March 26th 2016

The signal and the noise

S P E C I A L R E P O R T

T E C H N O L O G Y A N D P O L I T I C S

20160326_Politicsanddata.indd 1 10/03/2016 13:36

The Economist March 26th 2016 1

TECHNOLOG Y AND POLITICS

SPECIAL REPOR T

A list of sources is atEconomist.com/specialreports

An audio interview with the author is atEconomist.com/audiovideo/specialreports

CONTENT S

2 Election campaignsPolitics by numbers

4 Tracking protestmovementsA new kind of weather

5 Online collaborationConnective action

7 Local governmentHow cities score

9 Living with technologyThe data republic

1

DONALD TRUMP, THE Republican front-runner for the American presi-dency, is clearly ridinga wave ofanger—buthe is also wielding a huge vir-tual megaphone to spread his populist messages. “@realDonaldTrump”,the Twitter account of the property magnate turned politician, has morethan 7m followers and the number is rising by about 50,000 every day.Moreover, since each of his tweets is re-tweeted thousands of times andoften quoted in mainstream media, his real audience is much bigger. Andifhe doeswin the Republican nomination, itwill be hard to tune him out.“How do you fight millions ofdollars offraudulent commercials pushingfor crooked politicians?” he tweeted in early March. “I will be using Face-book& Twitter. Watch!”

If Ted Cruz, his fellow Republican, were to clinch the nomination,the campaign for America’s presidency would be quieter—but no lessdigital. Mr Cruz’s victory in the Iowa primaries was based on effective

number-crunching. He bombard-ed potential supporters withhighly targeted ads on Facebook,and used algorithms to label vot-ers as “stoic traditionalists”, “tem-peramental conservatives” or“true believers” to give campaignvolunteers something to go on.He also sent official-looking“shaming” letters to potentialsupporters who had previouslyabstained from voting. Under thehea�������

oting�

iolation”, theletters reminded recipients oftheir failure to do their civic dutyat the polls and compared theirvoting records with those of theirneighbours.

The way these candidatesare fighting their campaigns, eachin his own way, is proof that poli-tics as usual is no longer an op-tion. The internet and the avail-ability of huge piles of data on

everyone and everything are transforming the democratic process, justas they are upending many industries. They are becoming a force in allkinds of things, from running election campaigns and organising protestmovements to improving public policy and the delivery ofservices. Thisspecial report will argue that, as a result, the relationship between citi-zens and those who govern them is changing fundamentally.

Incongruous though it may seem, the forces that are now poweringthe campaign of Mr Trump—as well as that of Bernie Sanders, the sur-prise candidate on the Democratic side (Hillary Clinton is less of a suc-cess online)—were first seen in full cry during the Arab spring in 2011. Therevolution in Egypt and other Arab countries was not instigated by Twit-ter, Facebookand othersocial-media services, but theycertainly helped itgain momentum. “The internet is an intensifier,” says Marc Lynch ofGeorge Washington University, a noted scholar of the protest move-ments in the region.

In the course of just a few years digital technology has become anessential ingredient in any protest movement. The Arab spring is just one

The signal and the noise

Ever easier communications and ever-growing data mountains aretransforming politics in unexpected ways, says Ludwig Siegele.What will that do to democracy?

ACKNOWLEDGMENT S

The author would like to acknowl-edge the generous help he receivedfrom many people while preparingthis report. Apart from thosementioned in the text, particularthanks go to Ayad Al-Ani, NickBeauchamp, Dan Breznitz, RobynCaplan, Matthew Claudel, RonaldDeibert, Bruce Ettling, Rob Faris,Alessandro Flammini, Deen Freelon,King-wa Fu, Wael Ghonim, Erin Hill,John Kelly, Julia Lane, LawrenceLessig, Matt Lira, Carl Miller, CharlesMok, Beth Noveck, Andrew Rasiej,Ben Rattray, Hal Roberts, MichaelSchwemmle, Micah Sifry, Nick Sinai,Janice Stein, Josh Tucker, Milan deVries, Albert Wenger, Irving Wladaw-sky-Berger, Nicole Wong, TahaYasseri and Ethan Zuckerman.

On the coverThe cover of this special report is partof a visualisation of the Twitternetwork of America’s anti-vaccina-tion movement, developed by GiladLotan of betaworks, a startupincubator. The circles representTwitter accounts and their size theirauthority within the network. Thelines between them mean they followeach other. Different colours markneighbourhoods of people who havestrong connections and share similarmessages.

2 The Economist March 26th 2016

SPECIAL REPOR TTECHNOLOG Y AND POLITICS

2

1

example of how the internet has facilitated political mobilisa-tion. Others include the civil unrest in Istanbul’s Taksim GeziPark, the Maidan protests in Ukraine and the Umbrella Move-ment in Hong Kong, all in 2013 or 2014. In America the main in-stances have been Occupy Wall Street in 2011 and more recentlyBlack Lives Matter, a campaign drawing attention to violenceagainst African-Americans. In Europe, Spain’s Indignados, ananti-austerity coalition, in 2011became the first big protest move-ment to make extensive use of social media. Even Islamic Staterelies on its online propaganda and messaging apps, which al-low the self-styled caliphate to recruit new fighters and keep intouch with those on the ground.

However, this special report will argue that, in the longerterm, online crusadingand organisingwill turn out to matter lessto politics in the digital age than harnessing those ever-growingpiles of data. The internet and related technologies, such assmartphones and cloud computing, make it cheap and easy notonly to communicate but also to collect, store and analyse im-mense quantitiesofinformation. This isbecomingevermore im-portant in influencing political outcomes.

America’s elections are a case in point. Mr Cruz with hisdata savvy ismerely following in the footstepsofBarack Obama,who won his first presidential term with the clever applicationof digital know-how. Campaigners are hoovering up more andmore digital information about every voting-age citizen andstashing itawayin enormousdatabases. With the aid of complexalgorithms, these data allow campaigners to decide, say, whoneeds to be reminded to make the trip to the polling station andwho may be persuaded to vote for a particular candidate.

No hiding placeIn the case of protest movements, the waves of collective

action leave a bigdigital footprint. Usingevermore sophisticatedalgorithms, governments can mine these data. That is changingthe balance of power. In the event of another Arab spring, auto-crats would not be caught off guard again because they are nowable to monitor protests and intervene when they consider itnecessary. They can also identify and neutralise the most influ-ential activists. Governments that were digitally blind when theinternet first tookoffin the mid-1990s now have both a telescopeand a microscope.

But data are not just changing campaigns and politicalmovements; they affect how policy is made and public servicesare offered. This is most visible at local-government level. Cities

have begun to use them for everything from smoothing trafficflows to identifying fire hazards. Having all this information attheir fingertips is bound to change the way these bureaucracieswork, and how they interact with citizens. This will not onlymake cities more efficient, but provide them with data and toolsthat could help them involve their citizens more.

This report will look at electoral campaigns, protest move-ments and local government in turn. Readers will note that mostof the examples quoted are American and that most of the peo-ple quoted are academics. That is because the study of the inter-relationship between data and politics is relatively new andmost developed in America. But it is beginning to spill out fromthe ivory towers, and is gradually spreading to other countries.

The growing role of technology in politics raises manyquestions. How much of a difference, for instance, do digitallyenabled protest surges really make? Many seem to emerge fromnowhere, then crash almost as suddenly, defeated by hard politi-cal realities and entrenched institutions. The Arab spring upris-ing in Egypt is one example. Once the incumbent president,Hosni Mubarak, was toppled, the coalition that brought himdown fell apart, leaving the stage to the old powers, first the Mus-lim Brotherhood and then the armed forces.

In party politics, some worry that the digital targeting ofvotersmightend up reducingthe democraticprocess to a market-ing exercise. Ever more data and better algorithms, they fret,could lead politicians to ignore those unlikely to vote for them.And in cities it is not clear that more data will ensure that citizensbecome more engaged.

When the internet first took off, the hope was that it wouldmake the world a more democraticplace. The fearnowis that theavalanche of digital information might push things the otherway��iktor Mayer-Schönberger, a data expert at the UniversityofOxford, sums up the problem: “Data are mainly helping thosewho already have information power.”7

Strong feelings

Source: Catalist

Prediction of turnout and strength of party supportIn 2010 elections

90,000+

Most RepublicanPartisanship score

Turn

out

prop

ensi

ty s

core

Mos

t lik

ely

to v

ote

Leas

t lik

ely

to v

ote

Most Democratic

0Number of votersper square

WHEN TOM PITFIELD talks about the campaign of JustinTrudeau, who was recently elected Canada’s prime minis-

ter, he gets animated. Mr Trudeau’s Liberal Party could not afforda lot of television time and spent much of its advertising budgeton social media. That proved an inspired choice. “We wouldcreate an ad, see how people reacted to it on Facebook, tweakthecontent and test it again. On some days we would produce morethan 50 different ads,” explains Mr Pitfield, who was in charge ofthe campaign’s digital side. This rapid feedback, he says, allowedhis team to offer much more flexible and targeted messages thanthe competition.

Although the trend is obscured by Donald Trump’s tweetsand his other antics, Facebook will also play a big role in Ameri-ca’s presidential contest this year. The tools that the world’s big-gest social network offers to campaigners are getting better allthe time. Last year it provided a way to upload lists of people toits site so they could be sent targeted messages. Now it is offeringa further service that allows campaigners to reach Facebook us-ers who “like” and share a lot ofpolitical content.

Election campaigns

Politics by numbers

Voters in America, and increasingly elsewhere too,are being ever more precisely targeted

The Economist March 26th 2016 3

TECHNOLOG Y AND POLITICS

2

1

Even Barack Obama’s first presidential campaign in 2008was widely hailed as “data-driven”. But it was only in 2012 thathis team systematically used digital technology to deal with ev-ery campaign’s biggest challenge: how to make the best use of alimited budget to reach the right voters. In the past, geographyhad served as a proxy target: if a precinct was considered Demo-cratic, for instance, it would get a lot ofattention from Democrat-ic campaigners. But in recent years it has become possible to tar-get voters individually, thanks to the availability of ever moredata as well as ever cheaper computing power and better meth-ods to mine them.

To find out where to concentrate its resources, the Obamacampaign used polls and otherdata to generate a statistical mod-el of the attributes potential Obama supporters had in common.“When volunteersknocked on doors in 2008, fouroutoften peo-ple they met backed Obama. In 2012 the ratio was nine out often,” says Dan Wagner, who led the president’s data-scienceteam during his second campaign.

Hidden persuadersFrom a windowless office in Chicago that became known

as the “data cave”, Mr Wagner and his colleagues also pioneereda number of other methods of persuasion in that election. Theytested the subject lines of fundraising e-mails (“I will be out-spent” raised $2.6m; “Do this for Michelle” only about$700,000). They found out whether a group ofvoters they want-ed to target watched certain cable shows, which allowed them touse television advertising more cost-effectively. Beyond mobilis-ing their own voters, they also tried to identify others who mightbe persuaded to change their mind (with limited success).

Such novel approaches helped scupper the campaign ofMitt Romney, the Republican candidate last time round. At thetime the Democratswere widelyexpected to retain a lasting edgein data, not least because they find it easier than the Republicansto attract highly trained data scientists. But that turned out to bewishful thinking. “Most of what we did in 2012 is now a com-modity,” says Mr Wagner, who went on to found Civis Analytics,a startup which offers data-management and analytics servicesto left-leaning groups.

The starting-point for all this infor-mation-gathering was the controversialFlorida recount after the presidential elec-tion of 2000, which became necessarypartly because of incomplete voter rolls.To avoid a repeat, Congress in 2002passed the Help Americote Act(HAVA), which required states to main-tain a “single, uniform, official, central-ised, interactive computerised statewidevoter registration list”. This was a biggerdeal than it appeared at first sight. Alongwith improvements in database technol-ogies, HAVA made it possible for the firsttime forpolitical parties to compile an up-to-date list of all voters in the country. Inhis book, “Hacking the Electorate”, EitanHersh of�ale University argues that thislaid the groundworkfor individual target-ing. Campaigners were able to identifyvoterseasilyand linkthem to other publicinformation.

The legislation also kicked off whatmight be called “database politics”.Democrats were the pioneers: in 2006partyofficials setup a companycalled Ca-

talist which today offers one of the most comprehensive data-bases on Americans of voting age, covering more than 240mpeople. Apart from the official register of voters, it also includesother public records and information from commercial data bro-kers. Each entry contains hundreds of pieces of information,from race to the probability ofowning an SUV.

Catalist is best understood as a “data co-operative” forDemocratic campaigns, trade unions and other left-wing organi-sations, says Laura Quinn, its chiefexecutive. Fora fee, it gives cli-ents access to its common data pool. They can combine it withtheir own information and benefit from the firm’s analytics ex-pertise, which mainly comes in the form of statistical “scores”.These numbers predict, for instance, how likely someone is tovote Democrat and go to the polls (see chart, previous page).

But the Democratic Party did not want to rely on an outsidedatabase. When Mr Obama became president, it decided tocreate its own. C�� �oteBuilder, it also relies on the principleof sharing data. During the primaries all competing Democraticcampaigns can u� oteBuilder and combine it with data theygather on the trail. Much of that additional information is keptseparate, but campaigns share basic items such as name changesor deaths. Once a nominee has been chosen for the general elec-tion, a lot of these data become part of the common pool.

The Republicans’ central database�oter

ault, has com-manded less co-operation and has often been neglected be-tween campaigns. Warring factions have insisted on producingtheir own databases, often working with commercial vendors.After Mr Romney lost to Mr Obama in 2012, Charles and DavidKoch, billionaire brotherswith a passion forconservative causes,invested millions in i360, a for-profit firm that competes with

oter

ault. On both sides, access to all these databases can be ahighly political issue. Insurgent Democratic candidates in stateand local races frequentlycomplain that theyare being excluded;they need to show a minimum ofco-operation to get the data.

Thanks to all thisdata-gathering, campaignersnowseem to“know you better than you know yourself”, as CNN, an Ameri-can cable newschannel, once put it. But the reality is different, ex-plains Mr Hersh in his book. A lot of the available data, particu-larly the commercial sort, are of little value in helping

Trumping the enemy

SPECIAL REPOR T

4 The Economist March 26th 2016

SPECIAL REPOR TTECHNOLOG Y AND POLITICS

2

1

campaigners decide which voters to target. The best guide is thebasic demographic information taken from public records, suchas gender, age, voting history and party affiliation.

According to Mr Hersh, this explains why most lawmakersare in favour of allowing easy access to public records. In 2012 alegislator in Utah proposed giving voters the option of limitingaccess to theirdate ofbirth, but the idea wasquashed by the lead-ership of both parties. Campaigners also file lots of requestsbased on the Freedom of Information Act and state statutes thatgovern public access to administrative data. “Ironically, laws os-tensibly passed to help private citizens track the government’saction turn out to be laws that help political campaigns trackprivate citizens,” writes Mr Hersh. But his main complaint is theconflict of interest arising when parties control the sources ofdata which they themselves use extensively.

The big question is whether the use of such databases andthe algorithms that sift through them change the outcome ofelections. Recent estimates suggest that they can add betweentwo and three percentage points to a candidate’s result. In aclosely fought election that could be crucial, but in an emotional-ly charged race between Mr Trump and Mrs Clinton it may notplay a decisive role. Then again, Mrs Clinton will probably endup relying on technology more than she has done so far. TheDemocrats, explains Matt Hindman of George Washington Uni-versity, always have a harder time getting their supporters out tovote than the Republicans do. Since Mrs Clinton has not enjoyeduniversal enthusiasm for her candidacy, getting out the voters inNovember may be even more difficult than usual.

Besides, even if parties do not derive a lasting and decisiveadvantage from data and analytics, they will still be obliged to in-vest in technology to keep up with their competitors. And al-though social media give politicians direct access to their voters,as the Trump campaign shows, big parties still enjoy an advan-tage because theyhave the moneyto hire technical talent, payforpolls and buy advertising on Facebook.

Shrinking the public sphereMore broadly, some people worry about how all this num-

ber-crunching will affect democracy, in America and elsewhere.Mr Hersh does not see much of a problem, as long as data areused just to getpeople out to vote, rather than to try to make themchange theirmind. But Zeynep Tufekci ofthe University ofNorthCarolina argues that targeting voters with ever more personal-ised messages will shrink the “public sphere”, which Jürgen Ha-bermas, a German philosopher, once defined as the basis of de-mocracy. “This form of big-data-enabled computational politicsisa private one. At its core, it isopposed to the idea ofa civic spacefunctioning as a public, shared commons,” writes Ms Tufekci.And privacy is a growing concern. In December a database con-taining the records of191m voters found its way onto the internet.

Rasmus Kleis Nielsen of Oxford University, who has writ-ten a book on political campaigns in America, thinks that suchtargeting will remain largely confined to that country. Nowhereelse have party organisations access to so much money, data andtechnical talent. Moreover, America’s political system lends it-self well to analytics because once voters get to the polling sta-tion they often have only two options. �

et there are signs that some other countries are shifting inAmerica’s direction. Apart from buying lots of ads on Facebook,Canada’s Liberal Party in last year’s election used the services ofCivis Analytics, a firm spun off from the Obama campaign in2012. In Britain, too, targeted ads on Facebookhelped the Conser-vative Party win the general election last May. And when itcomes to using social media to influence the political weathermore generally, America is by no means alone. 7

WHEN BRITAIN’S MOST active followers of Islamic State(IS) sent tweets and posted on Facebook, they did not ex-

pect all their online output to end up in print on a wall at the In-ternational Centre for the Study of Radicalisation at King’s Col-lege London. “Allah gave you this life, so use it the way hecommanded you to,” writes one. “Whenever pain comes to me Ijust stop and think, ‘Allah does not give me more than I can bear.’Then I carry on,” says another.

The unusual display is an important part of a research pro-ject to find out if British IS followers who left for Syria use socialmedia differently from those who stayed at home. Shiraz Maher,the project’s leader, and his colleagues took months to classifyeach ofthe 50,000-odd messagesand postsunderheadings suchas “yearning for the afterlife”, “religiosity” and “desire to mi-grate”. “If we can find patterns, we can perhaps predict whensomebody is about to go,” says Mr Maher.

The research illustrates a basic dilemma for movementsthat live at least partly online: without the internet and socialmedia some of them would not exist, but the technology alsoopens them up to an unprecedented degree of scrutiny. And thetools for laying them bare are getting ever better.

Gilad Lotan, the chief data scientist at betaworks, an incu-bator for startups in New�

ork, did his military service in the in-telligence corps of the Israeli army, where he was trained to digup information. He was pleasantly surprised when one day thedata just started arriving without any effort on his part—on Twit-ter. The micro-blogging service was an important communica-tion channel for activists of Iran’s Green Movement in 2009 be-fore it was blocked by the government in Tehran.

Mr Lotan’s fascination with the data generated by socialmovements has grown further as the tools to analyse them haveimproved. “Topic modelling” reveals what people are talkingabout; “sentiment analysis” gives an idea of how they feel; “net-work mapping” identifies the most important “nodes”; “visuali-

Tracking protest movements

A new kind of weather

Social media now play a key role in collective action

This is so suddenCumulative percentage of messages on social media

Source: “Political Turbulence: How Social Media Shape Collective Action”, by T. Yasseri et al

#MikeBrown #BlackLivesMatter #EricGarner

#Ferguson #HandsUpDontShoot #ICantBreath(e)

2014 20150

20

50

75

100

FacebookTwitter

2014 20150

25

50

75

100

2014 20150

25

50

75

100

2014 20150

25

50

75

100

2014 20150

25

50

75

100

2014 20150

25

50

75

100

The Economist March 26th 2016 5

TECHNOLOG Y AND POLITICS

2

1

sation” software turns the information into colourful pictures.And using such programs is becoming cheaper all the time.

Last year Mr Lotan turned his attention to the anti-vaccina-tion movement in California, which relies heavily on social me-dia to spread its message. Analysing the hashtags—increasinglyused as the brands of social movements, as in #BlackLivesMat-ter—he found that most anti-vaccination messages came fromonlya dozen Twitteraccounts. After losinga legislative battle, themovement’s leaders changed their message: rather than makingdubious claims about the link between vaccines and autism,many accounts started to present vaccina-tion as a matter of freedom to choose (seethe visualisation on the cover of this spe-cial report, where the dark blue cluster ontop represents messages of that sort).

Mr Lotan is about to turn his hobbyinto a product. Other firms already pro-duce such network maps for money. Oneis Graphika in New York. Its speciality isidentifying communities of interest with-in social networks, finding the most influ-ential members and tracking what theyare talking about. Most of the firm’s cus-tomers are companies such as fashionbrands or media firms, but it also looks at political issues. Its soft-ware revealed, for instance, that during the Maidan protests inUkraine in 2013-14, Russian “spam bots”—programsthatautomat-ically send messages—had a much larger presence in Ukraine’sTwittersphere than tweets by the Russian political opposition.

Such analyses have answered many of the questions askedabout collective action online. They show that social media playa key role under any kind of regime. Sceptics about the Green

Movement in Iran pointed out thatmany ifnotmost tweets werein English and sent by people outside the country. But during theMaidan protests most messages were in Ukrainian or Russian,say researchers at New YorkUniversity’s Social Media and Politi-cal Participation (SMaPP) programme.

Similarly, to find out how Hong Kong’s Umbrella Move-ment evolved, a team at Hong Kong University looked at its pub-lic Facebook pages. For every day when protesters occupied sev-eral public spaces in the city in the second half of 2014, theymeasured how interlinked the pages were. This number turnedout to be a good predictor of the mood in the population at large.

In a widely cited article in the New Yorker in 2010, MalcolmGladwell, a bestselling author, argued that the form of protestspromoted by social media, such as signing online petitions,would neverhave the same impact as “high-risk” actions such asthose taken at the time of America’s civil-rights movements. Butonline “slacktivism”, as other critics have called it (combiningthe words“slacker” and “activism”), can make a difference. Look-

ing at the tweets of several recent protest movements, includingthe Indignados in Spain and the demonstrators at Istanbul’s Tak-sim Gezi Park, researchers at SMaPP concluded that participantsat their periphery were as important as the ones at the centre.

Last but not least, social media can isolate people fromviewpoints they disagree with. Eli Pariser, an internet activist,calls this the “filter bubble”. Researchers at Indiana University’sNetwork Science Institute who analysed links shared on Twitter

Politics in the age of social media is better described bychaos theory than by conventional social science

IT IS MORE than half a century old, but Man-cur Olson’s book, “The Logic of CollectiveAction”, is still hugely influential. In a nut-shell, the late economist argued that largegroups of people will organise only if theyhave some particular incentive: many willsimply “free-ride” on the efforts of others.Are the rules different online?

In recent years a number of academicshave tried to find an answer. One prominentattempt is a book called “The Logic of Con-nective Action”, by Lance Bennett and Alex-andra Segerberg. The authors contend thatwhen people express views online, they donot need to be part of a formal organisation.By sharing links or posting comments, theyare already engaging in political activity. Butthis diffuse political energy has to be bundledto become effective, hence the importance ofrallying cries such as the Occupy movement’s“We are the 99%”.

Yochai Benkler, a law professor atHarvard University and author of anotherinfluential book, “The Wealth of Networks”,

makes a similar argument. Just as far-flungcommunities of volunteer programmers areable to produce useful open-source software,he says, activists can further their cause bybanding together online. The best examplefor such political “peer production”, as hecalls it, was the successful campaign to stopSOPA and PIPA, two controversial bills inAmerica aimed at strengthening intellectual-property rights online. He and his collab-orators tracked the websites of the organisa-tions supporting the campaign and mappedlinks between them at different stages. Theyfound a network that was constantly evolv-ing, with different players taking the lead atdifferent points. “They managed to getthings done in a decentralised way,” says MrBenkler. He concludes that such movementshave become a “new source of power” inaddition to conventional ones, such as old-style media and political parties.

Henry Farrell, of George WashingtonUniversity, and Cosma Shalizi, of CarnegieMellon University, are more interested in the

Connective action

How the internet changes the way people club together

scope the internet might offer for experi-ments to improve democratic structures.These are not nearly as good as they could be,they reckon in a paper entitled “CognitiveDemocracy”. The internet is full of experi-ments in collective decision-making. Themost successful ones can be found in open-source software and content-sharing sites.Their governance structures range frombenevolent dictatorship (as practised byLinux, an operating system) to more decen-tralised organisations (for example, Wikipe-dia, an online encyclopedia).

None of this refutes Olson’s basicpremise that people do not automaticallycollaborate, even if they have a commoninterest. But the internet makes such col-laboration much easier. And in the onlineworld people take action for different rea-sons, argue the authors of another book,“Political Turbulence”: income levels matterless and personalities more. Extroverts, forinstance, are attracted by the prospect thatwhatever they do will be widely noticed.

SPECIAL REPOR T

6 The Economist March 26th 2016

SPECIAL REPOR TTECHNOLOG Y AND POLITICS

2

1

and searcheson AOL, a web portal, showed that the sites reachedfrom social media are much lessdiverse than those reached froma search engine. Pablo Barberá, formerly of SMaPP and soon tojoin the University of Southern California, who examined thepolitical Twitterspheres in America, Germany and Spain, foundthey were indeed polarised, particularly in America.

Random spikesSome researchers have now moved on to building comput-

er models to test theories about the behaviour of online socialnetworks. The spread of information on social media is typically“spiky”, with some posts suddenly becomingextremely popularwhereas others never take off, regardless of the topic.

To track how misinformation travels online, FilippoMenczer and his colleagues at Indiana University developed asystem that can simulate millions of users to reproduce the “spi-kyness” of social media. Even if the “users” are programmed toprefer worthy content, misinformation can go viral when it coin-cides with information overload, which is common online.“People just pass on stuffwithout thinking,” notes Mr Menczer.

A new book entitled “Political Turbulence” gives a taste ofwhere such research might lead. The four authors, most ofwhom work at the Oxford Internet Institute, come to an intrigu-ing conclusion: social media are making democracies more “plu-ralistic”, but not in the conventional sense of the word, involvingdiverse but stable groups. Instead, the authors see the emergenceof a “chaotic pluralism”, in which mobilisations spring from thebottom up, often reacting to events. Online mobilisation can de-velop explosively and seemingly at random. Most online peti-tions, the authors found, attract only a small number of signa-tures, but the successful ones took off in the first few days (seechart earlier in this article). Success does not seem to depend onthe subject matter: similar ones often fare quite differently.

Politics in the age of social media, the authors conclude, isbetterdescribed bychaos theory than byconventional social sci-ence: “Tiny acts of political participation that take place via so-cial media are the units of analysis, the equivalent of particlesand atoms in a natural system, manifesting themselves in politi-cal turbulence.” One day, say the authors, it will be possible topredict and trigger such surges, in the same way that meteorolo-gists have become good at forecasting the weather.

But who will be the political meteorologists? The chancesare that itwill notbe researchers such asMrLotan or MrMenczer.Access to data isgettingharder. ManyfollowersofIS, for instance,

have abandoned Twitter and now use encrypted messaging ser-vices considered safer, such as Telegram or TextSecure. Social-media services are also becoming less generous with their data,both for privacy and commercial reasons.

Only two groups of actors are sure to have good access tosocial-media and other types of internet data. One is the onlinegiants, such as Facebook and Google, which know much moreabout people than any official agency does and hold all this in-formation in one virtual place. It may not be in their commercialinterest to use that knowledge to influence political outcomes, assome people fear, but they certainly have the wherewithal. In2010 Facebook allowed James Fowler, a political scientist at theUniversity ofCalifornia, San Diego, to test the service’s influenceon people deciding whether or not to vote in the elections forCongress. About61m userswere presented with an “I voted” but-ton and shown pictures of friends who had clicked on it. Twosmallercontrol groups received eitherno message or just a noticeabout the elections and how many users had clicked the “I vot-ed” button. The results showed that the Facebookusers who hadseen pictures of their friends were 0.4% more likely to vote thanthose who had not.

The other group ofactors are governments, particularly theauthoritarian kind. Having been caught offguard by online prot-est movements, many are now investing heavily in their web-based propaganda infrastructure. Russian government agencies,for instance, are not just good at setting up social-media bots andother spamming weapons to drown out genuine online dis-

Umbrella power

A force for good or ill

Source: World Bank

E-government provision, by government type, 2015

Political regimeAutocratic Democratic

E-go

vern

men

t se

rvic

e in

dex

0

0.2

0.4

0.6

0.8

1.0

10 5 0 5 10+–

The Economist March 26th 2016 7

TECHNOLOG Y AND POLITICS

2

1

course. They also employ armies of “trolls” to fight on their be-half in Western comment sections and Twitter feeds.

China’s political weathermen are even more sophisticated.Researchers at Harvard University who studied millions of Chi-nese social-media posts found that censors mostly blocked con-tent designed to spur collective action but tended to toleratecomments critical of the Chinese leadership.

The longer-term worry is that the internet and related tech-nologies could strengthen authoritarian governments and maymake it harder for the countries concerned to move towards de-mocracy. In a recent report the World Bank pointed out thatamong non-democratic countries, the most autocratic have in-vested most in e-government services (see chart, previous page).They do this, says the bank, to strengthen control and solve whatit calls the “dictator’sdilemma”: the invidiouschoice between re-stricting the internet, which would hurt economic development,and leaving it unfettered, which could undermine the govern-ment’s power.

Similarly, in a recent paper Espen Geelmuyden Rod andNils Weidmann, both of the University ofKonstanz, find that theinternet tends to grow faster in countries in which regimes aremore concerned about the flow of information. They also arguethat there is no evidence to date “that democracy advances in au-tocracies that expand the internet”.

At least in democratic countries, though, there are some en-couraging signs that at the local level the internet has improvedparticipation in decision-making from the bottom up. 7

MARTIN WALSH, THE mayor of Boston, keeps on top ofwhat is going on in his city. His office is dominated by a

dashboard, a large screen packed with constantly changing snip-pets of text, numbers and charts (pictured, next page). One sec-tion shows the current traffic to the city’s call centre and the per-centage that has been answered within 30 seconds. Next to it is achart tracking the number of potholes filled every day, whichmakes way for a map of Boston’s neighbourhoods coloured ac-cording to how often Mr Walsh has visited them.

But the central piece ofinformation is the “CityScore”, a sin-gle numberto indicate Boston’soverall health. It combines 24 dif-ferent metrics, from crime to Wi-Fi availability, energy consump-tion and grants for the arts. A value above 1means that things aregoingbetter than planned; anythingbelow this, and the mayor islikely to pickup the phone. “Everybody knows that he is lookingat this,” says Daniel Koh, Mr Walsh’s chief of staff, who came upwith the idea for the index. Bostonians can check it online.

CityScore, launched last October, reflects a growing trendamong city governments in America. Led by Boston, Chicagoand New�ork, they have started to use the ever-increasingamounts of data they collect to improve planning, offer betterservices and engage citizens. To speed up the process, the WhiteHouse recently launched a new “smart-city” initiative.

Here we go again, youmight say: itwasonlya fewyears agothat big makers of computing and communications gear madean effort to persuade city halls to buy more of their machines.But this time the push is coming more from the city governmentsand even the citizens themselves. Cities are becomingaware thatdata, and the infrastructure to analyse them, will eventually be-come as important to their citizens’ welfare as the power gridand the transport system.

What most mayors have yet to realise, however, is howmuch their administrations will have to change to be able to getthe best out of these data—and use them to make their citiesmore democratic. More and better data could help governmentsensure that services in poor neighbourhoods are as good asthose in rich ones. Given a city-wide system of sensors, the lead-contaminated water that poisoned poor citizens of Flint, Michi-gan, in 2014-15 would probably have been spotted much earlier.

Pinpointing potholesIn some waysBoston hasbeen a digital pioneer. In 2006 the

previous mayor, Tom Menino, hired the city’s first cabinet-levelchief information officer. He was behind the launch of an appcalled “Citizen Connect” which made it easy for people to reportproblems, for instance by taking a picture ofgraffiti. Another firstwas the creation of an internal innovation team, perhaps bestknown for another app, Street Bump. This collects vibration datafrom moving cars to pinpoint potholes that need to be filled.

Now the city is putting more effort into learning from suchdata. About one-third of its rubbish bins are equipped with solarpanels and sensors that signal when a bin is full, making rubbishcollection more efficient. The city’s data scientists have also ana-lysed online classified ads to identify landlords who cram toomany tenants into their flats. And they are running experiments

Local government

How cities score

Better use of data could make cities moreefficient—and more democratic

SPECIAL REPOR T

8 The Economist March 26th 2016

SPECIAL REPOR TTECHNOLOG Y AND POLITICS

2

1

known as “A/B-tests” that have already become routine online.So far they have tried prioritising buses at traffic lights and in-creasing fines for blocking an intersection, and then used datafrom Waze, a popular navigation app owned by Google, to seehow this affects congestion, a big problem in Boston.

MIT’s Senseable City Lab in Cambridge across the CharlesRiver gives a taste of how much more cities could do with data.Researchers there are working on a cheap package of sensors tobe put on top of street lights, which if widely deployed wouldmake it possible to measure noise and pollution levels almosthouse by house in real time. A project called “Underworlds” en-visages small robots crawling through sewers, collecting sam-ples and perhaps one day analysing them on the spot. This couldreveal things such as what people eat and how many have theflu. “Imagine how many data get flushed down the toilet,” saysErin Baumgartner, one of the lab’s directors. The project is sup-ported by the government of Kuwait, which is looking for waysto measure its people’s excessive intake ofsalt.

However, this sort ofthing is not going to make much differ-ence if the bureaucratic structure of city governments remainsthe same. Most are collections of depart-mental silos that do not communicatemuch with each other, held together bycomplex hierarchies and rules. That mayhave worked when information wasscarce and moved slowly, but now it hasbecome an obstacle. City governmentshave to become more of a coherent whole—a “platform”, asgeeks put it.

This often starts with getting the technology right. City gov-ernments’ computer systems tend to reflect their fragmented na-ture. Information is typically kept in separate databases. Makingthese work together is crucially important, but the task is oftenunderestimated, explains Jascha Franklin-Hodge, Boston’s chiefinformation officer. The city still has more work to do, but mostof its digital information now sits in a “data warehouse”, a bigcomputer system where it can be easily accessed and analysed.

The next thing is better integration of a city’s administra-tion. To be able to improve existing services and develop newones, departments have to work together more closely, says Ste-

phen Goldsmith of Harvard University and co-author of “TheResponsive City”, a new book about urban government. Cityemployees also have to be able to act more independently andbe judged by their results, not have to follow rules slavishly.

Regulation, too, has to be rethought. When informationabout businesses was hard to come by, it made sense to imposeall kinds of rules and regularly check for compliance. But nowthatanalytics can point to likelyviolators, and business practicescan be tracked in real time, such regulation may amount to over-kill. In Chicago inspectors were sent mainly to restaurants whichan algorithm had identified as potential problems.�

et the biggest change will be of another order: cities needto play a more active role as broker of urban data. This meansmore than just sharing reams of their own administrative infor-mation, as many cities around the world already do, says Antho-ny Townsend, a researcherat New

�orkUniversity and author of

a book on smart cities. Municipal governments should becomethe guardians of the local data ecosystem, creating a frameworkthatencouragesothers to share data and offerservices to citizens.They could act, for instance, as a portal for information from util-

ities and online firms, while also protecting privacy and ensur-ing that the algorithms used do not discriminate against particu-lar groups ofpeople.

Some cities are beginning to take on this role. An early ex-ample is Boston’s data-sharing partnership with Waze on reduc-ing traffic congestion. In return for some of the service’s data, thecity is giving it early warning of any planned road closures. Chi-cago, meanwhile, has launched OpenGrid, a website which al-lowscitizensand businesseseasily to visualise public urban datausing online maps.

In New�

ork the Centre for Urban Science and Progress(CUSP) has launched a project called “Quantified Communities”to work out how people could use data generated by increasingnumbersofsensors in theirneighbourhoods. One idea is to mea-sure airquality in differentareasand compare itwith hospitalisa-tion rates for asthma. Constantine Kontokosta, who heads theprojectatCUSP, explains that “we want to define the problem be-fore we decide on the technology—not the other way around.”

Seattle, for its part, has discovered that citizens will insist onstringent protection of privacy. A few years ago it began using awireless police network that could track smartphones, alongwith automatic licence-plate readers. The programme was im-plemented without much public discussion or thought abouthowthe data would be managed. That led to a backlash from res-idents and a hasty about-turn. The city has since adopted de-tailed privacy principles and has just appointed a data-protec-tion officer—a standard requirement in European cities but a firstin America.

It is less clear what cities can and will do to prevent algo-rithms from becoming “Weapons of Math Destruction”, the titleof a forthcoming book by Cathy O’Neil, a blogger and formerquantitative analyst on Wall Street. Critics allege that local policeforces in America are the worst offenders. Their “predictive pol-icing”, which uses algorithms, crime statistics and other data topinpoint “hotspots” where further crimes are likely to be com-mitted, has sometimes proved quite accurate. But it can also leadto unnecessary questioning, excessive stopping and searchingand racial profiling in such hotspots.What the mayor saw

A project called “Underworlds” envisages small robotscrawling through sewers, collecting samples andperhaps one day analysing them on the spot

The Economist March 26th 2016 9

TECHNOLOG Y AND POLITICS

2

1

“TECHNOLOGY IS NEITHER good nor bad; nor is it neu-tral,” said the late Melvin Kranzberg, one of the most influ-

ential historians of machinery. The same is true for the internetand the use ofdata in politics: it is neither a blessing, nor is it evil,yet ithasan effect. Butwhich effect? And what, ifanything, needsto be done about it?

Jürgen Habermas, the German philosopher who thoughtup the concept of the “public sphere”, has always been in twominds about the internet. Digital communication, he wrote afew years ago, has unequivocal democratic merits only in au-thoritarian countries, where it undermines the government’s in-formation monopoly. Yet in liberal regimes, online media, withtheir millions of forums for debate on a vast range of topics,could lead to a “fragmentation of the public” and a “liquefactionofpolitics”, which would be harmful to democracy.

The ups and downs of the presidential campaign in Ameri-ca and the political turbulences elsewhere seem to support Mr

Habermas’s view. Indeed, it is tempting to ask whether all thisonline activism isnotwasted political energy thatcould be put tobetter use in other ways. Indeed, the meteoric rise of many on-line movements appears to explain their equally rapid demise:many never had time to build robust organisations.

But online activism cannot be dismissed. Some move-ments have had real impact, either by putting an issue on the po-litical agenda or by taking over an existing organisation. Without

the Occupy movement, the debate about income inequality inAmerica would be much less prominent. The same goes for theBlackLivesMattercampaign and violence againstAfrican-Amer-icans. In Britain, Jeremy Corbyn and his supporters managed tocommandeer the Labour Party. In America, Donald Trumpseems about to do the same with the Republican Party (thoughwhether he can do it to the whole country remains to be seen).

No going backOnly the most extreme critics want to go back to a time

when the flow of information was controlled mostly by govern-ments and mass media. And the current political turbulencesmay lead to the creation ofservices that calm them down. Earlierthis year, for instance, Change.org, a petition site with nearly140m members, launched Change Politics, which lets any user,including media companies and other organisations, post en-dorsements. The idea is that voters will be able to draw on re-commendations by people they trust, rather than being manipu-lated by political commercials and tweets.

The effect of vast quantities of data is both easier and hard-er to gauge. As this special reporthasshown, pilesof digital infor-mation and the algorithms to analyse them tend to be good forthose in power. Political parties with plenty of money can usethem both to target voters and to discipline recalcitrant candi-dates by cutting off access. Autocratic governments that wereblindsided when the internet took off in the mid-1990s have re-gained their vision. Data can make cities more efficient, but alsomore centralised and controlling.

All this suggests that data and analytics risk slowing downand perhaps even undoing the welcome redistribution ofpowerto ordinary people that the internet seemed to be able to offer.They create “points of control” in what used to be largely an“open system”, as�ochai BenklerofHarvard Universityputs it ina recent article in Daedalus, an American journal. The design ofthe original internet, he writes, was biased towards decentralisa-tion of power and the freedom to act. Along with other develop-ments such as smartphones and cloud computing, he now seesdata as a force for recentralisation that allows “the accumulationof power by a relatively small set of influential state and non-state actors”.

Does this matter? Another law of technology, particularlythe digital kind, is that it is never in equilibrium. Data can em-

power both empires and rebels. DavidKarpf, of George Washington University,expectsa rise in whathe calls “analyticac-tivism”, the title of a forthcoming book ofhis. One example is MoveOn.org, a left-wing advocacy group in America with avoracious appetite for data ofwhich evenmany of its 8m members are unaware.Among many other things, it closelytracks whether people have read themany messages it sends out.

Equally important, digital technol-ogy has a “capacity to surprise”, says Hel-en Margetts of the Oxford Internet Insti-tute (OII). The database politicking

within America’s parties has created room for non-partisan of-ferings. One is NationBuilder, a startup based in Los Angeles. Itsclients get access to a basic national voter file to which they canadd their own data and share it with other campaigns if theywish. “Unlike an organisation which keeps a big central data-base, we don’t have to make a decision on who can use it,” saysJim Gilliam, the startup’s chiefexecutive.

And then there is the blockchain. This technology, a version

Living with technology

The data republic

To safeguard democracy, the use of data should bemade as transparent as possible

Piles of digital information and the algorithms toanalyse them tend to be good for those in power

Even apparently neutral apps such as Street Bump mayhave unintended consequences: the service could give priorityto wealthier neighbourhoods where people can afford smart-phones, leavingpotholes in poorareas unfilled. To avoid such anoutcome, Boston first released the app to its road inspectors, whodrive all over the city. It has also negotiated a deal with Uber, thetaxi-hailing service, to get trip data so that its transport depart-ment can monitor, for instance, how long passengers in poorneighbourhoods have to wait for a car.

The big political question is whether data will simply makecity government more efficient—which in itself is a worthwhilegoal—or whether they will also empower citizens. Susan Craw-ford of Harvard University, co-author of “The Responsive City”,argues thathavingaccess to data will notonlyshowpeople whattheir tax money can achieve, but give them the tools to get in-volved in their city’s affairs.

Others are not so sure. Technology rarely fixes the underly-ing problem but mostly replicates it, says Benjamin Barber, anAmerican political theorist with an interest in local government.“Above all we need smart mayors and smart citizens, not smartcities.” The dashboard in the mayor’s office suggests that in Bos-ton, for now at least, efficiency and control win out. 7

SPECIAL REPOR T

en the chance to check the infor-mation held about them in cam-paign databases.

Transparency over the useof algorithms has its limits.Opening them up for inspec-tion, as some have proposed,can make them lose their valuebecause it will allow them to begamed. Others are so complexthat even their authors do notfullyunderstand howtheyoper-ate. One possibility is to developalgorithms that check on algo-rithms. Researchers at Colum-bia University have built a soft-ware tool called Sunlight toreveal why, say, users of onlineservices are presented with cer-tain ads.

Luciano Floridi, also of theOII, calls for an ethical frame-work for the use of data, muchlike that currently being devel-oped for reproductive technol-ogies. Some companies have al-ready started to move in thisdirection. Google has set up anethics committee forartificial in-telligence. And the British par-liament’s science and technol-ogy committee recently pro-posed the creation ofa national data-ethics council.

The debate about data and politics has only just begun andthese proposals need time to mature. But getting the rules formanaging digital information right is critically important. Societ-ies will have to decide how they want data to be used, in politicsas well as in other spheres. As Alec Ross, a former State Depart-

ment official who now works as an advis-er on technology politics for Hillary Clin-ton’s campaign, puts it in his new book“The Industries of the Future”: “Thechoices we make about how we managedata will be as important as the decisionsabout managing land during the agricul-tural age and managing industry duringthe industrial age.”

Data and politics are likely to be-come ever more intertwined, as science-fiction writers have long forecast. Theymay have got the details wrong, but someof their ideas are nevertheless worth con-sidering. Isaac Asimov, who died a quar-ter of a century ago, before the internettook off, invented a prophetic universeruled by a group of “psychohistorians”who forecast humanity’s future, using aset of complicated equations. To preventpeople from interfering with the predic-tions, they had to keep them secret, butthat in turn created untold complications.The story, like this special report, suggeststhat technology is morally neutral. Dataare neither good nor bad for democracy. Itall depends on how people use them. 7

10 The Economist March 26th 2016

SPECIAL REPOR TTECHNOLOG Y AND POLITICS

Offer to readers

Reprints of this special report are available. A minimum order of five copies is required.Please contact: Jill Kaletha at Foster Printing Tel: +1 866 879 9144 Ext: 168 e-mail: jillk@fosterprinting.com

Corporate offer

Corporate orders of 100 copies or more are available. We also offer a customisation service. Please contact us to discuss your requirements.Tel: +44 (0)20 7576 8148e-mail: rights@economist.com

For more information on how to order special reports, reprints or any copyright queries you may have, please contact:

The Rights and Syndication Department20 Cabot SquareLondon E14 4QWTel: +44 (0)20 7576 8148Fax: +44 (0)20 7576 8492e-mail: rights@economist.comwww.economist.com/rights

Future special reports

Previous special reports and a list of forthcoming ones can be found online: economist.com/specialreports

Business in Africa April 16th International banking May 7thMigration May 14thThe Middle East May 28th

2 of which powers bitcoin, a cryptocurrency, could prove to be abig democratic reset button. It is essentially a new type of data-base that isowned and maintained notbya single actor butby itsusers, who collectively agree to any changes. Such “distributedledgers”, as they are known, could one day become alternativesto big centralised databa�����enture-capital firms have madetheir first bets on such undertakings, including OpenBazaar, apeer-to-peer marketplace. Perhaps one day voter files will bekept in blockchain-like distributed ledgers, which allow citizensto reveal their data only to the candidates they like.

Taming the beastIt would be foolish, however, to base public policy solely

on the hope that some new service or technology will comealongto solve existingproblems. So whatsafeguardsmight be in-troduced to limit the power conferred by data? The most radicalproposal comes from Evgeny Morozov, a technology critic. Hethinks that big companies such as Facebook and Google shouldbe barred from owning certain types of data, such as the key-words users search for, and whether those users have voted inthe past. Instead, this information should belong to the individ-uals concerned and shared only if they so choose��et the politi-cal will to implement such a policy is lacking in much of theworld, says Mr Morozov.

A more practical idea comes from Gavin Starks, the execu-tive director ofLondon’s Open Data Institute. He argues that cer-tain types of data may need to be kept available to all: addressfiles and geospatial information, for instance, are akin to roadsand other public infrastructure and need to be treated in thesame way. “We need to discuss who owns our data infrastruc-ture, what roles the public and private sectors should have, andwhat role we as citizens play,” he recently wrote in a blog post.

Others think that more transparency would help. ZeynepTufekci of the University of North Carolina wants campaignersto be required to publish all the messages they pitch to voters—inthe same way as they are obliged, at least in America, to show indetail how they have spent their campaign money. And EitanHersh of�

ale University recommends that voters should be giv-

The internet, but not as you know it