. The Treatment group was more likely to click on a post-experimentemail providing information about tools to limit social media usage, and ve percent of the Treat-ment group still had their accounts deactivated nine weeks after the experiment ended. Our overallindex of post-experiment Facebook use is 0.61 standard deviations lower in Treatment than in Con-trol. In response to open-answer questions several weeks after the experiment ended, the Treatmentgroup was more likely to report that they were using Facebook less, had uninstalled the Facebookapp from their phones, and were using the platform more judiciously. Reduced post-experiment usealigns with our nding that deactivation improved subjective well-being, and it is also consistentwith the hypotheses that Facebook is habit forming in the sense of Becker and Murphy (1988) orthat people learned that they enjoy life without Facebook more than they had anticipated.Deactivation caused people to appreciate Facebooks both positive and negative impacts on theirlives. Consistent with our results on news knowledge, the Treatment group was more likely to agreethat Facebook helps people to follow the news. About 80 percent of the Treatment group agreedthat deactivation was good for them, but they were also more likely to think that people wouldmiss Facebook if they used it less. In free response questions, the Treatment group wrote more textabout how Facebook has both positive and negative impacts on their lives. The opposing e ectson these speci c metrics cancel out, so our overall index of opinions about Facebook is una ected.Our work also speaks to an adjacent set of questions around how to measure the economic3Correlation studies on active vs. passive Facebook use include Burke, Marlow, and Lento (2010), Burke, Kraut,and Marlow (2011), Burke and Kraut (2014), and Krasnova et al. (2013), and randomized experiments include Detersand Mehl (2012) and Verduyn et al. (2015).4gains from free online services such as search and media.4 In standard models with consumers whocorrectly optimize their allocation of time and money, researchers can approximate the consumersurplus from these services by measuring time use or monetary valuations, as in Brynjolfsson and Oh(2012), Brynjolfsson, Eggers, and Gannamaneni (2018), Corrigan et al. (2018), and others. But ifusers do not understand the ways in which social media could be addictive or make them unhappy,these standard approaches could overstate consumer surplus gains. Sagioglu and Greitemeyer(2014) provide suggestive evidence: while their participants predicted that spending 20 minutes onFacebook would make them feel better, it actually caused them to feel worse. Organizations suchas Time to Log O argue that a 30-day digital detox“ would help people align their social mediausage with their own best interest.To quantify the possibility that deactivation might help the Treatment group to understandways in which their use had made them unhappy, we elicited willingness-to-accept at three separatepoints, using incentive-compatible Becker-DeGroot-Marschak (1964, BDM“) mechanisms. First,on October 11th, we elicited WTA to deactivate Facebook for weeks 1-4 of the experiment, betweenOctober 12th and November 8th. We immediately told participants the amount that they hadbeen o ered to deactivate ($102 for the Treatment group, $0 for Control), and thus whether theywere expected to deactivate over that period. We then immediately elicited WTA to deactivateFacebook for the next four weeks after November 8th, i.e. weeks 5-8. When November 8th arrived,we then re-elicited WTA to deactivate for weeks 5-8. The Treatment groups change in valuationfor weeks 5-8 re ects a time e ect plus the e ect of deactivating Facebook. The Control groupsparallel valuation change re ects only a time e ect. Thus, the di erence between how Treatmentvs. Control change their WTAs for deactivation for weeks 5-8 re ects projection b