RESULTADOS Y DISCUSION
3.1. DE LA MATERIA PRIMA
3.2.2. Experimento N° 2: Piqueladode la Piel
Table4reportstheresultsoftheimpactofinvestorattentiononthestockliquidity.AsshowninEq.(3),weregressthe relativebid-askspreadontheSVI,onlinemediacoverage,advertisingexpenditureandasetofcontrolvariables12.Inmodel
I,weincludeonlytheSVIandthecontrolvariables.Capturingtheimpactofactiveattention,thecoefficientoftheSVIis significantlynegative,whichsuggeststhathigherlevelofinvestorattentionreflectedbysearchfrequencyleadstoareduced bid-askspreadandthereforeimprovedstockliquidity.ModelsIIandIIIincorporatepassiveattentionmeasuresincluding Googleonlinenewscoverageandadvertisingexpenditure,whilemodelIVadjuststheclusteredstandarderrorbyapplying thebootstrappingregression.ModelVisbasedonthereducedsampleof6742observationswithnon-missingadvertising expenditure.ThecoefficientsoftheSVIremainnegativeandsignificantacrossthemodels,suggestingthatinvestors’active attentionhelpstoimprovestockliquidity.ConsistentwithGrullonetal.(2004),thecoefficientofadvertisingexpenditure issignificantlynegativeinmodelV.
WealsoreplicatetheanalysisbyreplacingtheSVIwiththechangeinSVI,andpresenttheresultsinPanelBofTable4.In modelI,thecoefficientoftheSVIchangeissignificantlynegative,suggestingthatanincreaseininvestorattentionimproves liquidity.TheresultsremainrobustaftercontrollingforthepassiveattentionmeasuresinmodelsIIandIII,afterapplying thebootstrappingregressionmodeltoadjusttheclusteredstandarderrorinmodelIVandafterdroppingtheobservations withnoadvertisementexpenditureinmodelV.Inuntabulatedtest,wealsoapplyfirmfixedeffectsorreplacethelevelof liquiditywiththechangeinliquidity13asthedependentvariable.Theresultsremainconsistentinthatthesignificantly
negativecoefficientsofthechangeinSVIsupportourpredictionthatincreasedinvestorattentionhelpstopromotestock liquidity14.
Thereareseveralaspectsofliquidity,andbid-askspreadreflectstheinventoryaspect.Stoll(1978)andHoandStoll (1981)arguethatliquiditydependsonfactorsthatinfluencetheriskofholdinginventoryandextremeeventsthatprovoke orderimbalanceandtherebycauseinventoryoverload.Tocompensatetheinventoryholdingcost,dealerswillpurchase sharesatthebidpricebelowthe“true”priceandsellsharesattheaskpriceabovethe“true”price,generatingthebid-ask spread.
Bid-askspread,however,mayfailtocaptureotheraspects.Forexample,GrossmanandMiller(1988)showthatliquidityis alsodeterminedbythedemandandsupplyofimmediacy,andbid-askspreaddoesnotreflectthecostofsupplyingimmediacy tothemarket.Kyle(1985)notestheinformedspeculationaspectofliquiditywhenmarketmakerscannotdistinguish betweenorderflowgeneratedbyinformedtradersorliquiditytraders,theysetthepriceasanincreasingfunctionofthe imbalanceintheorderflow,whichmayindicateinformedtrading.Thissuggestsapositiverelationbetweenthetransaction volumeandpricechange,knownaspriceimpact.Itcanbeconsideredasthefirstderivativeoftheeffectivespreadwith respecttotheordersizeandreflectsthecostofdemandingadditionalinstantaneousliquidity.Amihud(2002)measuresit asthedailypriceresponseassociatedwithonedollaroftradingvolume,calculatedasthedailyratioofabsolutestockreturn todollartradingvolumeaveragedoverallpositivevolumedays.
12AsshowninthePanelBoftheAppendix,wedivideoursamplefirmsintolow-attentionandhigh-attentionsubsamplesbasedonthemedianlevel
oftheSVIandtestthedifferenceinthemeansoftherelativebid-askspread.Theresultsshowthattherelativebid-askspreadissignificantlysmallerin thehigh-attentionsubsample.Whenwefurtherdividethesub-samplesaccordingtomarketcapitalization,intosmallandlargefirms,thedifferencein thebid-askspreadstillexistsinbothtypesoffirm.Thedifferenceismorepronouncedforsmallerfirmsbecausetheyare,ingeneral,lessrecognizedby investors,andthereforemorelikelytobenefitfromincreasedactiveattentionfrominvestors.Theresultsareavailableuponrequest.
13Thechangeinliquidityismeasuredbytherelativespreadinmonthtminustheaveragerelativespreadbetweenmontht−1andmontht−3. 14ThepanelCinappendixalsoreportstheresultsbyapplyingthefirmfixedeffects.Theresultsofthreeoutoffourregressionmodelsarebroadlyin
linewithourmainresults.Petersen(2009)givespossibleexplanationsfortheminorinconsistency:thestandarderrorsclusteredbyfirmareunbiased andproducecorrectlysizedconfidenceintervalswhetherthefirmeffectispermanentortemporarywhilethefixedeffectandrandomeffectsmodelonly producesunbiasedstandarderrorswhenthefirmeffectispermanent.
20 R.Ding,W.Hou/Int.Fin.Markets,Inst.andMoney37(2015)12–26
Table4
Activeattentionandstockliquidity.
PanelA.SVIandrelativebid-askspread
Dependentvariablerelativespreadin%
ModelI ModelII ModelIII ModelIV ModelV
SVI −0.01**(−2.18) −0.01**(−2.17) −0.01***(−2.11) −0.01*(−1.73) −0.01**(−1.87) Ln(numberofnews) 0.01*(1.71) 0.01*(1.8) 0.01(1) 0.01(0.15) Ln(advertising) −0.003(−1.42) −0.01(−0.34) −0.02***(−2.56) Ln(firmage) −0.01(−0.73) −0.01(−0.69) −0.01(−0.70) −0.02***(−2.44) −0.04***(−3.72) ROA −0.01(−0.46) −0.01(−0.41) −0.01(−0.44) −0.03***(−2.72) −0.12***(−5.88) Ln(firmsize) 0.003(0.39) −0.001(−0.17) 0.001(0.16) 0.01(0.43) 0.03***(4.22) 1/shareprice 2.95***(8.04) 2.93***(7.99) 2.94***(7.99) 3.17***(11.58) 3.64***(9.77) Ln(turnover) 0.12***(7.68) 0.12***(7.64) 0.12***(7.69) 0.15***(13.24) 0.16***(8.32) Ln(returnvolatility) 2.47***(84.99) 2.47***(84.95) 2.47***(84.86) 1.93***(87.23) 1.82***(38.39) NASDAQ −0.03(−1.28) −0.03(−1.14) −0.03(−1.14) 0.02(1.43) 0.01(0.11) Year Y Y Y Y Y Industry Y Y Y Y Y
Adjustclustererrors N N N Y Y
Excludemissingadvertising N N N N Y
Observation 14,690 14,690 14,690 14,690 6742
Adj.R2 0.793 0.793 0.793 0.56 0.55
PanelB.ChangeinSVIandrelativebid-askspread
Dependentvariablerelativespreadin%
ModelI ModelII ModelIII ModelIV ModelV
ChangeinSVI −0.24***(−4.30) −0.24***(−4.30) −0.24***(−4.30) −0.19***(−4.58) −0.27***(−6.45) Ln(numberofnews) 0.01*(1.7) 0.01*(1.79) 0.01(0.88) 0.01(0.12) Ln(advertising) −0.003(−1.42) −0.01(−0.16) −0.01**(−2.47) Ln(firmage) −0.01(−0.67) −0.01(−0.63) −0.01(−0.64) −0.02***(−3.23) −0.04***(−4.05) ROA −0.01(−0.39) −0.01(−0.35) −0.01(−0.39) −0.03***(−3.85) −0.11***(−6.11) Ln(firmsize) 0.003(0.38) −0.001(−0.19) 0.001(0.16) 0.01(0.32) 0.03***(3.56) 1/shareprice 2.97***(8.11) 2.95***(8.05) 2.96***(8.06) 3.25***(10.31) 3.74***(10.54) Ln(turnover) 0.12***(7.76) 0.12***(7.71) 0.13***(7.77) 0.15***(17.47) 0.16***(8.3) Log(returnvolatility) 2.48***(85.28) 2.47***(85.24) 2.47***(85.16) 1.93***(89.67) 1.83***(43.63) NASDAQ −0.03(−1.24) −0.02(−1.11) −0.02(−1.11) 0.01(0.94) 0.01(0.2) Year Y Y Y Y Y Industry Y Y Y Y Y
Adjustclustererrors N N N Y Y
Excludemissingadvertising N N N N Y
Observation 14,690 14,690 14,690 14,690 6742
Adj.R2 0.794 0.794 0.794 0.56 0.55
PanelC.Activeattentionandamihudilliquiditymeasures
Dependentvariable
ILLIQ ILLIQMA SD(ILLIQ)
ModelI ModelII ModelIII
ChangeinSVI 0.02(1.01) 0.01(1.4) 0.01(0.22) Ln(numberofnews) 0(0.81) 0(0.81) 0(0.79) Ln(advertising) 0(1.37) 0(1.38) 0(1.36) Ln(firmage) −0.01(−1.16) −0.01(−1.19) −0.01(−1.12) ROA −0.02(−1.33) −0.02(−1.32) −0.02(−1.34) Ln(firmsize) −0.02(−1.48) −0.02(−1.51) −0.02(−1.47) 1/shareprice −0.08(−1.17) −0.1(−1.17) −0.1(−1.24) Ln(turnover) −0.14(−1.52) −0.16(−1.51) −0.16(−1.52) Ln(returnvolatility) 0.1(1.53) 0.11(1.47) 0.11(1.53) NASDAQ 0.04(1.44) 0.04(1.43) 0.04(1.53) Year Y Y Y Industry Y Y Y Observation 14690 14690 14690 AdjR2 0.2423 0.1561 0.2311
PanelAandBofthistablereportstheestimatesfrompanelregressionsrelatingtherelativebid-askspreadtoactiveinvestorattention(SVI).Thesample includestheconstituentsoftheS&P500overaperiodofsixyearsfromJanuary2004toDecember2009.TheSVIisthesearchfrequencyofthestockticker, obtainedfromGoogleTrends.ThechangeintheSVIisthedifferencebetweentheSVIinweektandthemedianvalueoftheSVIduringthepreviouseight weeks.The“numberofnews”istheonlinemediacoveragefromGoogleNews.AdvertisingexpensesareobtainedfromCompustat.Therelativebid-ask spreadisthemonthlyaverageoftheratioofthedailyinsidespreadtothemidpointofthedailyinsidespread,obtainedfromCRSP.Firmageisthenumber ofyearsforwhichthefirmhasbeenincludedinCRSP.ReturnonassetsisconstructedfromCompustatastheannualoperatingincomebeforedepreciation, scaledbytotalassets.Firmsizeisthemarketcapitalization,calculatedastheproductofthetotalnumberofoutstandingsharesandtheannualclosing price.ShareturnoverisconstructedfromCRSPasthemonthlyaverageofthesharevolume,dividedbysharesoutstanding.Returnvolatilityisthemonthly averageofthestandarddeviationofdailyreturns,fromCRSP.NASDAQequals1forfirmslistedontheNASDAQ,and0otherwise.
PanelCpresentstheresultsbasedontheAmihud(2002)liquiditymeasures.ILLIQisthemonthlyaverageforthedailyratioofabsolutereturntothedollar volumeofthestock.ILLIQMAistheratioofthevariableILLIQtoitsmonthlymeanacrossallstocks.SD(ILLIQ)isthemonthlystandarddeviationofILLIQ (LangandMaffett,2011).Wecorrectforstandarderrorsusingclustering.
R.Ding,W.Hou/Int.Fin.Markets,Inst.andMoney37(2015)12–26 21
Toexplorewhethertheactiveattentionofretailinvestorsalsoaffectsotheraspectsofliquidity,wereplacethedependent variablewiththeAmihud(2002)liquiditymeasure(ILLIQ)andtheresultsarereportedinPanelC.ThecoefficientofChange inSVIisinsignificant.WhenweincorporateILLIQMA,estimatedastheratiooftheILLIQtotheaverageILLIQofallstocksin themarket,orincorporatethestandarddeviationofILLIQ(LangandMaffett,2011),thecoefficientsofattentionmeasures remaininsignificant.Theresultssuggestthattheactiveattentioncouldsignificantlymitigatetheadverse-selectiontypeof illiquidity,butnottheprice-impacttypeofilliquidity.Amihud(2002)alsoindicatesthatalthoughbid-askspreadisafiner andbettermeasure,thereisnoonesinglemeasurethatcapturesallitsaspects.AssuggestedbyDaetal.(2011),SVIlargely reflectstheattentionofretailinvestorsbecauseinstitutionalinvestorshaveaccesstoprofessionalinformationvendors.In generalthetradingvolume(indollarterm)ofanaverageretailinvestorislesslikelytobelarge,whichsuggeststhattheir tradingbehaviormighthavediminishedeffectonthestockpricereactiontotrading.