IsGoogledegradingsearch?

ConsumerHarmfromUniversalSearch
MichaelLuca,TimWu,andtheYelpDataScienceTeam1
Abstract
WhileGoogleisknownprimarilyasasearchengine,ithasincreasingly
developedandpromoteditsowncontentasanalternativetoresultsfromother
websites.ByprominentlydisplayingGooglecontentinresponsetosearch
queries,Googleisabletoleverageitsdominanceinsearchtogaincustomersfor
thiscontent.Thisyieldsseriousconcernsiftheinternalcontentisinferiorto
organicsearchresults.Toinvestigate,weimplementarandomizedcontrolled
trialinwhichwevarythesearchresultsthatusersareshowncomparing
GooglescurrentpolicyoffavorabletreatmentofGooglecontenttoresultsin
whichexternalcontentisdisplayed.Wefindthatusersare45%morelikelyto
engagewithuniversalsearchresults(i.e.prominentlydisplayedmapresultson
Google)whentheresultsareorganicallydetermined.Thissuggeststhatby
leveragingdominanceinsearchtopromoteitsinternalcontent,Googleis
reducingsocialwelfareleavingconsumerswithlowerqualityresultsandworse
matches.

Introduction
Universalsearchreferstoasearchenginefeaturethroughwhichthe
searchengineoperatorblendsspecializedsearchpropertiesoftenfrom
1

Harvard Business School and Columbia Law School, respectively. The Yelp data science team consists
of Travis Brooks, Sebastien Couvidat, Daniel Frank, and Will Seltzer. Luther Lowe on the public policy
team also contributed. The work for this paper was financially supported by Yelp.

proprietarydatabasesinpriorityovertheresultsgeneratedbyanorganic,or
algorithmicsearch.PioneeredandwidelyusedbyGoogle,universalsearch,by
itsnature,isaformofselfdealingthatexcludescompetitorsinspecialized
search.Assuch,overthelastdecade,thepracticehasattractedscrutinyfrom
competitionlawenforcersintheUnitedStates,Canada,Europe,andLatin
America.
Initsdefenseofuniversalsearch,Googlehasreliedonasimpleand
powerfulargument:ithasdonenothingmorethancreateabetterproduct.
YoursearchjustgetsbetterandbetterovertimesaidEricSchmidtrecently,
GoogleBerlinweatherandyoullnolongergettenbluelinksyoullgetthe
weatherforecastforthenextfewdaysatthetopresult,savingyoutimeand
effort.2 Whenchargedthatitsuniversalsearchunfairlyexcludescompetitors,
Googleargues,inlegalterms,thatuniversalsearchisaprocompetitive
3

innovationthatservestheinterestsofusers.
Throughaseriesofonlineexperiments,thispaperrevealsamore
complexrealityandsharplyquestionstheassertionthatGooglesimplementation
ofuniversalsearchhasbeenuniformlygoodforconsumersorprocompetitive.
OurfindingssuggestthatGoogleisinsomeinstancesactuallymakingits

Eric Schmidt,
The New Grndergeist
, Google Europe Blog (Oct 13, 2014, 10:28
AM), http://googlepolicyeurope.blogspot.com/2014/10/the-new-grundergeist.html
3 Communication From the Commission: Guidance on the Commissions
Enforcement Priorities in Applying Article 82 of the EC treaty to Abusive
Exclusionary Conduct by Dominant Undertakings, 2009 O.J. (C 45) 7, 12.
2

overallproductworseforusersinordertoprovidefavorabletreatmenttoGoogle
content.
Ontheonehand,thereareinstancessuchasdisplayingtime,or
presentingacalculatortoarithmeticproblemswhereuniversalsearchhasbeen
deployedinamannerthatlikelybenefitsconsumers.Afterall,tothequery2+2
thereisnobetterorworseanswer,butsimplytherightanswerorwronganswer
andtheanswercanbeeasierorhardertofind.Therefore,acalculator
presentedviauniversalsearch,aswedemonstratebelow,ispreferredbyusers.
Incontrast,considerthequestionwhichdoctorshouldIseeforthis
condition?Clearly,thereisawidevarietyofcontentthatonemightprovideto
facilitateauserssearchforadoctor.Competingrankings,ratings,andreviews
rangingfromUSNews&WorldReporttoZocDoctoHealthGradesinvestheavily
indevelopingsuchcontent.Inthissituation,Googlescontentmaybemoreor
lessusefultousersthanothercontent.IfGoogleprovidesfavorabletreatmentto
GooglecontentinaworldinwhichZocDocismoreusefultousers,thiscomes
witharealcosttoconsumers.Ourresultsprovideempiricalevidencethat
universalsearchhasinsomesituationsbeendeployedinamannerthat
degradesthesearchproductandharmsGooglesusers.Suchconducttherefore
cannotbedescribedasprocompetitive.

Localintentbasedsearchescomprisethelargestsinglecategoryof
search,representingroughlyonethird4 oftotaldesktopsearchvolume,andover
onehalfofmobilesearch.5 Today,Googlepresentsuserswithlocalsearch
resultsthatareamixtureofitsorganicresultsalongwithauserinterfaceobject
knownastheLocalOneBox.6 TheOneBoxtypicallyincludesalistofseven
businesspinspopulatedbyexclusivelyqueryingGooglesproprietarylocal
product,Google+LocalthissetofsevenbusinesspinsisattachedtoaGoogle
Map.

Mike Blumenthal,
Ed Parsons: 1 in 3 Searches at Google are Local
, Understanding Google My Business
& Local Search (Nov. 14, 2012, 9:26 AM),
http://blumenthals.com/blog/2012/11/13/ed-parsons-1-in-3-searches-at-google-are-local/
5
Greg Sterling,
Research: 56 Pct. of On the Go Searches Have Local Intent
, Search Engine Land (May 6,
2014, 2:07 PM),
http://searchengineland.com/research-56-percent-go-searches-carry-local-intent-190655
6
Danny Sullivan,
Meet the Google OnBox, Plus Box, Direct Answers & the 10-Pack
, Search Engine Land
(Sept. .28, 2009, 6:12 PM),
http://searchengineland.com/meet-the-google-OneBox-plus-box-direct-answers-the-10-pack-26706
4

However,Googlesorganicsearchemploysameritbasedalgorithmthatcan
easilybeusedtoidentifybettercandidatestopopulateitslocalsearchboxes,
enablingthecreationofanalternativeversionofthesearchengineresultspage.
Basedonasampleof2,690Internetusers,thisstudydemonstratesthatusers
wouldbemorelikelytoengagewithlocalspecializedsearchresultsifGoogle
weretoreplaceitsproprietaryanswersinuniversalsearchwithresultsdrawn
fromthewholewebbasedonthesamemeritbasedalgorithmthatitusesto
populateorganicsearch(asopposedtobeingexclusivelydrawnfromGoogle+).
Thestudyisconductedbyrandomlydisplayingoneoftwosetsofsearchresult
screenshotstothousandsofinternetusers:onesetofusersseestheGoogle
universalsearchpageasitiscurrentlyconstitutedthesecondsetseesan
alternativeversionofuniversalsearch.Inthealternativeversion,abrowser
pluginnicknamedFocusontheUserLocal(FOTUL)7hasqueriedthirdparty
reviewsitesandrankedthem,usingGoogle'sownorganicalgorithm,according
towhichsitedeliversthebestormostrelevantinformationforthecontentin
question.

See
Focus On The User - Local,
Github,
https://github.com/FocusOnTheUserLocal/FocusOnTheUserLocal
. (last visited June 23, 2015)
(explaining the Focus On The User - Local as a widget formed in collaboration by local search
companies to educate about consumer harm brought about when Google+ is artificially promoted in
local search).
See also
Focus On The User,
http://focusontheuser.eu/
(last visited June 23, 2015).
7


Theresultsdemonstratethatconsumersvastlypreferthesecondversionof
universalsearch.Stateddifferently,consumersprefer,ineffective,competitive
results,asscoredbyGooglesownsearchengine,thanresultschosenby
Google.ThisleadstotheconclusionthatGoogleisdegradingitsownsearch
resultsbyexcludingitscompetitorsattheexpenseofitsusers.Thefactthat
GooglesownalgorithmwouldprovidebetterresultssuggeststhatGoogleis
makingastrategicchoicetodisplaytheirowncontent,ratherthanchoosing
resultsthatconsumerswouldprefer.
Ineconomicterms,wecanunderstandGooglesutilityasareductionof
searchfrictionsbyhelpingtocreateamatchbetweenusersandtheirpreferred
goodsorservices.Inatwosidedmarketanalysis(suchasRochetandTirole),
Googleistheintermediaryorplatformthatputstogetherbuyerswithsellersof
desiredgoodsorservices.Byprovidingsuboptimalresults,Googlesconduct

eliminatessomeoftheaforementionedwelfaregains,henceunderminingthe
valueofsearch.Assuch,whenGoogleknowinglydegradesitssearchsoasto
harmitscompetitors,theimpactcanbefeltbybothsidesofthemarketbythe
consumerswhodontgettheresults(andproducts)theyvaluemosthighly,and
themerchantswhomightotherwisehavesoldtothem.
Thedemonstrationofconsumerharmis,wethink,animportant
conclusionstandingonitsownthatshouldinfluenceanycompetitionlaw
analysis.However,itintersectswithseveralwidelyrecognizedcriteriafor
enforcementactionincompetitionlaw.First,whateverthegeneralutilityof
universalsearch,wehaveshownthat,asimplementedinsomesegments,
universalsearchisharmfulbothtomerchants,consumersandcompetitorswhile
lackingredeemingqualities.Assuch,insomeimplementationsitmaybe
categorizedasaspeciesofnakedexclusioninotherwords,conductthat
excludescompetitorswithoutanycountervailingbenefit.8
Alternatively,Googlesconductcanbeunderstoodastheknowingneglect
ofalessrestrictivealternativeforachievinglegitimategoals.9Googles
developmentofuniversalsearch,ingeneral,canbeacceptedasanimportant
innovationthatcanimproveconsumerwelfare.Butitseizesonthefactthat,as
implemented,Googleappearstohavechosentodosoinawaythatneglectsan
obviousandclearlymoreeffectivealternative,resultinginharmtoconsumers,

See Eric B. Rasmusen, J. Mark Ramseyer, & John S. Wiley, Jr.,

Naked Exclusion
, 81 Am. Econ. Rev. 1137
(1991) (describing exclusion as naked when the conduct unabashedly excludes rivals without offering
any efficiency justification).
9
See Scott Hemphill, Less Restrictive Alternatives in Antitrust & Constitutional Law working paper
(2015).
8

merchants,anditscompetitors.Importanttothisconclusionisevidencethat
Googleissacrificingahigherqualityandpotentiallymoreprofitableproductin
favorofamoreexclusionaryoption.Thatfortifiestheintuitionthattheconductis
suspect.
Fromthispaperonethingshouldbeabundantlyclear.Theeasyand
widelydisseminatedargumentthatGooglesuniversalsearchalwaysserves
usersandmerchantsisdemonstrablyfalse.Instead,inthelargestcategoryof
search(localintentbased),Googleappearstobestrategicallydeploying
universalsearchinawaythatdegradestheproductsoastoslowandexclude
challengerstoitsdominantsearchparadigm.

I.

IndustryBackgroundandtheDevelopmentofUniversalSearch
Sinceitsinception,theWorldWideWeb,nowrepresentingmorethan4.7

billionlinkedwebpages,hasdependedforitsutilityontoolsforlocatingdesired
information.Thetoolsforfindinginformationonthewebhavealwayshadan
importantinfluenceonnearlyallaspectsofeconomicactivityandinnovationon
thewebanditsconnectiontoMainStreetsmallbusinesses.Theabilityof
buyersandsellerstoconnectcandomuchtodeterminewhichbusinesses
succeedandwhichfail,whatinnovationscatchon,andwhichflounder.They
alsohaveapreeminentroleininfluencingthespeechenvironmentcenteredon
theweb.
Thehistoryoftoolsforfindinginformationonthewebisoneof
technologicalevolution.Intheearlydaysoftheworldwideweb,simplelistsof
linksanddirectoryservices,(liketheearlyYahoo!)wereused,incompetitionwith
early,primitivesearchengines.Eventually,intheearly2000s,thegeneral
10

searchengine(whichsearchestheentireweb)
becamethetoolofchoiceof
whichGoogleisthepreeminentexample.Generalsearch,today,remainsa
dominanttoolbywhichwebsitesandtheirusersfindeachother,andbuyersand
sellerarematched.
Theevolutionofthewebsinformationretrievaltoolsdidnotendwith
generalsearch.Overthe2010s,justasGooglesgeneralsearchreplaced

10

More precisely, everything that makes itself available or crawlable with [a

standard internet DNS address and uses the HTML protocol] and does not
specifically tell the search engine not to make its content searchable.

directoryservices,itsgeneralsearchbegantofaceitsownchallengers.The
mostimportantchallengehascomefrom
specialized
searchthatis,search
enginesthatdeliverinformationbasednotonsearchingtheentireweb,butrather
aspecificcategoryofinformation.Prominentexamplesincludethesearchfor
booksonAmazon,planeticketsonKayakorOrbitz,hotelreviewson
TripAdvisor,orforalocaldoctoronYelporZocDoc.Asitstands,specialized
searchisnotindirectcompetitiontoGoogleinstead,itrepresentsathreattothe
generalsearchparadigm(inthesensethatNetscape,incombinationwithother
11

middlewarewereaparadigmthreattotheWindowsOSparadigm).
Eric
Schmidtcapturedthisrealitywhenherecentlysuggestedourbiggestsearch
competitorisAmazon.12
Googlehas,overthedecade,aggressivelyfoughtthethreatfrom
specializedsearchinthefollowingways.First,beginninginthe2010s,itmade
copiesofthemostsuccessfulofthespecializedsearchengines.Manyofthese
aresimpleclonesofcompetitors,likeKayakorYelp(yielding,forexample,
GoogletravelGoogleshopping,Google+Local,etc.).Toimprovetheir
popularity,theearlierversionsofGooglesclonesborroweddatafromthe
originalsinparticular,earlyversionsofGooglelocaltookthereviewscollected
byYelpandTripAdvisorandincorporatedthemintoGooglesproduct.

11

The two paradigms are not direct competitors in the market sense, but rather in
the Schumpeterian sense (in the sense that Netscape & middleware were not a
direct threat to Microsoft Windows, but as a platform).
See
United States v.
Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001).
12 Schmidt,
The New Grndergeist,
supra note 1.

Overall,withnotableexceptions,likeGoogleMaps,thecloneshavenot

provedaspopularorsuccessfulwithusersastheoriginals(forexample,
GooglesclonestypicallysufferfrompoorsearchrankingsinGoogle'sorganic
searchresults).TodealwiththischallengeGooglehasleveragedthepowerof
itsdominant,generalsearchenginetogiveadvantagestoitsotherwise
unpopularspecializedsearchproducts.Theprimarymeansfordoingsoiswhat
iscalledtheuniversalsearchortheOneBox.Universalsearchoperatesby
incorporatingthegeneral(tenbluelinks)andspecialized(OneBox)search
paradigmsintothesameuserinterfaceonthesearchengineresultspage.
GoogledeploystheOneBoxwhenitdetectskeywords,orthefactthatanatural
searchreturnsaspecializedcompetitorasaleadingresult.


(AnexampleofGooglesUniversalsearchforthequerypancakeshowsablendoforganic
resultsalongsideothersourcesofdata:Google+Local,GoogleMaps,KnowledgeGraph,
GoogleNews)

Forexample,asearchforrestaurantsSanFranciscomighttriggerYelpasthe
firstnaturalresult,causingGoogletodeployaLocalOneBox.13TheOneBox,
withafewexceptions,ispopulatedbyGooglesownclonesofitscompetitors
specializedsearchservices(e.g.,Google+Localortravel).

Universalsearch,initsoperationisinherentlyexclusionary,foritusesthe
dominantgeneralsearchenginetodiverttrafficfromGooglesspecialized
competitors(Expedia,Yelp,etc.)toitsownversionsofthosecompanies.That
facthas,unsurprisingly,ledtoscrutinybothbyEuropeandAmericancompetition
regulators.
Googlesstrategyfollowsawellknownhistoricpattern.Technological
monopolists,facingthreatsfrominnovativecompetitors,oftenengageinvertical

Memorandum, Recommendation to the Federal Trade Commission on Google

Antitrust Litigation 24, Page 70 (Aug. 8, 2012), available at

http://graphics.wsj.com/google-ftc-report.
13

selfdealingtoprotecttheirmonopoly.Aprominentexamplewasthedefensive
selfdealingbytheAmericanphonemonopolistAT&T,whichledtoprolonged
litigationoverthe1970sand80s,whereAT&Twasaccusedofexcludingboth
handsetandlongdistancecompetitors,despitefederalregulationsrequiring
interconnection.14AnotherhistoricalparallelcanbeseenintheUSand
EuropeanMicrosoftlitigationofthe1990s,whereMicrosoftwasaccusedofusing
thepowerofitsoperatingsystemmonopolytoexcludevariouscompetitorsin
favorofitsownclonesoftheirproducts.Themostprominentvictimofthat
15

conductwasNetscape.
AccordingtoGoogle,aprincipaldifferencebetweentheearliercasesand
itscurrentconductisthatuniversalsearchrepresentsaprocompetitive,
userservinginnovation.Bydeployinguniversalsearch,Googleargues,ithas
madesearchbetter.AsEricSchmidtargues,ifweknowtheansweritisbetter
forustoanswerthatquestionso[theuser]doesnthavetoclickanywhere,and
inthatsenseweusedatasourcesthatareourownbecausewecant

S
ee
AT&T, 38 F.C.C.2d 213 (1972); AT&T, 64 F.C.C.2d 1 (1977); Communications
Satellite Corp., 56 F.C.C.2d 1101 (1975); United States v. AT&T, 461 F. Supp 1314
(D.D.C. 1978); United States v. AT&T Co., 524 F. Supp 1336 (D.D.C. 1981); United
States v. AT&T Co., 552 F. Supp 131 (D.D.C. 1982), affd sub nom. Maryland v. United
States, 460 U.S. 1001 (1983).
15
See
United States v. Microsoft Corp., 56 F.3d 1448 (D.C. Cir. 1995); United States v.
Microsoft Corp., 147 F.3d 935 (D.C. Cir. 1998); United States v. Microsoft Corp., 253
F.3d 34 (D.C. Cir. 2001); Case COMP/C-3/37.792 Microsoft, 2004 O.J. (C 290).
14

engineeritanyotherway.16Theobjectofthispaperistotestthetruthofthat
proposition.

Testimony of Eric Schmidt, Executive Chairman, Google Inc., before the Senate

Committee on the Judiciary Subcommittee on Antitrust, Competition Policy, and

Consumer Rights (September 21, 2011).
16

II.

MethodologyandSummaryofResults
A.

A/Btestingincompetitionanalysis

TheA/Btestingmethodologyisaformofstatisticaltestingthatiswidely
usedforwebdevelopment.PioneeredinmajorpartbyGoogle,italsoservesas
aparticularlyusefultoolfordetectingconsumerharmincompetitionanalysis,as
itcreatesadirectmeansforobservingtheimpactofdesigndecisionson
consumerbehavior.A/Btestingpresentsarandomsampleofuserswithoneof
twovariantsofawebsiteandthenrecordstheusersresponsetothe
presentation.TheA/Btest,alsoknownasasplittest,istypicallyusedin
commercialwebdevelopmenttodeterminetheeffectivenessofcompeting
designsforsomegoalorobjective,suchasinducinguserstoclickon
advertising.Butthatsamemethodcanalsoservecompetitionanalysisby
allowingotherinterestedpartiestotesttheimpactonconsumersofdifferent
designs.Assuchitallowsadirectandrealistictestofimpactsonconsumers.
InasimpleimplementationtodemonstratetheusageofanA/Btestinthe
searchcontext,wetestedtheeffectsofpresentinguserswithtwodifferent
searchengineresultpagesinresponsetoanarithmeticproblem(42people
havegivenyou$54.YoudecidetousethesearchengineGoogletodetermine
howmuchtotalmoneyyouhavebeengiven.YoutypecalculatorintoaGoogle
searchbarandcomeuponthefollowingpage.Wheredoyouclickfirst?)
VariantApresenteduserswithmerelybluelinks,followedbybluelinksearch
resultsVariantBpresentedaninteractivecalculatoronthesearchengineresult

page.Whenpresentedwiththeinteractivecalculatorabovetheorganiclinks,
over90%oftheusersclickedonit,suggestingagaininconsumerutilityfromthe
presentationofthecalculatorOneBox.Thisresult,incidentally,isconsistent
withGooglesclaimsabouttheefficiencyofuniversalsearch,but,asweshall
see,onlytellspartofthestory.

Position1:
31
Position2:
63
Position3:
6

AnswerBox(Calculator):
96
Positions13:
8

Turningtothepresentcase,howmightA/Btestingmeasuretheeffectsof
asearchenginedesignhereuniversalsearchonconsumerwelfare?A
searchengineisaplatformorintermediarywhoseutilitycomesfrommatching
buyersandsellersatminimumcosts,whethermeasuredintimeorotherefforts.
Universalsearch,therefore,servestheinterestsofconsumersifithelpsto
facilitatematchingbetweenbuyersandsellersatminimalcost.
Theteststhatfollowcenteronuniversalsearchasimplementedtoanswer
questionsthattriggeralocalspecializedresult,suchasnearbyrestaurantor
pediatriciannyc.Wefirstposeaquestiontousers,andfollowbyrandomly
showingtheuseroneoftwosetsofresults.17Thefirst,thecontrol,isexclusively
populatedbyGooglesownproducts,andreflectstheresultscurrentlycreatedby
Google.Thevariant,oralternative,populatestheuniversalsearchresultsby
relyingonGooglesorganicalgorithmtodeterminewhatmightbethebestsites
forprovidingtheinformation.Thisvariant,hereincalledFocusontheUser
LocalorFOTUL,isdescribedinthedetailedmethodologysectionbelow.Users
arerandomlypresentedwitheitherthecontrolorthevariant.
WeconductourexperimentonUsabilityHub,18anonlineplatformthatis
specificallygearedtowardtestingwebsitedesignsbeforeawebsiteisintroduced
toamarketthesurveyrespondentsweresourcedfromMechanicalTurk.This

17

For example: You're visiting Louisville, Kentucky for the first time with a friend.

Your friend suggests you both get a cup of coffee. You type "coffee louisville ky" into
Google and click "Search", which brings you to the following screen. What do you
click first?
18

See UsabilityHub,
http://www.usabilityhub.com
(last visited June 24, 2015).

makesanappealingcontextforthisexperimentforseveralreasons.First,it
providesalargersubjectpoolthanwewouldbeabletoobtaininatypicalsocial
sciencelaboratory.Second,thisistheexactplatformthatbusinessesusetotest
websites,suggestingtheexternalvalidityoftheresults.
Thereisgrowingacknowledgmentthatonlinelabormarketsprovidea
compellingsettinginwhichtoinvestigaterealworldproblems.Forexamplesof
fieldexperimentsinonlinelabormarkets,seeGilchrist,Luca,andMalhotra
(forthcoming)andPallais2014.Moreover,Hortonetal2011ranaseriesof
classiceconomicsexperimentsonMechanicalTurk(asimilar,butless
specialized,market)andfindthattheresultsarequalitativelysimilartofindings
fromaphysicallaboratory.Thissuggeststhatwemightexpectfindingsto
generalizeacrosssettings.
Nonetheless,onemightstillbeconcernedthatresultsfrom
MechanicalTurkandUsabilityHubdonotgeneralizetoactualGooglesearches.
WhilewedonothaveaccesstointernalGoogledata,wedohaveaccessto
internalYelpdata,whichweusetocrossvalidateaseriesofexperimental
findingsaboutYelpfromUsabilityHub.Whilethereweredifferencesbetweenthe
clicksurveysandactualsearchresults,wefindthatclicksurveysareboth
directionallycurrentandprovidereasonableestimatesofactualuser
engagement.Thisprovidesfurtherevidenceoftheexternalvalidityofour
experimentaldesign.

B.

Results

Basedonclicksurveysof2,690users,thisstudydemonstratesthat
Internetuserswouldbeapproximately45%morelikelytoengagewithlocal
specializeduniversalresultsiftheresultsaredrawnfromGooglesmeritbased
algorithmicprocessusedtopopulateorganicresults,comparedwithGoogles
currentlocalsearchresults.
Theclicksurveyrevealedthat32%ofuserswouldclickonGoogles
currentlocal7packoflocalresults,whenpromptedwithaneedthatwould
triggeralocalquery,andshowncurrentresults.Bycontrast47%ofusers
clickedonthesame7packpresentation,inthesamescenarios,whenthe
7packwasfilledwiththeFOTULresults,usingGooglesmeritbasedalgorithm
rd
torankresultsfrom3
partyreviewsites.

Anearly50%increaseinCTRisimmenseinthemodernwebindustry,
andissurprisinggiventheabsenceofanymodificationtotheuniversalsearch
userinterface.Wecanbecomemoreconfidentintheresults,andtheuser
reactioncanperhapsbeunderstood,byexaminingthealternativesthatwere
presentedtotheusers.TheFOTULalternative,sinceitbringsincontentfrom
reviewecosystemslargerthanGooglesown,fillsthelocal7packswithresults
thathavemorereviewcontentthattheGooglelocalresults.Thepresentationof
theresultsmakesthereviewcountvisibleasaclearindicatorofthequalityand
depthofcontent.

Thequeryforpediatriciannycyields31reviewsfromGoogle+.

Thesamequeryforpediatriciannycyields719reviewswhentheFocus
ontheUsersoftwareisactivated.

Users,unsurprisingly,areattractedtothiscontentwhenattemptingto
carryoutthetasksetoutbeforethem,suchasfindingahotelorpediatrician.
TheseresultsstronglysuggestthatGooglespresentimplementationofuniversal
searchismakingconsumersworseoffthantheobviousalternativepresented
here.Usersareengagingatalowerratewiththeuniversalresults,which
suggeststheywouldpreferthealternativeversionpresented.Statedsimply,

whenitcomestolocalsearch,Googleispresentingitsuserswithadegraded
versionofitssearchengine.
ThesuggestionthatGooglepurposelyandconsciouslydeployeduniversal
searchtomaximizedamagetoitscompetitors,evenattheexpenseof
consumers,ismoreoverbackedbyevidencecompiledbytheFederalTrade
CommissionBureauofCompetitionoverthecourseofitsinvestigation.Asthe
Commissionconcluded,EvidenceshowsthatGooglesoughttoincreasesuch
"triggering"ofUniversalSearchresultsnotonlytoprovideuserswiththe"right"
19

answertotheirqueries,butalsotodrivetraffictoGoogleproperties.
Wenowconsiderinmoredetailsomeofthetheoriesofconsumerharm
thataresupportedbytheseresults.

III.

TheoriesofHarm
Beginninginthe2000s,economistshaveusedthephrasetwosided

marketstodescribethoseenvironmentswherebuyersandsellersarebrought
togetherbyacompanyactingasanintermediary,orplatform.Asdescribedby
JeanCharlesRochetandJeanTirole,insuchatwosidedmarket,theeconomic
utilityoftheplatformisabyproductofthematchingofthebuyersandsellerson
bothsidesofthemarket.20 Forexample,theutilityofabazaarliesinmatching
buyersandsellersadatingapplicationlikeTindercreatesutilitybysuccessfully
19

Memorandum, Recommendation to the Federal Trade Commission on Google

Antitrust Litigation 24 (Aug. 8, 2012), available at
http://graphics.wsj.com/google-ftc-report.
Jean-Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets,
J. Eur. Econ. Ass'n
990 (2003).
20

matchingcouplestheutilityofatravelplatformismatchingpeoplewithflights,
andsoon.
Thisanalysisprovidesthefirstpartofahelpfulandsimplewayof
understandingtheutilityofGooglesearch.Websearchingcomesclosetobeing
aparadigmaticexampleofthetwosidedmarketintermediary.Ontheoneside
ofthemarketwehaveconsumerswithpreexistingpreferencesforgoodsand
services(Europeansoccergames,vinyasayoga,NewYorkstylebagelsandso
on.)Ontheothersidearemerchantsandwebsiteoperatorswhoseek
consumers(tosimplifytheanalysis,weignoreadvertisers,orassumethatthey
arealsoamongthoseseekingamatch).Searchcreateseconomicwelfareby
matching
consumerswiththeobjectsoftheirpreferences.Inthelanguageof
twosidedmarkets,thesearchengineservesastheplatformorintermediary,
21

whomatchesbuyerswithsellersofdesiredgoodsorservices.
Search,stated
moresimply,generatesutilitywhensomeonefindsayogastudiotheydecideto
join,locateswheretobuyticketstoasoccergame,identifiesagoodpediatrician
fortheirsickchild,andsoon.Inthismanner,bymatchingbuyersandsellers,
searchdirectlygenerateseconomicwelfare.Theflipsideofunderstandingthe
utilityofsearchisthatapurposefullydegradedsearchproductcanthereforedo
economicharm,particularlyifitiswidelyreliedupon.Forcompleteness,wenow
considerfivetheoriesbywhichdegradingofsearchcreatesharmtoconsumers
andmerchants.

21

WelfareLoss.
Wehavedescribedtheutilityofasearchengineinterms
ofmatchingbuyersandsellersandthereductionofsearchcostsinthatprocess.
EllisonandEllison22(2009)demonstratethatfirmscanengageinobfuscationto
makeitharderforcustomerstoacquireinformationinanattempttomaintain
marketpower.Weshowthateventheplatformitselfmayhavetheincentiveto
increasesearchfrictionsfortheirownbenefit.Becausetheresultingqualityand
volumeofmatchesisworsewhensearchfrictionsareincreased,thereisadrop
intotalwelfare.Thiswelfarelossisthensplitbetweenconsumersand
merchants.

Thereare,tobemoreprecise,severalspeciesofharmcaused.First,
someconsumersmaysimplynotfindwhattheyarelookingforinthetimethey
have,andwillgiveup,yieldingsomenumberofunconsummatedtransactions.
Second,someconsumerswillbe,infact,determinedenoughtoeventuallyfind
theirdesiredtarget,butsimplysuffergreatersearchcostsintheprocess.A
thirdkindofharmariseswhenabuyerendsuppatronizingabusinessorother
serviceproviderwhowould
not
havebeentheirfirstchoice,butforthedegrading
ofthesearch.Consider,forexample,aconsumerwhoismisdirectedandends
upatabadrestaurantortheparentswhoarelookingforatopnotch
pediatrician,butbecauseofsearchdegradation,patronizeasubparpractitioner.
Theharmcausedbysuchmisdirectionwhenitoccurs,willvary,butis
undeniableintheaggregate.Thepointissimplythatadegradedsearchengine

22

Search, Obfuscation, and Price Elasticities on the Internet

, Econometrica, 2009, Vol. 77, No. 2

willinvariably,ascomparedtoitsalternative,yieldsomeconsumerharmfrom
misdirection.
SearchAdvertisingMonopolyMaintenance.
Ourresultssuggestthat
Googlehaschosenapaththatexcludesitsspecializedsearchcompetitorsatthe
expenseofitsusers.Theresultistoweakennascentcompetitorstothegeneral
searchparadigm.Whileitisbeyondthescopeofthispapertoprovethat
Googlesdominanceofsearchadvertisingallowsittochargesupracompetitive
pricesforsearchadvertising,assumingitdoes,theexclusionofcompetitors
likelysustainsthoseelevatedprices.Suchconductmightbedefensibleif,as
Googleclaims,itsexclusionaryconductwasjustifiedbyprocompetitivebenefits
forconsumers.However,asthispaperhasshown,Googlesimplementationis
insomeareasactuallyharmfultoconsumers.Assuch,themaintenanceof
inflatedpricesconstitutesaformofeconomicharmthatthecompetitionlaws
weredesignedtoremedy.
InnovationHarms.
ThepatternbywhichGoogleusesgeneralsearchto
excludeandharmcompetitorsinspecializedsearchposesseveralthreatsto
innovation,ofwhichtwomaybehighlighted.
Itisimportanttonotethatmostofthesuccessful,pioneerspecialized
searchoperatorsearnedtheirsuccessthroughinvestmentandinnovation.Yelp
madedeepinvestmentsinthesuccessfulcultivationofofflinecommunitieswhich
encouragedprolificcreationofdetailedreviews,andalsodevelopedimportant
technologiestoweedoutvariousformsoffraudandspam.Kayakdeveloped

newwaysofpresentingflightinformationandincorporatedpredictionalgorithms
thatsuggestthebesttimetobuyaticket.TripAdvisorsucceededwhereothers
hadfailedinachievingacriticalmassofreviewsofhotelsandattractionsfor
muchoftheglobe.
Seriousharmtofutureinnovationiscausedbyreducingtheincentivesof
existingandpotentialverticalstoinvestintheinnovativeanddisruptive
technologiesofspecializedsearch.Ifitisunderstoodbyentrepreneursand
innovatorsthatanyfirmthatreliesonaspecializedsearchwillfaceaneffortby
Googletocloneitsproductandbeleveragedbythepowerofadominantsearch
engine,theshadowcastbyGooglessearchenginebecomeslongindeed.Just
asMicrosoftwasabletodampeninnovationinsoftwarethatitmightincorporate
intoWindows,sotooGooglecanusetheterroroflinkingspecializedsearchesto
generalsearchtodiscourageinvestmentinspecializedsearchproductsthat
mightultimatelychallengeitsdominance.
Second,Googlesconductmaycreateharmbyslowingtheevolutionof
searchtechnology.Aswehaveseen,thehistoryofthewebhaswitnessed
marketalteringimprovementsininformationlocationtechnologyarisingeach
decadeorso(TheriseofYahoo!,Google,etc).Tothedegreethatuniversal
searchdelaysorpreservesthegeneralsearchparadigmovertheriseof
specializedsearchalternatives,itmaybeunderstoodasthepotentialslowingor
blockageofSchumpeteriancompetitionforthemarketthathasbeena
trademarkofthehightechandinternetindustriesformuchofthelast40years.

Thetendencymaybeparticularlyimportantinthemobileenvironment,where
generalsearchhasbeenlesswellestablishedandpresumablyweakerthan
specializedsearch.
Tobesure,theveryintroductionofuniversalsearch,asGooglehas
highlighted,isitselfaformofinnovation.However,everything,onceagain,
dependsonimplementation.Totheextentthatuniversalsearchisimplemented
inamannerthatbenefitsbuyersandsellersitsvaluecannotbecontested.To
theextentitisdeployedtodamagecompetitorsattheexpenseofconsumers,it
representsnorealinnovationatall.
SpeechandSelfExpression.
TheWorldWideWebhasbeencelebrated
overthelastdecadeforitswidespreadpromulgationofspeechandotherforms
ofselfexpressioncapturedinthephraseusercreatedcontent.Thatphrase
reflectsinamultitudeofformsrangingfromblogs,usercreatedvideos,reviews
offilms,products,orrestaurants,andsoon.Moregenerally,Googleoperatesin
theinformationindustries,wherethecostofmonopolymustnotbemeasuredin
dollarsalone,butalsoinitseffectontheeconomyofideasandimage,the
23

restraintofwhichcanamounttocensorship.
Searchenginesarewidelyunderstoodaskeymediatorsofthewebs
speechenvironment,giventhattheyhaveapowerfulimpactonwhogetsheard,
whatspeechisneglected,andwhatinformationgenerallyisreached.Itistelling
thatincensorialregimes,searchenginesareinvariablythetargetsofstrict

23

Tim Wu, The Master Switch: The Rise and Fall of Information Empires 69 (2011).

24

governmentcontrols.
AsprofessorJeffreyRosenhaswritten,Googlehasa
25

particularlycentralroleinthisprocess.
SearchengineslikeGooglehave
substantialinfluencetothepointthat[u]nderstandingfreespeechinAmerica
hasbecomeamatterofunderstandingthebehaviorofintermediaries,whether
26

motivatedbytheirownscruples,law,orpublicpressure.
Thedecisionsmadeaboutsearch,andinparticular,thedecisionto
selfdealattheexpenseofotherentitiesontheweb,haveimplicationsbothfor
webspeakersandalsolisteners,orusers.ThemorethatGoogledirectsusers
toitsowncontentanditsownproperties,themorethatspeakerswhowrite
reviews,blogsandothermaterialsbecomeinvisibletotheirdesiredaudiences.
Similarly,thoseuserswhomightwanttosampleabroader,morediverserange
ofopinionswill,unlesstheyundertakemoreefforts,findthemselveswithamore
constrainedrangeofviews.Thisfitswiththegeneralimplicationsofvertical
integrationinaninformationindustrythatselfdealingtendstoyieldamore
centralizedandnarrowavailabilityofviewsascomparedtothebroaderand
decentralizedpresentationofviewpointsthathascharacterizedthewebsinceits
infancy.

S
ee
Jack Goldsmith & Tim Wu, Who Controls the Internet? (2006)
See

Jeffrey Rosen,
The Deciders: The Future of Privacy and Free Speech in the Age
of Facebook and Google
, 80 Fordham L. Rev. 1525 (2012).
26 Tim Wu, Is Filtering Censorship? The Second Free Speech Tradition, Governance
Studies at The Brookings Institute (Dec. 27, 2010), available at
http://www.brookings.edu/~/media/research/files/papers/2010/12/27-censorsh
ip-wu/1227_censorship_wu.pdf;
see also
Jeffrey Rosen,
Googles Gatekeepers
, N.Y.
TIMES MAG. (Nov. 28, 2008), available at
http://www.nytimes.com/2008/11/30/magazine/30google-t.html).
24
25


IV.

DetailedMethodology
Thetestingproceduredescribedhereinconsistedofthreeseparate

efforts.Firstly,theFOTULsoftware,whichgeneratedanalternativesetoflocal
searchresults,usingGooglesownalgorithm.Secondly,aclicksurveytotest
userreactiontoFOTUL,comparedtouserreactiontoGoogleslocalresults.
Finally,anattempttounderstandthedegreetowhichtheclicksurveyresultsare
indicativeofrealworlduserbehavior.
A.

TheFOTULSoftware
SoftwarewasdevelopedintheformofaChromeBrowserExtension

calledFocusontheUserLocal(referredtohereinasFOTUL)designedto
detectLocalOneBoxes.UpondetectingthepresenceofaLocalOneBoxin
Googlesearchresults,FOTULconductedaseparatesearchinthebackground
includingmajorthirdpartylocalreviewwebsitesseparatedbyORstatements
(e.g.site:yelp.comORsite:zocdoc.comORsite:tripadvisor.comOR)Thefirst
tenpagesofresultsareanalyzedforlinkstobusinesspagesappearingon
popularthirdpartyreviewwebsitessuchasYelpandTripAdvisor.Googles
universeoflocalreviewcontentonthesubdomainplus.google.comisalso
included.
FOTULextractedandrankedtheseresultsaccordingtoacombinationof:
(a)Googlesorganicranking
(b)averagestarrating

(c)quantityofreviews
Theextracted,orderedliststhenreplacedthedefaultcontentsoftheLocal
OneBox,withanewsetofresultsthatGooglesalgorithmselectedasrelevant,
andthathadhighreviewcontent.Theseresultscouldcomefromlocalreview
websiteslikeYelp,TripAdvisorandZocDocorfromGooglesownecosystemof
localreviewsonplus.google.com.
InthiswayFOTULwasabletopresenttheuserwithalivecomparisonof
theGooglesexistinglocalresultsandanalternativemethodofpresentinglocal
results,basedonthealgorithmthatpowersGoogleorganicsearch.Because
Googlesearchresultpages,localcontent,anddomainschangeoften,this
methodologymaynotworkinalllocalesoratalltimes,howeveratthetimeof
thisstudy,thisworkdidproduceacomparisonofthecontentasdescribed.
B.

ClicksurveyTestingofFOTUL
AtotaloffivedistinctvarietiesofLocalOneBoxeswereidentified.While

FOTULwasdisabled,screenshotsweretakenofthesearchresultstoestablisha
controlgroup.FOTULwasthenenabledandusedtogenerateexperimental
variationswithineachofthefivevarietiesofGoogleLocalOneBoxes.
Clicksurveyswerepresentedtoasampleof2,690Internetusers,sourced
usingMechanicalTurk.Respondentswerepromptedwithahypotheticalquestion
basedoncontextualinformationderivedfromeachillustrationandaskedto

answerbyclickingontheillustration.

Therelativeperformance,asmeasuredbytheclickthroughrate(CTR)ofeach
LocalOneBoxwasanalyzedandcompared.

Anexamplefromtheseriesoftestsrunshows182clickson7packregion
poweredbyGoogle+(iestatusquoGoogle).

Anexamplefromtheseriesoftestsrunshows241clickson7packregion
variantpoweredbyFOTUL.


Thefollowingcalculationswereusedtodeterminerelativeperformanceof
theExperimentalGroupagainsttheControlGroup
OneBoxClick
(TotalClicksOtherClicks)

= P erformanceScore

PerformanceScoreExperimental
PerformanceScoreControl

= P erformanceScoreRatio

CTRonOneBox+Map

Google+Local

32%

1,121

FOTULpowered

47%

1,442

ComparingtheexperimentalgroupsgivesaPerformanceRatioof1.44ora44%
increaseinengagementwhenuserswereshownFOTULpoweredresults.In
ordertodeterminethestatisticalvalidityofobservedvaluesofthePerformance
ScoreRatio,95%confidenceboundswerecomputedusingthemethodology
describedin[
http://www.jstor.org/stable/2531405
]producingaboundof[1.303,
1.594].

C.

ClickTextSurveyValidation
ClicktestsurveysonsitessuchasUsabilityHubareacommonlyusedtool

intheindustry,anditiswidelyunderstoodthatsuchtoolsprovideuseful

guidance,butarenotasaccurateapredictorofuserbehaviorasdirectA/B
testingonaliveplatformwiththerealusercommunityinquestion.
Factorssuchasintentoftheuserinthemoment,demographicsofthe
studiedpopulation,andthecontextofthelivesiteallcreatedifferencesbetween
theresultsofusabilitytestingsitesandliveA/Btests.Ofcourse,inthisparticular
case,onlyGoogleisinapositiontodoareal,liveA/Btestonitsusers,soa
relevantquestionbecomestowhatextentarethesetestresultsreflectiveof
reality?
Wegainconfidenceinthevalidityofusingaclicksurveytodraw
conclusionsaboutactualinternetbehaviorbyexaminingA/BtestsrunonYelp
internettraffic,andcomparingthesewithclicksurveysdesignedtomimicsimilar
experimentalconditions.Thiscomparisonshowsthatchangesinbehaviorina
clicksurveydoindicatethedirectionalityofchangesinactualinternetbehavior.
Specifically,Yelpinternettrafficwasexposedtoacontrolcondition
involvingahiddensetofsearchfiltersversusanexperimentalconditioninvolving
anexposedsetofsearchfilters(seeimagebelow).Aclicksurveywasalso
createdwhichreproducedtheseconditionsandwasadministeredtoasetof

7,062usersinthesamemannerastheFOTULstudydescribedabove.

Theimageaboveisanexampleofsurveyperformed:fortherestaurants
nearSanFrancisco,CAquery,userswereshownthesearchresultswithfilters
hidden(leftpanel)orvisible(rightpanel).ForbothYelptrafficandthe
clicksurvey,userengagementwiththemapwasmeasuredandissummarized
below.

Map
engagement

Yelp

UsabilityHub

Filtersshown

X%

57131

0.12*X%

3531

Filtershidden Y%

530

0.127*Y%

3531

CTRRatio

1.29

1.22

TheactualCTRonYelpishiddenbecausethedataisproprietary,howevernote
thatwhileClickteststudyCTRwasmuchsmallerthanthatoftheYelplivetraffic,
thechangeduetotheexperimentalconditionremainedquitesimilar:1.22vs
1.29.Thisagreementofthemagnitudeanddirectionoftheuserresponsetothe
experimentacrossdifferentexperimentaltechniquessuggeststhattheclicktest
techniquecandeliverrelevantinsightstorealworlduserbehavior.
Differencesinthecontextofaclicksurveyascomparedtoactualinternet
usageclearlyaffectuserbehavior.Thiscomparisondoesnotaimtoestablish
thatthebehaviorofclicksurveysisidenticaltothebehaviorofactualinternet
users,butinsteadshowsthatwhilesuchdifferencesobviouslyexist,the
directionalityofthedifferenceispreservedand,forexample,positivechangesin
engagementinaclicksurveyaremirroredbypositivechangesinactualinternet
traffic.Thus,anobservedclicksurveystudychangecanbeusedasevidenceof
anactualinternettrafficchangeinthesamedirection.
ClicksurveysareusedhereasasubstituteforaliveA/BtestonGoogle
withFOTULasanexperimentalconditioncomparedtoacontrolconditionusing
currentsearchresults,becauseGoogledoesnotprovidethesedata,whichcould
definitivelyshowthesuperiorityorinferiorityofthecurrentlocalsearchalgorithm.
Theraisondetreofclicksurveysistoprovideinsightintouserbehaviorwhen

liveA/Btestingdataisnotavailable,andthususingthemforthispurpose,while
perhapsnovelinthiscontext,alignswithindustrypractices.
Indeed,intheabsenceofinternaldatafromsitessuchasGoogle,
clicksurveysappeartobeausefulanddirectionallyaccuratewaytoestimate
userreactionstoproposedchanges.Wesuggestmoregenerallythatthelow
costandhighspeedofA/Bclicksurveysmaygenerallyserveasausefuland
accuratetoolforcompetitionenforcementauthoritiesintheirinvestigationsof
anticompetitiveconductasitprovidesadynamicmeasurementtoolinthese
fastmovingmarkets.

V.

LegalCriteriaofActionableExclusion
Ademonstrationofconsumerharmisrelevanttonearlyanycompetition

lawtheoryonemightcaretoinvoke.Wethereforethinktheconclusionthat
Googleisknowinglydegradingitssearchattheexpenseofconsumersstandson
itsownasancrucialfact.However,weheredevelopthreewaysofthinking
abouttheinformationwhichmightproverelevanttothequestionastowhether
anenforcementauthorityoughttakeaction.
A.

NakedExclusion,NeglectofLessRestrictiveAlternatives&

SacrificeofProductQuality
Inanycompetitioncasecenteredonexclusionaryconduct,authorities
facethefollowinghardquestion:Whenshouldconductthatexcludes
competitorsbeexcusedbyvirtueofthefactthatitmayalsobeefficientor

beneficialforconsumers?Howcanprocompetitiveefficienciesbeweighed
againstanticompetitiveexclusion?Theproblemshowsupwithgreatfrequency
incasesarisingunderSection2oftheShermanActorPara102oftheTreatyon
27

theFunctioningoftheEuropeanUnion.
Americanjudgeshavebeenparticularlysensitivetothischallenge,and
havewarnedofthedangersofthefalsepositivethatistosay,condemnation
ofpracticesthatmight,onbalance,begoodforconsumersortheeconomy.As
theAmericanSupremeCourthasstressed,tocondemnexclusionaryconduct
thatisprocompetitivemightchilltheveryconductthattheantitrustlawsare
28

designedtoprotect.
TheEuropeanCommissionisalsowaryofthisconcern,
andallowsthejustificationofconductleadingtoforeclosureofcompetitorson
thegroundofefficienciesthataresufficienttoguaranteethatnonetharmto
29

consumersislikelytoarise.
Asaresponsetothisconcern,courtsandcommentatorshaveisolated
scenarioswhereactionincasesofexclusionismostclearlywarranted.We

S
ee
Sherman Antitrust Act, 15 U.S.C. 2 (2000). See
also
Consolidated Version of
the Treaty of the Functioning of the European Union, art. 102, Oct. 26, 2012, 2012
O.J. (C 326) 47.
28 Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 575 (1986). The
challenge is augmented in software or technologies cases where exclusion is often
accomplished by product design, for competition authorities have sometimes been
inclined to defer to the designer. In the American Microsoft case, for example, the
court pointed out that In a competitive market, firms routinely innovate in the hope
of appealing to consumers, sometimes in the process making their products
incompatible for those of rivals; the imposition of liability when a monopolist does
the same thing will inevitably deter a certain amount of innovation. United States v.
Microsoft Corp., 253 F.3d 34, 65 (D.C. Cir. 2001).
29 Communication from the Commission, supra note 4.
27

considerthreehere:nakedexclusion,neglectoflessrestrictivealternatives
andsacrificeofproductquality.
Thefirstisthecategoryofnakedexclusion,wheretheexclusionof
competitorsissimplynotjustifiedbyanyrealefficienciesorbenefitsfor
consumers(eveniftheymaybeclaimed.).AsSusanCreightonputsit,such
30

conductisthatwhichislikelytohaveonlyanticompetitiveeffects.
Among
31

manyclassicexamplesisthe
LorainJournal

case,whereanewspaperrefused
32

toprintadvertisementsfromcompaniesthatpatronizeditsrivals,or
AlliedTube
,
wherethemakersofaformofsteelpipeconspiredtopreventplasticpipesfrom
beingacceptedbyastandardsbody.
Thesecondcategoryisonewhereadominantfirmforgoesanobvious,
lessrestrictivealternativecourseofconductthatwouldbeequally,ormore
effectiveinservingtheprocompetitivegoalsarticulated.Therelianceona
lessrestrictivealternativeanalysis,ascommentatorspointout,servesseveral
purposes.Foronething,itservesasanaidtothebalancingofprocompetitive
efficiencieswiththeharmsofexclusion.AsHerbertHovenkamphaswritten,
givencomplexbalancing,firstandforemost,theantitrustdecisionmakermust

See

Susan A. Creighton, Dir. Bureau of Competition, Fed. Trade Commn, Remarks

at the Charles River Associates 9th Annual Conference: Current Topics in Antitrust
Economics and Competition Policy (Feb. 8, 2005), available at
https://www.ftc.gov/sites/default/files/documents/public_statements/cheap-excl
usion/050425cheapexclusion.pdf.
31 Lorain Journal Co. v. United States, 342 U.S. 143 (1951).
32 Allied Tube and Conduit Corp. v. Indian Head, Inc., 486 U.S. 492 (1998).
30

33

lookforlessrestrictivealternatives.
Second,asScottHemphillpointsout,the
useoflessrestrictivealternativesmayalsoserveasimportanttoolfor
smokingoutanillicitmotive.34
InEuropeanpractice,anindispensabilityanalysismirrorstheAmerican
lessrestrictivealternativesjurisprudence.AstheCommissionhaswritten,a
firmthatoffersaprocompetitiveefficiencymustshowthatitsconductis
indispensabletotherealizationofthoseefficiencies:theremustbenoless
anticompetitivealternativestotheconductthatarecapableofproducingthe
35

sameefficiencies.
Theindispensabilityorlessrestrictivealternativesanalysisaimstoidentify
thefollowingsituation.Consideradominantfirmwhoclaimsthatits
exclusionarypracticesservesalaudable,procompetitivegoal.Itfacestwo
equallyeffectivestrategiesfordoingso:strategyAandB,butstrategyAismore
exclusionaryofitscompetitors.Inthatcontext,thechoiceofAissuspiciouson
itsface.ButmattersbecomemuchworseifstrategyAis,moreover,notequally,
butactually
less
effectiveinachievingthestatedprocompetitivegoalsinthat
casesuspicionmaybereplacedwithoutrightincredulity.Thechoiceofaless
effectivebutmoreexclusionaryalternativestronglysuggestsboththeintentand

33

Herbert Hovenkamp & Areeda Phillip, Antitrust Law: An Analysis of Antitrust

Principles and Their Applications, 2111c (2nd ed. 2005).
34 C. Scott Hemphill, Less Restrictive Alternatives in Antitrust and Constitutional
Law (working paper 2015) , available at
http://scholarship.law.berkeley.edu/cgi/viewcontent.cgi?article=1031&context=la
w_econ
.
35
Communication from the Commission, supra note 4.

implementationofananticompetitiveandconsumerharmingcourseofconduct
thatshouldattractseriousscrutiny.
Third,enforcersmayconsider,asevidenceofactionableexclusion,the
sacrificeofprofitorproductqualitysoastodamagecompetitors.Whena
dominantfirmdegradesitsownproductssoastodamagecompetitors,itisoften
reasonablyinferredthatthegoalofsuchconductisnot,infact,innovation,but
themaintenanceofdominance.Here,Googleissacrificingqualityandprofitson
itssearchplatformtoexcluderivalstoitslocalproduct.Suchsacrificesareoften
linkedtoachangeinanexistingcourseofdealing.
Theideaofasacrificeinproductqualityorofprofithasanimportant
pedigreeincompetitionlaw.IntheEuropeanCommissioncaseBritishMidland
36

v.AerLingus
,forexample,regulatorsweregreatlyconcernedthatAerLingus
sacrificedabetterproducttheissuanceofinterlinedairlineticketsforthe
evidentpurposeofslowingarisingcompetitor(BritishMidland).Atissuewasan
existingpracticewherebytheairlines,whosharedaroute,issuedaticketthat
wouldbehonoredbyeitherairline.TheCommissionnotedthatthepractice
wasbothpreferredbyconsumersandrevenuegeneratingforallinvolved.
WhenAerLingusbegantorefusetointerlinewithBritishMidland,itconcluded
thatthedenialofinterliningwasintendedtodamagecompetitionandwouldharm
consumers.Asignificantnumberofpassengersconsiderthepossibilityto
changeticketsandorganizecomplexjourneysonasingleticketasnecessary

36

Commission Decision No. 92/213/EEC, 1992 O.J. (L 96) 34 (EC).

wrotetheCommission,andarefusaltointerlinewillhavetheeffectofdiverting
manyofthesepassengersawayfromthenewentrantairline.Inthisrespect,a
refusaltointerlineaffectsinparticularthewellinformedbusinesstravellerswho
requirefullyflexibleticketsandwhomakeadisproportionatelylargecontribution
totherevenueofthenewentrant.

37

Thebasicpremiseofthesacrificeanalysisisarecognitionthatthateven
thoughexclusionaryconductcanbeambiguousinitseffectonconsumers,
wheretheexclusioncreatesademonstrablylowerqualityproductitbecomes
inherentlysuspicious.AnAmericantakeonsuchconductcanbefoundinthe
AspenSkiing
case,wherethedominantskiingcompanyintheAspenarea
purposelydestroyedapreviouslypopularallmountainpassfornoreasonother
38

thandoingdamagetoitscompetitor.
Thestrategyhastheconsequenceof
bothharmingconsumersanddamagingacompetitor,andassuch,bearssome
resemblancetotheconductatissuehere.
B.

AsApplied

OuranalysisandresultssuggestthatGooglesimplementationof
universalsearchinthelocalcategorysetsofwarningsigns,whetheritis
consideredthroughaninstanceofnakedexclusion,neglectofalessrestrictive
alternative,orasacrificeofproductquality.Thatshouldleadenforcerstotreat
theimplementationofuniversalsearchwithsuspicion.

37
38

I
d.
at 40.
Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585 (1985).

Nakedexclusionisanappropriateconclusionifconsideringuniversal
searchnotasawhole,butspecificimplementations.Thispaperdemonstrates
that,inlocalsearch,theeffectofGooglesOneBoxonconsumersisclearly
negative.Theremaybeotherexamples,discoverablethroughfurthertesting.
Itmaybeappropriatethen,toconsiderimplementationsparticularlythose
servingupnonfactbasedinformationofuniversalsearchasharmfultoboth
consumersandcompetitorsassimplytheexerciseofnakedexclusion.
Alternatively,onemightchoosetoviewuniversalsearchmorebroadly,
andcredittheinstanceswhereUniversalSearchdoesbenefitconsumers(such
asitsmorerapidresponsestoinquiriesforrawfactualdata).Inthatcase
Googlesimplementationofuniversalsearchmorereadilyfitswithinthecategory
offorgoingmoreeffective,lessrestrictivealternativeinfavorofastrategythat
bothdoesmoretoexcludecompetitorsandlesstoserveusers.Assuch,the
testgivesstrongreasontosuggestthatGoogleisopportunisticallydeploying
universalsearchtopreventanythreatstothegeneralsearchengine,atthe
expenseofbothcompetitionandconsumers.
Third,itshouldnotbedifficulttoconcludethatGoogleissacrificing
productqualityinthepursuitoftheexclusionofitscompetitors.Asdemonstrated
above,consumersactuallypreferredauniversalsearchthatincludesboth
Googleanditscompetitors,justastravelerspreferredtheinterlinedairline
tickets.Assuch,justasintheAerLinguscase,thereisnopersuasiveand

39

legitimatebusinessjustificationforitsconduct.
Nor,astheCommissionhas
recognized,canGooglesdesiretogrowitsownspecializedsearchproduct
competitorsserveasajustificationfortheconduct.Thedesiretoavoidlossof
40

marketshare[does]notmakethisalegitimateresponsetonewentry.
ThestatementstendtosupportthetheorythatUniversalSearchwasat
minimumatwoedgedinstrument.Ithasbeen,attimes,deployedinawaythat
benefitsconsumers.Buteverythingalsopointstotheuseofthesame
instrument,inatleastsomeareas,toexcludecompetitorsattheexpenseof
consumers,afindingthatmustbetakenseriously.

39
40

I
d.
at 41.
Id.