您的当前位置：首页 Dividend changes and stock price informativeness

Dividend changes and stock price informativeness

来源：化拓教育网

JournalofCorporateFinance35(2015)1–17

ContentslistsavailableatScienceDirect

JournalofCorporateFinance

journalhomepage:www.elsevier.com/locate/jcorpfin

Dividendchangesandstockpriceinformativeness

AmedeoDeCesaria,⁎,WinifredHuang-MeierbabManchesterBusinessSchool,UniversityofManchester,BoothStreetWest,ManchesterM156PB,UnitedKingdomEssexBusinessSchool,UniversityofEssex,WivenhoePark,ColchesterCO43SQ,UnitedKingdom

articleinfoabstract

Weinvestigatehowprivateinformationinstockpricesimpactsquarterlydividendchanges.Weﬁndthatthepositiverelationshipbetweenpastreturnsandcurrentdividendchangesstrengthenswhenreturnsconveymoreprivateinformation.Thisﬁndingisrobusttotheuseofseveralpriceinformativenessmeasuresandtheinclusionofmanagerialprivateinformationandstockover-valuationmeasures.Managersseemtolearnnewinformationfromstockpricesthattheyusewhendecidingontheirdividendpolicy.Thisstudyhighlightsprivateinformationinstockpricesasanimportantdeterminantofdividendpolicyandcontributestotheliteratureontherealeffectsofﬁnancialmarkets.

Articlehistory:

Received27November2014

Receivedinrevisedform10August2015Accepted12August2015

Availableonline20August2015JELclassiﬁcation:G35G14

Keywords:

Dividendchanges

StockpriceinformativenessStockreturns

1.Introduction

Acommonly-heldbeliefthathasinformedmostﬁnanceliteratureisthenotionthatcorporateinsiderspossessmorecompleteinformationontheirﬁrmthancorporateoutsiders.Whilethisisareasonablebelief,wecannotruleoutthepossibilitythatoutsidersmayholdinformationunknowntoinsidersthatcouldallowthislattergrouptomakebetterandmoreinformeddecisions.Forinstance,foreigninvestorsmaybeneﬁtfrominformationspeciﬁctotheircountry'seconomythatamultinationalﬁrminvestinginsuchcountrydoesnotpossess.Firmsfrequentlyhireconsultantstogaininformationthatmaybevaluableinthedecision-makingpro-cess.Moregenerally,professionalfundmanagersmaybebetterthanindustrialﬁrmsatanalyzingpublicly-availableinformationandproducingusefulprivateinformation.

Oneimportantchannelthroughwhichmanagerscangainnovelinformationontheirﬁrmistheobservationofthelevelanddy-namicsoftheﬁrm'svaluationonsecondaryﬁnancialmarkets(Bondetal.,2012;DowandGorton,1997;SubrahmanyamandTitman,1999).AsemphasizedbyDowandGorton(1997)andChenetal.(2007),stockpricescanconveyprivateinformationpossessedbytradersonthedemandforaﬁrm'sproducts,theﬁrm'sinvestmentopportunities,thecompetitiveenvironmentinwhichitoperates,andtheimplicationsofthepastdecisionsoftheﬁrm'smanagers.Incontrast,traders,unliketheﬁrm'smanagers,maylackdetailedinformationonthetechnologiesusedbytheﬁrm.Thissuggeststhatstockmarketsmayprovideinformationthatiscomplementarytothatheldbycorporateinsiders.Thus,ﬁnancialmarketsmaynotsimplybeasideshowinthattheymayinﬂuencedecisionmakersintherealsideoftheeconomythroughaninformationalchannel.

Theexistingliteratureondividendpolicy,whileextensive,hasessentiallyoverlookedtherolethatprivateinformationconveyedbystockpricesmayplayasadeterminantofcashdividends.Managerscantakeintoaccountbothpublicandprivateinformationwhendecidingdividendpayments.Itiswell-establishedintheliteraturethatpublicly-availableinformationonvariablessuchasproﬁtability,growthopportunities,andﬁrmsizeinﬂuencesdividendpolicy(e.g.,FamaandFrench,2001;GrullonandMichaely,

⁎Correspondingauthor.Tel.:+44(0)16130603.

E-mailaddresses:amedeo.decesari@mbs.ac.uk(A.DeCesari),w.huang-meier@essex.ac.uk(W.Huang-Meier).

2A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

2002).Yet,thelinkbetweentheprivateinformationconveyedbystockpricesanddividendpolicyhasreceivedlittleattention.Themainobjectiveofthispaperistoﬁllthisgapintheliterature.Inotherwords,ourstudyprovidesimportantinsightsintothedetermi-nantsofdividendpolicybytakinginformedtradingintoaccount.Tothisend,weinvestigatewhetherthedegreeofinformativenessofstockmarketvaluationsaffectstherelationshipbetweenpastabnormalchangesinstockprices(i.e.,pastabnormalreturns)andcur-rentdividendchanges.

Wearguethatpastabnormalstockreturnsmaybeconsideredbymanagerswhendecidingwhether,inwhichdirection,andtowhatextentcashdividendpaymentsshouldberevised.Thisisbecauseunexpectedchangesinassetvaluesareinformativeinthattheyshouldreﬂectnewsaboutcashﬂowsand/ordiscountrates(e.g.,CampbellandShiller,1988;ChenandZhao,2009).Newsonlarg-erthanexpectedcashﬂowsand/orunexpecteddecreasesindiscountratesshouldgeneratepositivemarketreactionsandmayleadmanagerstoincreasecashdividends.Thus,consistentwithpreviousevidence(e.g.,Brownetal.,2007;Chemmanuretal.,2010;Grullonetal.,2002),weexpectcurrentdividendchangestobepositivelyrelatedtolaggedabnormalstockreturns.However,reportingthispositiverelationshipisnotsufﬁcientevidenceofmanageriallearningofprivateinformationconveyedbystockpricechanges.Marketinvestorscan,infact,reacttobothnovelpublicandprivateinformation.Inordertodistinguishbetweenthesetwotypesofinformationweconsiderseveralmeasuresofprivateinformationinstockprices.Assumingthatmanagersuseprivateinformationlearnedfromabnormalrevisionsinmarketvaluations,weexpectcurrentdividendchangestobemorestronglyrelatedtopastabnormalreturnswhenmoreprivateinformationisimpoundedintostockprices.

Despitethis,managerscouldbereluctanttoincreasedividendpaymentsevenafterlearningnovelprivateinformation.AshighlightedbyJensen(1986)amongtheothers,dividendscanbecostlytomanagersbecausetheyreducetheirfreedomtopursueempire-buildingstrategiesandengageinotheractivitiesthatmainlybeneﬁtthemselvesattheexpenseoftheirshareholders.Cashdividendscanalsoreduceﬁnancialﬂexibility,potentiallyleadingtounderinvestmentcostsandcausingﬁnancialdistressandmana-gerialjoblosses(BlauandFuller,2008;Bonaiméetal.,2014).However,thereisevidencethatshareholderscansuccessfullyputpres-sureonmanagersandforcethemtoberesponsivetotheirdemandsfordividends,atleastinstronginvestorprotectionenvironments(AllenandMichaely,2003;DeAngeloetal.,2008).Forinstance,LaPortaetal.(2000)showthatstrongerminorityshareholderrightsareassociatedwithhigherdividendpayments.Moreimportantly,managersarelikelytoalsoconsiderthebeneﬁtsthataccruetothemwhendividendsarepaid.Theycouldmakedividendpaymentstoshareholdersinordertoreduceagencycostsandboostﬁrmvalue.Ahigherﬁrmvaluationwouldbeneﬁtmanagersbydecreasingthelikelihoodofinterventionsanddisciplinaryactionsbyshare-holdersandmakinghostiletakeoverslessprobable(AllenandMichaely,2003;DeAngeloetal.,2008).1Moreover,executivecompensationisoftenlinkedtoequityvaluationsandmanagerswithlargeequityinvestmentsintheircompaniesmaybeeagertoreceivecashdividendstodiversifytheirportfolios(AllenandMichaely,2003;Brownetal.,2007).Whenweighingthecostsandbeneﬁtsofcashdividends,newsoflargerthanexpectedcashﬂowsand/orunexpectedreductionsindiscountratesshouldmakemanagersmorewillingtoincreasedividendsmainlyowingtothegreateravailabilityofcash,higherﬁnan-cialﬂexibility,anddecreasedﬁnancialdistressrisk.

Ourexpectationsaretestedbyusingasampleofquarterlydividendchangesovertheperiod1962–2010fornon-ﬁnancialandnon-utilityUSindustrialﬁrmswithshareslistedonNYSEandAMEX.Foraparticularﬁrm,thevariabledividendchangeisdeﬁnedasthedifferencebetweenthecurrentandthepreviousquarterlydividendpayment,dividedbythepriorquarterlydividendpayment.Theaverageabnormaldailyreturnovertheperiodbetweenthecurrentandthepreviousquarterlydividendisusedasproxyforthelaggedabnormalrevisioninthevalueoftheﬁrm'sstock.

Consistentwithpreviousevidenceandwithourpredictions,wereportthatthelikelihoodofadividendincreaseandthemagni-tudeofadividendchangearebothpositivelyrelatedtolaggedabnormalstockreturns.Incontrast,higherpastabnormalreturnsmakeadividendcutlesslikely.Moreimportantly,weﬁndthatthedegreeofinformativenessofstockprices,asmeasuredbyﬁrm-speciﬁcstockreturnvariation(Roll,1988),strengthenstherelationshipbetweenlaggedabnormalstockreturnsanddividendchanges.Abnor-malrevisionsinthevalueofastockaremorestronglypositively(negatively)associatedwithfutureincreases(decreases)indivi-dendswhenthemarketvaluationofthestockcontainsmoreprivateinformationthatmanagerscanexploit.Theserelationshipsarerobusttotheuseofalternativeprivateinformationmeasures,namelytheilliquidityratiobyAmihud(2002),thetradingmeasuredevelopedbyLlorenteetal.(2002),andtheprobabilityofinformedtrading(PIN).Further,ﬁndingsarequalitativelysimilarwhenweincludeaproxyformanagerialprivateinformationtocontrolfortheprivateinformationconveyedbystockpricechangesthatisal-readypossessedbymanagers.Finally,wereportthattherelationshipsbetweenpastreturnsandcurrentdividendchangesbecomeweakerifstocksareovervalued,probablybecausechangesinmarketvaluationsarelessinformativeforthesestocksthattendtobemisvalued.Moreimportantly,aftercontrollingforstockovervaluationinourregressions,westillconﬁrmthemainﬁndingsofthestudy.

Wecanconcludethatmanagersseemtorelymoreonpastabnormalvariationsinthevalueoftheirstockwhendecidingcurrentdividendchangesifsuchvariationsarelikelytobemoreinformative.Ourevidencesupportsthenotionthatmanagerstakeadvantageofnovelprivateinformationimpoundedinstockpriceswhensettingthedividendpolicyoftheirﬁrms.ConsistentwithKauetal.(2008),ourstudysuggeststhatmanagers“listentothemarket”sincechangesincorporatepoliciesaresensitivetomarketreactions.Moregenerally,ourstudycontributestothegrowingliteratureontheeffectsofsecondarymarketstockpricesoncorporateﬁnancialdecisions.Notablepreviousstudiesfocusoninvestments(e.g.,Chenetal.,2007;Durnevetal.,2004;Wurgler,2000),CEOturnover

Eventhoughcashcanalsobedisbursedbyrepurchasingstock,buybacksareseenbymanagersandinvestorsasbeingmoreﬂexiblethancashdividends.Hence,itisgenerallyarguedthatpayingdividendsisamorecredibleandeffectivestrategytomitigateagencyconﬂictsandincreaseﬁrmvalue.

1A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–173

(DefondandHung,2004),mergersandacquisitions(Luo,2005),andcashsavings(Fresard,2012).2Ourﬁndingsﬁllagapinthislit-eraturebyhighlightingthatdividenddecisionsareaffectedbyinformationlearnedbymanagersfromstockpricechanges.

Moreover,ourstudyfocusesonapolicythatﬁnancialmanagersconsiderasimportantastheinvestmentpolicyoftheirﬁrms(Bravetal.,2005).Dividendpaymentsinvolvelargesumsofmoney,arequitefrequenttransactions(i.e.,everyquarterinourstudy)thatcangeneratelargemarketreactions,andareinterconnectedwithkeyﬁnancialdecisionsregardingrealinvestments,issuesofdebtandequity,mergersandacquisitions,andtheretentionofearnings(AllenandMichaely,2003).Enhancingourunderstandingofwhichvariablesaffectdividendpolicyis,therefore,usefultobuildbettertheoriesandexplanationsconcerningothercorporateﬁnancede-cisions.Importantimplications,bothformanagersandshareholders,arisefromourstudy.Ithighlightsthenotionthatmanagerscanpotentiallyoptimizetheirdividendpolicybyexploitinginformationconveyedbystockpricechanges.Shareholderscanalsousesuchinformationtoascertainwhetherthefutureprospectsoftheirﬁrmsareimprovingandlargerdividendpaymentsshouldbeexpected.Overall,thisstudyoffersanimportantcontributiontothecorporateﬁnanceliterature.

Thepaperisstructuredasfollows.InSection2wereviewtheexistingliteratureanddiscusstheresearchquestions.Section3de-scribesthedatausedinthisstudyandtheempiricalmethods.InSection4wepresentanddiscusstheempiricalﬁndings,includingthosefromseveralrobustnesstests.Section5providesconcludingremarks.2.Relatedliteratureandresearchquestions

Inthissection,wesummarizepriorstudiesoftherelationshipbetweentheinformationcontentofstockpricesandcorporatedecision-making.Wealsoanalyzeliteratureontheinter-relationsbetweenstockreturns,priceinformativeness,anddividendchang-es.Ourresearchquestionsarediscussedandmotivatedattheendofthissection.2.1.Theinformationalroleofﬁnancialmarketpricesinmanagerialdecision-making

Inrecentyears,therehasbeenagrowinginterestintheeffectsofsecondarymarketpricesondecisionstakenbyindividualsintherealsideoftheeconomy.Forinstance,anargumentthatisoftenadvancedisthatcorporateinsiderscangainusefulinformationfromstockmarketvaluationstomakebetterandmoreinformeddecisions(e.g.,DowandGorton,1997;SubrahmanyamandTitman,1999).Intheircomprehensivereviewoftheexistingliterature,Bondetal.(2012)positthatﬁnancialmarketpricesaresigniﬁcantlymean-ingfulto“realdecisionmakers”forthreemainreasons.First,marketpricesconveyusefulinformationtodecisionmakers.Forinstance,stockpricescontaininformationabouttheﬁrm'sinvestmentopportunitiesandthemanagers'pastdecisions(DowandGorton,1997).Second,decisionmakerstakemarketpricesintoaccounteveniftheydonotlearnanygenuinelynovelinformationfromthem.Third,behavioralbiasesaffectdecisionmakersandleadthemtoirrationallyrelyonmarketpricesandusethemasanchors.

Inthispaper,wefocusontheﬁrstargumentaboveand,morespeciﬁcally,onthelinksbetweenstockmarketpricesanddecisionsbycorporatemanagers.AshighlightedbyBondetal.(2012),whileacorporateinsiderisprobablymoreinformedthantheaverageoutsider,stockmarketpricesmaystillbeusefultoimperfectlyinformedmanagerssimplybecausetheymaynothavesomeinforma-tionthattradershave.Themarketvalueofacompany,infact,reﬂectsaggregateinformationfromaverylargesetofoutsideinvestors.Further,corporateoutsidersmaypossessmorecompleteexternalinformationthaninsidersthatcanalsobeimportantinthecontextofmanagerialdecisions.

Theexistingempiricalliteraturesupportsthenotionthatcorporatedecisionmakerslearnusefulinformationfromﬁnancialmar-kets.Severalauthorshavestudiedcorporateinvestmentpolicy.Forinstance,Wurgler(2000)studiestheﬁnancialmarketsof65coun-triesandreportsthattheefﬁciencyofcapitalallocationispositivelycorrelatedwiththeamountofﬁrm-speciﬁcinformationinstockreturns.HearguesthatmoreinformativepricesallowmanagerstobetterdistinguishbetweengoodandbadinvestmentopportunitiesthroughmorereliableTobin'sQmeasures.Similarly,Durnevetal.(2004)ﬁndsupportforthepredictionthatmoreinformativestockpricesshouldenhancecapitalbudgetingdecisions.Chenetal.(2007)showthatstockpriceinformativenesshasapositiveeffectonthesensitivityofcorporateinvestmenttostockprice.Theyalsoreportthatpriceinformativenessmeasuresareassociatedwithfutureﬁrmoperatingperformance.Thisevidencesuggeststhatprivateinformationconveyedbystockpricesallowsmanagerstolearnabouttheirﬁrms'fundamentalsandenhancemanagerialcorporateinvestmentdecisions.Relatedly,FoucaultandFresard(2014)reportastron-gerpositiverelationshipbetweenaﬁrm'sinvestmentsandthevaluationsofitspeerswhentheﬁrm'sstockpriceislessinformativeandwhentheﬁrm'smanagerspossesslessinformation.BakkeandWhited(2010)showthatwhilestockmarketmispricingdoesnotinﬂuenceinvestments,managersuseprivateinformationinstockmarketpriceswhendecidingtheirinvestmentpolicy.Fromarelat-edperspective,Kauetal.(2008)documentthatmanagerslistentothemarketandtendtoabandoninvestmentprojectswhenthemarketreactsnegativelytotherelatedannouncements.

Therealeffectsofﬁnancialmarketshavealsobeenstudiedinrelationtoothercorporateeventsandpolicies.Forexample,incoun-trieswithastronglawenforcement,thereisevidenceofamoresigniﬁcantrelationshipbetweenCEOturnoverandpoorstockperfor-mancewhenstockpricesaremoreinformative(DefondandHung,2004).Thisﬁndingindicatesthatexecutivesaremorelikelytobeevaluatedusingthestockmarketperformanceoftheircompanieswhenmoreﬁrm-speciﬁcinformationisimpoundedinstockprices.

OurstudyisrelatedtothecateringtheoryofdividendsdevelopedbyBakerandWurgler(2004)inthatitshowsthatstockmarketvaluationsinﬂuencedividendpayments.Yet,whileBakerandWurgler(2004)considertime-varyingmarket-widemeasuresofinvestordemandfordividends,wefocusonﬁrm-speciﬁcchangesinstockvaluationsandarguethatdividendchangesarepartlyassociatedwiththeinformationconveyedbysuchchanges.

24A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

ArelationshipbetweenthemarketreactiontoanM&AannouncementandthecompletionofthedealisreportedbyLuo(2005).ItappearsthatcompaniesinvolvedinanM&Adealextractinformationfromtheinvestorreactiontotheannouncementofthetransac-tioninordertodecidehowtoproceedwiththedeal.Finally,Fresard(2012)concentratesoncashsavingsandreportsahighersavings-to-pricesensitivitywhenstockpriceismoreinformative.Onthewhole,agrowingliteraturedocumentsthattheinformationcontentofstockpricesisendogenoustocorporatedecisionssuchasinvestmentandcashpolicydecisions.

2.2.Stockreturns,priceinformativeness,anddividendchanges

Managerscanusebothpublicandprivateinformationtodecidetheircompanies'dividendpolicies.Thereisnolackofliteraturereportingsigniﬁcantrelationshipsbetweenlaggedandcurrentvaluesofpublicly-knownvariablessuchasproﬁtability,growthoppor-tunities,andﬁrmsizeandcashdividends(e.g.,FamaandFrench,2001;GrullonandMichaely,2002).Sincemanagersmaylearnfromstockpriceswhenmakingdecisions,animportantresearchquestiontoaddressiswhethersuchlearninginﬂuencesdecisionsondiv-idendchanges.

Ithasbeenarguedthatmanagerschangethelevelofdividendpaymentsfordifferentreasons.Someauthorsarguethatdividendscanconveyinformationaboutfuturecashﬂows(MillerandModigliani,1961)andthatmanagersmayusedividendchangesascostlysignalsoffutureearningschanges(Bhattacharya,1979;JohnandWilliams,1985;MillerandRock,1985).Basedontheirarguments,wecouldexpectdividendincreases(decreases)tobefollowedbypositive(negative)changesinproﬁtabilityandcashﬂows.Howev-er,withfewexceptions(e.g.,NissimandZiv,2001),theresearchondividendincreaseshasnotprovidedsupporttothisexpectation.Forexample,Benartzietal.(1997)ﬁndthatcompaniesthatincreasedividendsdonotbeneﬁtfromunexpectedfuturepositivechang-esinearnings.SimilarevidenceonﬁrmperformancehasbeenreportedbyDeAngeloetal.(1996),Grullonetal.(2002),andGrullonetal.(2005).3Asfordividendcuts,whileDeAngeloetal.(1992)showthattheymaybeusefultopredictfutureearnings,otherstudiesshowinsigniﬁcantorevenpositivechangesinﬁrmperformanceafterareductionincashdividends(e.g.,Benartzietal.,1997;Grullonetal.,2005;Jensenetal.,2010).

Changesindividendpaymentsmayalsoconveyinformationaboutfuturerisklevels.Grullonetal.(2002)ﬁndthatthesystematicriskofﬁrmsthatincrease(decrease)theirdividendssigniﬁcantlydeclines(rises)afterthedividendchange.Theyhighlightthattheseempiricalresultsareconsistentwiththeargumentthatﬁrmsstartpayinghigherdividendswhentheybecomemorematureandhavefewerinvestmentopportunities.This“maturityhypothesis”explainsboththedecliningsystematicriskfollowingdividendincreasesandthelackofapositiveassociationbetweendividendchangesandfutureproﬁtability.Otherrelatedstudiesﬁndthatoperatingin-comevolatility(Lie,2005)anddefaultrisk(Charitouetal.,2011)declinearounddividendincreases.

Insummary,theexistingdividendliteraturesuggeststhatcompaniesuseinformationonlaggedandcurrentvaluesofseveralvar-iables(e.g.,proﬁtability)togetherwithexpectationsonothervariables(e.g.,systematicrisk)whendecidingwhether,towhatextent,andinwhichdirectiontheyshouldadjusttheirdividendpolicy.Bothpublicandprivateinformationcanhelpcompaniesoptimizetheirdividendpayments.Asmentionedabove,theusefulnessofprivateinformationconveyedbyoutsideinvestorsthroughtradinghasbeenoverlookedinpriorresearch.Inthispaper,weaimtoﬁllthisgapbyinvestigatingwhetherprivateinformationconveyedbylaggedabnormalchangesinthemarketvaluationofacompanyisusedbythecompany'smanagerstodecidechangesindividendpolicy.Thisisbecausedividendannouncementsaregenerallymadeeveryquarter.Ouraimistoinvestigatewhethermanagerstakeprivateinformationinstockpricesintoaccountwhentheymakefrequentdecisions.Wefocusonabnormalstockreturnsbecauseunexpectedchangesinassetvaluesshouldreﬂectand,therefore,conveynewsaboutcashﬂowsand/ornewsaboutdiscountrates(e.g.,CampbellandShiller,1988;ChenandZhao,2009).Forinstance,theempiricalstudyofChenetal.(2013)reportsthatbothcashﬂowanddiscountratenewsaffectunexpectedmovementsinstockpricesinasigniﬁcantway.

Wheninvestorexpectationsonthecashﬂowsproducedbyastockarerevisedupward,themarketvalueofthestockshouldincrease.Incontrast,higherexpectationsondiscountratesshoulddepressmarketvaluations.Wethusexpecttoﬁndapositiverelationshipbetweendividendchangesandlaggedabnormalstockreturns,alsoinlightofpriorempiricalﬁndings(e.g.,Brownetal.,2007;Chemmanuretal.,2010;Grullonetal.,2002).However,identifyingsuchrelationshipisnotenoughtoconcludethatcompaniesexploitnewprivateinformationlearnedfromstockpricemovementswhensettingtheirdividendpolicy.Newpublicinformationfromothersourcesmay,infact,drivebothamendmentsindividendpolicyandvariationsinstockmarketvaluations.

Inourstudy,wegoastepfurtherandtestwhetherlaggedabnormalreturnshavemoresigniﬁcanteffectsondividendchangeswhenstockpricesaremorelikelytoimpoundandconveyinvestorprivateinformation.Assumingthatmanagerscanlearnusefulprivateinformationfromvariationsinstockprices,weexpectthatmoreprivateinformationinstockpriceswouldmakethepositiverelationshipbetweendividendchangesandlaggedabnormalstockreturnsstronger.Whentestingthisprediction,wealsoattempttocontrolformanagerialprivateinformationinordertoseparatetheeffectsofthisinforma-tionfromthatpotentiallylearnedbymanagersfrompaststockpricechanges.Finally,sincestockreturnsarelikelytobelessinformativewhenstockstendtobemisvalued,insomeofouranalysesweinvestigatetherobustnessofourﬁndingstotheinclusionofmarketmisvaluationproxies.

Studiesofdividendinitiationsoffermixedconclusions.Forinstance,HealyandPalepu(1988)ﬁndthatinitiationssignalfutureearningsgrowthwhereasBulanetal.(2007)showthatproﬁtabilityandgrowthdonotimproveafteracompanystartspayingdividends.

3A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–175

3.Sample,dataandempiricalmethods3.1.Sampleofdividendchanges

WeconsiderannounceddividendpaymentstocomputeourDividendchangevariable.DividendannouncementsareselectedfromCRSPmainlyfollowingtheﬁltersofGrullonetal.(2002).Tobespeciﬁc,weincludedividendannouncementsmadebyNYSEandAMEXﬁrmsbetween1962and2010thatsatisfythefollowingcriteria:

a)TheannouncementisforaquarterlycashdividendpaymentinUSdollars;

b)Thedividendpaymentisnotaninitiation(i.e.,thereisaquarterlycashdividendpaymentinthepreviousquarter);

c)Thepreviousquarterlycashdividendpaymentwasmadewithinawindowof20to90daysbeforethecurrentdividendannouncement;

d)Thedividendpaymentisnotmadebyaﬁrmthatiseitherautilityﬁrm(SICs4900to4949)oraﬁnancialﬁrm(SICs6000to6999);e)TheCRSPsharecodefortheﬁrmannouncingthedividendpaymentiseither10or11.

Dividendchangeisthequarterlydividendannouncedinthecurrentquarterminusthequarterlydividendannouncedinthepre-viousquarter,alldividedbythequarterlydividendforthepreviousquarter.Thisvariablecantakepositive,negative,andzerovalues.Afterexcludingobservationswithmissingvaluesfortheothervariablesusedinthestudy,ourbaselinesampleincludes98,535obser-vationsfor2510uniqueﬁrmswithvalidvaluesofthevariableDividendchange.Inthissample,thereare16,240observationsfor1994uniqueﬁrmswithnon-zerovaluesofDividendchange.3.2.Baselinespeciﬁcationanddatasources

Inordertostudyhowthesensitivityofdividendchangestolaggedabnormalreturnsvarieswithprivateinformationinstockprices,weemploythefollowingbaselinemodel:

Dividendchangei;q¼β0þβ1Abnormalreturni;qþβ2Privateinformationi;q

󰀁󰀃

þβ3Abnormalreturni;qÂPrivateinformationi;qþβ4Dividendyieldi;q

þβ5Marketcapitalizationi;qþβ6Debti;qþβ7Cashi;q

þβ8Market‐to‐booki;qþβ9Operatingincomei;qþβ10Dividendpremiumi;qþεi;q

ð1Þ

whereiidentiﬁestheﬁrmthatisannouncingthequarterlydividendpaymentandqisthequarterwhentheannouncementtakesplace.TheinclusionoftheinteractiontermAbnormalreturn×Privateinformationallowsustotestwhetherthelevelofprivateinformationinstockpriceshasasigniﬁcantimpactontherelationshipbetweendividendchangesandlaggedabnor-malreturns.

Stockmarketdatasuchasquarterlydividendpayments,stockreturns,tradingvolumes,turnovers,dividendyields,andmarketcapitalizationsarecollectedfromCRSP.Compustatisoursourceofaccountingdata.OurstudyalsousesSørenHvidkjær'sdata,avail-ablefromhiswebsite,4tomeasuretheprobabilityofinformedtrading(PIN)asoneofouralternativeproxiesforstockpriceinforma-tiveness.Deﬁnitionsforalltheindependentvariablescanbefoundinthefollowingsections.

Inallourregressionswecontrolfortwo-digitSICindustrydummies.5Standarderrorsareadjustedfortwo-wayclusteringbyﬁrmandcalendarmonth.Inordertominimizetheinﬂuenceofverylargeabnormalstockreturnsthatcouldpossiblyreﬂectmajorcorpo-rateevents,wetrim1%oftheobservationsfromeachtailofthedistributionofAbnormalreturn.Similarly,bothtailsofthedistributionsofalltheothervariablesusedinthisstudyarewinsorizedatthetopandbottom1%level.3.3.Abnormalreturn

WefollowamethodsimilartothatusedbyFaulkenderandWang(2006)tocomputetheabnormalreturns.Speciﬁcally,wedeﬁnetheabnormalreturnasthereturninexcessofthevalue-weightedreturnfortherelevantFama–Frenchbenchmarkportfolio.ForeachDividendchangeevent,thebenchmarkportfolioisselectedfromthesetof25Fama–Frenchportfoliosformedonsize(i.e.,marketvalueofequityME)andbookvalueovermarketvalueofequity(BE/ME).Toidentifytherelevantportfolio,weconsiderﬁrm-speciﬁcMEandBE/MEvaluesandusematchingcriteriainaccordancewiththemethodsoriginallyfollowedtocreatetheportfolios.6Overthepre-eventperiod,Abnormalreturn,whichisexpressedinpercentterms,istheaveragedailydifferencebetweenthereturnon

https://sites.google.com/site/hvidkjaer/data.

Resultsarequalitativelysimilarifwealsocontrolforcalendarmonthdummies.TheonlysigniﬁcantdifferenceisthatcoefﬁcientsforDividendpremiumbecomein-signiﬁcantinallregressionswhenmonthdummiesareincluded.6Eachcalendaryear,theFama–FrenchportfoliosandtheirrespectivebreakpointsaregeneratedattheendofJune.FordailystockreturnsofaﬁrmbetweenJuly1stofyeartandJune30thofyeart+1,weusetheyeartbreakpointsandthefollowingﬁrm-speciﬁcvariablestoselectasuitablebenchmarkportfolio.Theﬁrm'sMEismea-suredattheendofJuneofyeartwhileBE/MEistheratiobetweenBEfortheﬁscalyearendingincalendaryeart−1andMEasoftheendofthesameyear.Portfoliovalue-weightedreturnsandportfoliobreakpointsareobtainedfromKennethFrench'swebsite(http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/).

546A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

thestockandthereturnontherelevantbenchmarkportfolio.7Thepre-eventperiodforthiscomputationistheperiodbetweenthepreviousandthecurrentquarterlydividendannouncements.However,wedonotusedatafortheﬁvetradingdaysbeforethecurrentdividendannouncementandtheﬁvetradingdaysafterthepreviousdividendannouncement.Theseexclusionsaremeanttomini-mizeand/oreliminatetheimpactofthemarketreactionstodividendannouncementsonourabnormalreturnestimates.

Apositive(negative)Abnormalreturncanbeseenasanindicationthat,duringtheperiodbetweenthecurrentandthepreviousdividendannouncements,marketinvestorshaveincreased(decreased)theirvaluationofthestockfollowingtheacquisitionofnovelinformationthatwasunknowntoinvestorsatthetimeofthepreviousdividendannouncement.3.4.Measuresofprivateinformationinstockmarketprices

Awidelyusedmeasureofprivateinformationinstockpricesisﬁrm-speciﬁcstockreturnvariation(orpricenon-synchronicity).Thismeasure,whichwasﬁrstlydevelopedbyRoll(1988),reﬂectsthevariationinthereturnonastockthatcannotbeexplainedbymarketandindustryreturns.Foragenericstocki,ﬁrm-speciﬁcreturnvariationcanbedeﬁnedasψ=ln((1−R2)/R2),whereR2isestimatedfromthefollowingregression:

ri;j;t¼aiþbi;mrm;tþbi;jrj;tþεi;t

ð2Þ

whereri,j,tisthereturnforﬁrmithatispartofindustryjattimet,rm,trepresentsthemarketreturnattimet,andrj,tisthereturnforindustryjattimet.ThemarketreturnisbasedontheCRSPvalue-weightedmarketportfolioandﬁrmsaregroupedintoindustriesusing3-digitCRSPSICcodes.Weexcludeﬁrmifromthemarketandindustryindiceswhenwecalculatemarketandindustryreturns.Foreachdividendchangeannouncementinoursample,weestimateitsﬁrm-speciﬁcreturnvariation(ψ)usingdailystock,market,andindustryreturnsoverthecalendaryearbeforetheannouncementdate.8Roll(1988)arguesthatﬁrm-speciﬁcreturnvariationmaybecorrelatedwithprivateinformation.He,infact,ﬁndsthatidentiﬁablenewsreleasesdonotdriveﬁrm-speciﬁcreturnvariation.Sincemovementsinstockpricesaregenerallycausedbyeitherreleasesofnewpublicinformation(e.g.,corporateearningsandGDPﬁgures)orbytradesofinvestorspossessingprivateinformation,Roll'sﬁnd-ingsmaysuggestthatonlythislattertypeofinformationisassociatedwithpricenon-synchronicity.However,asconcededbytheau-thor,tradesbyuninformednoisetraderscouldalsoproducechangesinstockpricesthatareunrelatedtomarketandindustryreturns.

Despitethiscautionarynote,alargeandgrowingbodyofempiricalresearchlendscredencetothenotionthatﬁrm-speciﬁcreturnvariationisavalidmeasureofprivateinformationinstockprices.Theempiricalresearchthatvalidatesthismeasureinthemostcon-vincingwayisprobablyDurnevetal.(2003).ThisstudyreportsanegativerelationshipbetweenR2andthestrengthoftheassociationbetweencurrentstockreturnsandfutureearnings.Thisresultimpliesthattheinformationimpoundedincurrentstockpricesismoreusefultopredictfutureearningswhenﬁrm-speciﬁcreturnvariationrepresentsalargerfractionoftotalreturnvariation.

Relatedly,Morcketal.(2000)provideevidenceofalowerlevelofﬁrm-speciﬁcreturnvariationinlow-incomeeconomiesandofadecliningtrendinpricesynchronicityintheU.S.Durnevetal.(2004)reportthattheefﬁciencyofcorporateinvestmentsisanincreas-ingfunctionofﬁrm-speciﬁcreturnvariation.Thisempiricalregularitysuggeststhatlesssynchronouspricesconveymoreprivatein-formationthatmanagersusetoenhancetheirinvestmentpolicy.PiotroskiandRoulstone(2004)ﬁndpricesynchronicitytobepositivelyassociatedwiththeactivityofﬁnancialanalysts.Incontrast,theyshowthattradesbyinsidershaveanegativeeffectonsyn-chronicity.Financeresearchershaveveryfrequentlychosenﬁrm-speciﬁcreturnvariationasameasureofprivateinformationinstockprices.9Inspiteofﬁrm-speciﬁcreturnvariationbeingawidelyusedmeasureofprivateinformationinstockprices,somerecentstudiescastdoubtsonthereliabilityofthismeasure.Throughasimplemodelandconsistentempiricalﬁndings,Dasguptaetal.(2010)showthat,contrarytoconventionalwisdom,amoretransparentinformationenvironmentshouldbeassociatedwithhigherstockre-turnsynchronicity.XingandAnderson(2011)reportanon-monotonicrelationshipbetweenpricesynchronicityandproxiesofpublicﬁrm-speciﬁcinformation.Assuminganinverseassociationbetweenprivateandpublicﬁrm-speciﬁcinformation,thisﬁndingunder-minesthenotionthatprivateinformationvariesmonotonicallywithsynchronicity.Similarly,LeeandLiu(2011)reportU-shapedrelationsbetweenﬁrm-speciﬁcreturnvariationandseveralalternativepriceinformativenessmeasures.ChanandChan(2014)reportanegativerelationbetweenstockreturnsynchronicityandSEOdiscountandconcludethattheinformativenessofstockpricesgrowswithreturnsynchronicity.Inlightoftherecentstudiesaboveandtoevaluatetherobustnessofourﬁndings,inourstudyweconsiderseveralalternativestostockpricenon-synchronicity.

Forthesameperiod,weconsiderseveralalternativemeasuresofabnormalreturn.First,wesimplyusetheaveragedailyreturnonthestock,essentiallyassumingthatthebenchmarkportfolioproducesazeroreturn.Second,werundailytime-seriesregressionstoestimatetwoversionsofJensen'sAlpha,onebasedonthethreeFama–Frenchriskfactorsandoneaugmentedbythemomentumfactor.WereplicateallouranalysesreplacingAbnormalreturnwiththesealternativemeasuresandﬁndqualitativelysimilarresults.8ThisperiodcouldoverlapwiththatforthecalculationofAbnormalreturn.Weevaluatewhetherthiscircumstanceaffectsourﬁndingsbyusing,instead,laggedvaluesofﬁrm-speciﬁcreturnvariationandtheotherprivateinformationmeasures.Weconﬁrmthatthemainﬁndingsofourpaperarequalitativelysimilarwhenlagsareused.9Themeasurecanbefoundinempiricalstudiesofcapitalallocationandinvestments(BakkeandWhited,2010;Chenetal.,2007;FoucaultandFresard,2014;Wurgler,2000),cashsavings(Fresard,2012),CEOturnover(DefondandHung,2004),privatebeneﬁtsofcontrol(DyckandZingales,2004),analystcoverage(ChanandHameed,2006;Crawfordetal.,2012),stockmarketopaqueness(Huttonetal.,2009;JinandMyers,2006),corporategovernancecharacteristics(FerreiraandLaux,2007;Ferreiraetal.,2011),cross-listings(FernandesandFerreira,2008),blockownership(BrockmanandYan,2009),insidertradinglaws(FernandesandFerreira,2009),andtheassociationbetweenassetgrowthratesandstockreturns(Watanabeetal.,2013).

7A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–177

Theilliquidityratio(Illiq),introducedbyAmihud(2002),hasbeenrecentlyusedasapriceinformativenessmeasurebyFerreiraetal.(2011)andFresard(2012).Thisvariableistheaverage(multipliedby106)ofthedailyratiobetweentheabsolutevalueofthereturnonastockandthestock'sdollarvolume:

󰀄󰀄Di󰀄r󰀄1Xi;t

Illiqt¼

Dit¼1Volumei;t

ð3Þ

whereDiisthenumberofvaliddailyobservationsforﬁrmiovertheperiod,while|ri,t|andVolumei,tare,respectively,theabsolutevalueofthereturnandthedollarvolumeforﬁrmiattimet.Illiqiscomputedforeachdividendchangeannouncementusingdailydataforthecalendaryearbeforetheannouncementaslongasthereareatleast200tradingdayswithnon-zerotradingvolume.Theapplicationofthisﬁlterallowsustodiscardobservationsforthinlytradedstocks.Theilliquidityratioisaproxyforthepriceim-pactoftradesbecauseitmeasurestheextenttowhichstocktransactionscausestockpricestochange.Thus,inthespiritofthemodelbyKyle(1985),thisratioshouldalsoreﬂecttheinformativenessofstockmarkettransactions.

WealsoconsidertheprivateinformationtradingmeasureofLlorenteetal.(2002).ExamplesofstudiesthathavechosenthismeasurearethoseofFerreiraandLaux(2007),FernandesandFerreira(2008),FernandesandFerreira(2009),andFresard(2012).Werunthefollowingregressiontoestimatethemeasureofinformation-basedtradingγ:

󰀁󰀃

ri;t¼aiþbiri;t−1þcirm;tþγiri;t−1ÂVi;t−1þεi;t

ð4Þ

whereri,tisthereturnforﬁrmiattimet,rm,trepresentsthemarketreturnattimet(basedontheCRSPvalue-weightedmarketre-turn),andVi,t−1isthelogarithmofﬁrmi'sturnoverattimet−1,detrendedbysubtractingitsaveragevalueintheprevious200tradingdays.10Thecoefﬁcientsoftheregressionaboveareestimatedforeachdividendchangeannouncementusingdailyreturnandturnoverdataforthecalendaryearprecedingtheannouncement.InthemodelbyLlorenteetal.(2002),thecoefﬁcientyisnor-mallyinterpretedasameasureofinformation-basedtrading.Theintuitionbehindthisinterpretationisthatstockreturnsshouldbemorepositivelyseriallycorrelatedwhentradesbyinformedinvestorsaremorefrequent.

Theﬁnalpriceinformativenessproxyweconsideristheprobabilityofinformedtrading(PIN)measureofEasleyetal.(2002).Thismeasureisbasedonamarketmicrostructuremodelinwhichtradescanbeexecutedbynoiseorinformedtraders.PINhasbeencho-senasaproxyofprivateinformationinstockpricesbymanystudies.NotableexamplesarethoseofChenetal.(2007),FerreiraandLaux(2007),BrockmanandYan(2009),andFerreiraetal.(2011).WeobtainallavailableyearlyestimatesofthePIN,from1983to2001,foroursampleofNYSE/AMEXcommonstocksfromtheaforementionedsource—SørenHvidkjær'swebsite.Foraparticulardiv-idendchangeannouncement,weconsiderthePINforthelatestcalendaryearbeforetheannouncement.

3.5.Controlvariables

Wecontrolforseveralvariablesthathavebeenshowntobesigniﬁcantdeterminantsofchangesinquarterlydividendsinthepre-viousliterature.Inparticular,werelyonthesetofcontrolsusedbyLiandLie(2006)thatincludesDividendyield,Marketcapitalization,Debt,Cash,Market-to-book,Operatingincome,andDividendpremium.

DividendyieldistheCRSPannualreturnfromdividendpayments.Marketcapitalization(inbillionUSdollaranddeﬂatedto1980usingtheCPI)istheCRSPmarketvalueofequity.Debtistheratiooflong-termdebt(Compustatitem9)tototalassets(Compustatitem6).Cashiscashandshort-terminvestments(Compustatitem1)overtotalassets(Compustatitem6).Market-to-bookisthemar-ketvalueofassetsdividedbythebookvalueofassets(Compustatitem6).Themarketvalueofassetsisdeﬁnedasthebookvalueofassetsminusthebookvaluesofcommonequity(Compustatitem60)anddeferredtaxes(Compustatitem74),plusthemarketvalueofequity(Compustatitem199timesCompustatitem25).Operatingincomeisoperatingincomebeforedepreciation(Compustatitem13)scaledbytotalassets(Compustatitem6).FollowingBakerandWurgler(2004),wedeﬁneDividendpremiumasthedifferencebe-tweenthelogofthevalue-weightedaverageMarket-to-bookratioofdividendpayersandthesamemeasurefornon-dividendpayers.WecarefullyreplicatethestepsfollowedbyBakerandWurgler(2004)tocomputethisvariableforeveryyearinoursampleperiod.However,inordertobeconsistentwithoursampleofdividendchangesthatonlyincludesobservationsforNYSEandAMEXﬁrms,weexcludedataforNASDAQcompanieswhenbuildingtheDividendpremiumvariable.

ExceptforMarketcapitalizationandDividendpremium,weconsiderlaggedvaluesofallcontrolvariablesforﬁscalperiodsendingbeforethedateofthecurrentquarterlydividendannouncementthatisusedtocomputetheDividendchangevariable.ThevariableMarketcapitalizationisbasedondatafromonedaypriortothecurrentquarterlydividendannouncement.Dividendpremiumisforthelatestcalendaryearimmediatelybeforetheannouncementofthecurrentquarterlydividend.

Inordertokeepobservationswithzerodailyturnovers,wefollowLlorenteetal.(2002)andadd0.00000255totheactualturnoverbeforeapplyingthelogarithmictransformation.

108A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table1

Descriptivestatistics.Thetableshowsdescriptivestatisticsforthedependentandindependentvariablesusedinthebaselineregressionmodels.Dividendchangeisthedifferencebetweenthequarterlydividendpaymentannouncedinthecurrentquarterandthequarterlydividendpaymentannouncedinthepreviousquarter,overtheannouncedquarterlydividendforthepreviousquarter.Abnormalreturnistheaveragedailypercentabnormalreturnovertheperiodstartingfromthesixthtradingdayaftertheannouncementofthequarterlydividendinthepreviousquarterandendingonthesixthtradingdaybeforetheannouncementofthequarterlydividendinthecurrentquarter.Dailyabnormalreturnisthedailystockreturninexcessofthedailyvalue-weightedreturnfortherelevant(basedonﬁrmsizeandbook-to-marketratio)Fama–Frenchbenchmarkportfolio.ψ=ln((1−R2)/R2),whereR2isforaregressionofthedailystockreturnondailymarketandindustryreturns.IlliqistheaveragedailyvalueoftheilliquidityratiobyAmihud(2002).γistheprivateinformationtradingmeasurebyLlorenteetal.(2002).Itisthecoefﬁcientoftheinteractiontermbetweenthelaggeddailystockreturnandthelaggeddailydetrendedlog-turnoverfromaregressionofthedailystockreturnontheinteractionterm,thelaggeddailystockreturn,andthedailymarketreturn.PINistheprobabilityofinformedtradingmeasureofEasleyetal.(2002).ForeachvalueofDividendchange,thevaluesofψ,Illiq,γ,andPINareforthelatestcalendaryearbeforetheannouncementdateofthecurrentquarterlydividend.Marketcapitalization(inbillionUSdollaranddeﬂatedto1980usingtheCPI)isthemarketvalueofequity.Marketcapitalizationismeasuredonthetradingdaybeforethedayofthecurrentquarterlydividendannouncement.Dividendyieldistheannualreturnfromdividendpayments.Debtistheratiooflong-termdebttototalassets.Cashiscashandshort-terminvestmentsovertotalassets.Market-to-bookisthemarketvalueofassetsdividedbythebookvalueofassets.Operatingincomeisoperatingincomebeforedepreciationoverthebookvalueofassets.Dividendpremiumisthelogofthevalue-weightedaveragemarket-to-bookratioofdividendpayersminusthelogofthevalue-weightedaverageMarket-to-bookratioofnon-dividendpayers(BakerandWurgler,2004).ThevaluesofDividendyield,Debt,Cash,Market-to-book,andOperatingincomeareforthelatestﬁscalperiodsendingbeforethedateofthecurrentquarterlydividendannouncementwhileDividendpremiumisforthemostrecentcalendaryear.

Obs.

DividendchangeAbnormalreturnψIlliqγPIN

Dividendyield

MarketcapitalizationDebtCash

Market-to-bookOperatingincomeDividendpremium

98,53598,53598,53598,53594,47935,35998,53598,53598,53598,53598,53598,53598,535

Mean0.02250.00181.85497.2026−0.00250.18530.03661.66090.18720.08131.46930.16740.0024

Min−8E−07−0.625−0.93150.0039−0.28450.07310.00060.01140

0.00160.63060.0205−0.0244

Median0

−0.00231.80611.8535−0.00090.180.02510.35280.18020.05181.21430.15970.0029

Max0.40.69835.370262.68830.260.35660.13530.39730.55210.4415.33150.39870.0338

Skewness3.62390.1230.29782.55−0.080.51741.112.88580.511.97972.34230.6995−0.0729

Kurtosis16.8623.13363.20529.193.21382.99683.459329.63192.9057.34259.68263.79272.204

St.dev.0.06740.22371.216211.96040.10350.05970.032.22730.12660.08450.81670.07040.0142

Table2

Firm-speciﬁcreturnvariationanddividendchanges:MultinomialLogitmodels.ThetablepresentsMultinomialLogitregressionresultsforthelikelihoodofanincrease(columns(1),(3),and(5))oradecrease(columns(2),(4),and(6))inquarterlydividendpaymentsonthepre-eventabnormalreturn,theﬁrm-speciﬁcreturnvariation(ψ),theinteractionbetweenthesetwovariables,andasetofcontrolvariablesincludingtwo-digitSICindustrydummies(notreportedinthetable).Thebaseoutcomeinthemodelisthezerochangeinquarterlydividendpayments.ThedependentvariableisbasedonthecontinuousvariableDividendchange.DeﬁnitionsforthisandalltheothervariablescanbefoundinTable1.z-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

DividendchangeN0(1)

Abnormalreturnψ

Abnormalreturn×ψDividendyieldMarketcapitalizationDebtCash

Market-to-bookOperatingincomeDividendpremiumConstantObservationsPseudo-R2−1.460⁎⁎(−2.129)0.016⁎⁎⁎(4.219)0.074(0.556)0.172(1.046)−0.052⁎⁎(−2.233)4.576⁎⁎⁎(19.123)0.541(0.369)−2.443⁎⁎⁎(−40.935)98,5350.0210

12.900⁎⁎⁎(10.473)−0.078⁎⁎⁎(−3.026)0.203(0.747)−0.466(−0.911)0.286⁎⁎⁎(3.385)−5.0⁎⁎⁎(−7.062)−3.154(−0.583)−4.203⁎⁎⁎(−28.047)98,5350.0210

−1.035(−1.483)0.010⁎⁎(2.577)0.036(0.269)0.203(1.221)−0.052⁎⁎(−2.252)4.523⁎⁎⁎(18.974)0.188(0.129)−2.312⁎⁎⁎(−36.212)98,5350.0192

12.660⁎⁎⁎(10.313)−0.073⁎⁎⁎(−2.8)0.232(0.849)−0.491(−0.959)0.295⁎⁎⁎(3.568)−5.112⁎⁎⁎(−7.108)−3.197(−0.595)−4.222⁎⁎⁎(−23.776)98,5350.0192

0.501⁎⁎⁎(9.719)

Dividendchangeb0(2)−1.081⁎⁎⁎(−6.133)

DividendchangeN0(3)

Dividendchangeb0(4)

DividendchangeN0(5)0.260⁎⁎⁎(2.3)−0.069⁎⁎⁎(−5.144)0.138⁎⁎⁎(3.813)−1.018(−1.451)0.010⁎⁎⁎(2.581)0.030(0.224)0.182(1.092)−0.050⁎⁎(−2.152)4.524⁎⁎⁎(18.874)0.031(0.021)−2.310⁎⁎⁎(−36.117)98,5350.0219

Dividendchangeb0(6)−0.523(−1.612)0.007(0.216)−0.267⁎⁎(−2.008)12.855⁎⁎⁎(10.46)−0.073⁎⁎⁎(−2.807)0.218(0.799)−0.462(−0.906)0.287⁎⁎⁎(3.406)−5.069⁎⁎⁎(−7.028)−3.061(−0.566)−4.229⁎⁎⁎(−23.833)98,5350.0219

−0.0⁎⁎⁎(−4.773)

0.021(0.656)

⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table3

Firm-speciﬁcreturnvariationanddividendchanges:OLSmodels.ThetablepresentsOrdinaryLeastSquaresregressionresultsforthechangeinquarterlydividendpaymentsonthepre-eventabnormalreturn,theﬁrm-speciﬁcreturnvariation(ψ),theinteractionbetweenthesetwovariables,andasetofcontrolvariablesincludingtwo-digitSICindustrydummies(notreportedinthetable).Observationswithzerodividendchangesareexcludedfromthesample.DeﬁnitionsforallthevariablescanbefoundinTable1.t-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

Dividendchange(1)

Abnormalreturnψ

Abnormalreturn×ψDividendyieldMarketcapitalizationDebtCash

Market-to-bookOperatingincomeDividendpremiumConstantObservationsAdjustedR2⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

−1.555⁎⁎⁎(−13.710)−0.000(−1.199)0.022(1.020)0.205⁎⁎⁎(6.326)−0.032⁎⁎⁎(−8.708)0.534⁎⁎⁎(11.506)1.359⁎⁎⁎(5.674)0.233⁎⁎⁎(5.202)16,2400.091

−1.507⁎⁎⁎(−13.066)−0.001⁎⁎⁎(−2.888)0.022(1.003)0.215⁎⁎⁎(6.668)−0.032⁎⁎⁎(−8.877)0.533⁎⁎⁎(11.326)1.306⁎⁎⁎(5.445)0.253⁎⁎⁎(5.872)16,2400.075

0.128⁎⁎⁎(13.637)

(2)

(3)0.076⁎⁎⁎(4.512)−0.009⁎⁎⁎(−3.994)0.029⁎⁎⁎(3.472)−1.495⁎⁎⁎(−13.174)−0.001⁎⁎⁎(−2.687)0.018(0.809)0.209⁎⁎⁎(6.434)−0.031⁎⁎⁎(−8.653)0.526⁎⁎⁎(11.319)1.300⁎⁎⁎(5.434)0.254⁎⁎⁎(5.533)16,2400.094

−0.008⁎⁎⁎(−3.451)

4.Empiricalﬁndings4.1.Descriptivestatistics

ThedescriptivestatisticsforallthevariablesincludedinourbaselinespeciﬁcationsarepresentedinTable1.TheaveragevalueofDividendchange(2.2%)showsthat,onaverage,companiesincreasetheirquarterlydividendsratherthancuttingthemorkeepingthemunchanged.However,themedianvalueofDividendchangeis0andthisvariablerangesfromaminimumthatisjustbelowzerotoaverylargevalueof40%.11Onlyin1366casesthechangeinquarterlydividendsisnegativewhileazerochangeindividendsisverycommon.

ItisalsoworthnotingthatannouncementsofquarterlydividendchangesaregenerallyprecededbypositiveabnormalchangesinstockpricessincethemeanAbnormalreturnis0.0018%.However,therangeofvariationofthisvariableisverywide.Thedescriptivestatisticsforsomeoftheothervariablesindicatethat,onaverage,theDividendyieldis3.66%,andlong-termdebtandcashandshort-terminvestmentsrepresent18.72%and8.13%oftotalassetsrespectively.Finally,themean(median)valueofOperatingincomeis16.74%(15.97%).

4.2.Firm-speciﬁcstockreturnvariationanddividendchanges

Inourﬁrstsetofmultivariateanalyses,weconsiderﬁrm-speciﬁcstockreturnvariation(ψ)asproxyfortheprivateinformationimpoundedinstockprices.Also,weuseMultinomialLogitregressionstoestimatethebaselinespeciﬁcationofEq.(1).Tobuildourcategoricaldependentvariable,weconsiderthefollowingthreepossibleoutcomesbasedonthevariableDividendchange:thevalueofDividendchangeiszero(nochangesindividends),positive(dividendincreases),ornegative(dividenddecreases).We

ThewinsorizationthatisappliedtotheextremevaluesofDividendchangesigniﬁcantlyincreasestheminimumvalueofthisvariable.Thepre-winsorizationmin-imumvalueofthevariableisaround−0.9792.

1110A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table4

Alternativeprivateinformationmeasuresanddividendchanges:MultinomialLogitmodels.ThetablepresentsMultinomialLogitregressionresultsforthelikelihoodofanincrease(columns(1),(3),and(5))oradecrease(columns(2),(4),and(6))inquarterlydividendpaymentsonthepre-eventabnormalreturn,ameasureofprivateinformationinstockprices(alternativetotheﬁrm-speciﬁcreturnvariation),theinteractionbetweenthesetwovariables,andasetofcontrolvariablesincludingtwo-digitSICindustrydummies(notreportedinthetable).Thebaseoutcomeinthemodelisthezerochangeinquarterlydividendpayments.ThedependentvariableisbasedonthecontinuousvariableDividendchange.DeﬁnitionsforthisandalltheothervariablescanbefoundinTable1.z-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

DividendchangeN0(1)

AbnormalreturnIlliq

Abnormalreturn×Illiqγ

Abnormalreturn×γPIN

Abnormalreturn×PINDividendyieldMarketcapitalizationDebtCash

Market-to-bookOperatingincomeDividendpremiumConstantObservationsPseudo-R2−1.367⁎⁎(−1.984)0.015⁎⁎⁎(3.5)0.068(0.511)0.178(1.082)−0.053⁎⁎(−2.254)4.588⁎⁎⁎(19.062)0.625(0.423)−2.437⁎⁎⁎(−40.57)98,5350.0213

12.131⁎⁎⁎(9.415)−0.068⁎⁎⁎(−2.86)0.274(1.012)−0.505(−0.997)0.302⁎⁎⁎(3.656)−5.175⁎⁎⁎(−7.241)−3.900(−0.706)−4.259⁎⁎⁎(−27.674)98,5350.0213

−1.3⁎(−1.958)0.015⁎⁎⁎(3.9)0.084(0.618)0.182(1.055)−0.051⁎⁎(−2.112)4.654⁎⁎⁎(18.517)0.758(0.515)−2.465⁎⁎⁎(−40.539)94,4790.0216

12.943⁎⁎⁎(9.967)−0.074⁎⁎⁎(−2.97)0.212(0.737)−0.839(−1.566)0.294⁎⁎⁎(3.272)−5.669⁎⁎⁎(−7.636)−2.873(−0.515)−4.113⁎⁎⁎(−26.753)94,4790.0216

0.400⁎⁎⁎(6.558)−0.001(−0.803)0.013⁎⁎⁎(3.746)

Dividendchangeb0(2)−1.117⁎⁎⁎(−5.535)0.007⁎⁎(2.455)0.003(0.304)

DividendchangeN0(3)0.511⁎⁎⁎(9.704)

Dividendchangeb0(4)−1.076⁎⁎⁎(−5.942)

DividendchangeN0(5)−0.049(−0.188)

Dividendchangeb0(6)−0.063(−0.069)

−0.248⁎⁎⁎(−2.863)0.296(0.766)−0.515⁎(−1.726)−1.680(−1.103)

0.011(0.029)3.202⁎⁎(2.542)−5.458⁎⁎(−2.156)0.024⁎⁎⁎(3.688)−0.2⁎⁎⁎(−3.108)−0.252(−0.948)0.046(1.255)4.501⁎⁎⁎(11.008)−13.654⁎⁎⁎(−5.552)−2.512⁎⁎⁎(−19.977)35,3590.0320

−2.315⁎(−1.861)−6.386(−1.338)44.737⁎⁎⁎(8.568)−0.134⁎⁎⁎(−2.724)0.957⁎⁎(2.073)−0.995(−1.45)0.474⁎⁎⁎(3.98)−6.547⁎⁎⁎(−5.853)21.876⁎⁎⁎(3.438)−4.590⁎⁎⁎(−13.608)35,3590.0320

⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

focusondividendincreasesanddecreasesseparatelybyusingtheﬁrstoutcome(nochangeinquarterlydividends)asthebasecase.12EmpiricalestimatesforthreeMultinomialLogitmodelscanbefoundinTable2.

Incolumns(1)and(2)ofTable2,thesetofindependentvariablescomprisesAbnormalreturntogetherwithallthestandardcon-trols.Asexpected,thereisapositiveandstatisticallysigniﬁcantrelationshipbetweenAbnormalreturnandthelikelihoodofadividendincrease(column(1)).Incontrast,anincreaseinAbnormalreturnmakesadividendcutlesslikely(column(2)).Firmstendtoincrease(decrease)dividendsafteraperiodofstrong(weak)stockmarketperformance.However,theseﬁndingsmaynotnecessarilyindicatethatmanagerslearnnewinformationfromrecentstockpricechangesthattheyexploitwhendeterminingdividendchanges.Forinstance,thedisclosureofunexpectednewsonacompanymay,atthesametime,affectitsmarketvaluationsigniﬁcantlyandcausefuturevariationsindividendpayments.

Next,weconsiderameasureofprivateinformationinstockprices.Byaddingthismeasuretoourmodels,wecantestwhethermanagerialpayoutdecisionsaremoreconsistentwiththesignandmagnitudeofrecentstockpricechangeswhensuchchangesconveymoreprivateinformation.Incolumns(3)and(4),wereplaceAbnormalreturnwiththeﬁrm-speciﬁcstockreturnvariationψ.Weﬁndthatthismeasureofprivateinformationinstockpriceshasastatisticallysigniﬁcantnegativeeffectondividendincreases.Theempiricalﬁndingssuggestthatpositivechangesinquarterlydividendsbecomemorelikelyifthestockoftheﬁrmconveyslessprivateinformation.Incolumns(5)and(6),weincludeAbnormalreturn,theﬁrm-speciﬁcstockreturnvariation,and,moreimportantly,thein-teractionbetweenthesetwovariables.ThesignsofthecoefﬁcientsonAbnormalreturnandψconﬁrmtheconclusionsfromthepreviouscolumnsofTable2.Ofgreaterimportance,incolumn(5),thecoefﬁcientontheinteractiontermAbnormalreturn×ψispositiveand

WehavealsoestimatedaLogitmodelinwhichthedependentvariableisadummythatissettoonewhenDividendchangeisnotequaltozero.Otherwise,thede-pendentvariableissettozero.Sincethegreatmajorityofnon-zerodividendchangesareincreases,theresultsoftheLogitmodelarequalitativelysimilartothemul-tinomialLogitﬁndingsforpositivedividendchanges.

12A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table5

Alternativeprivateinformationmeasuresanddividendchanges:OLSmodels.ThetablepresentsOrdinaryLeastSquaresregressionresultsforthechangeinquarterlydividendpaymentsonthepre-eventabnormalreturn,ameasureofprivateinformationinstockprices(alternativetotheﬁrm-speciﬁcreturnvariation),theinteractionbetweenthesetwovariables,andasetofcontrolvariablesincludingtwo-digitSICindustrydummies(notreportedinthetable).Observationswithzerodividendchangesareexcludedfromthesample.DeﬁnitionsforallthevariablescanbefoundinTable1.t-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

Dividendchange(1)

AbnormalreturnIlliq

Abnormalreturn×Illiqγ

Abnormalreturn×γPIN

Abnormalreturn×PINDividendyieldMarketcapitalizationDebtCash

−1.679⁎⁎⁎(−15.119)−0.000(−0.598)0.028(1.272)0.197⁎⁎⁎(6.084)−0.029⁎⁎⁎(−8.118)0.524⁎⁎⁎(11.229)1.261⁎⁎⁎(5.312)0.222⁎⁎⁎(5.544)16,2400.095

−1.539⁎⁎⁎(−13.151)−0.000(−0.802)0.021(0.924)0.220⁎⁎⁎(6.1)−0.032⁎⁎⁎(−8.611)0.536⁎⁎⁎(11.277)1.349⁎⁎⁎(5.467)0.198⁎⁎⁎(5.291)15,5340.091

0.103⁎⁎⁎(9.601)0.001⁎⁎⁎(4.706)0.003⁎⁎⁎(3.288)

(2)0.131⁎⁎⁎(13.485)

(3)0.088⁎(1.776)

0.000(0.001)0.207⁎⁎(2.524)

0.301⁎⁎⁎(5.014)0.196(0.768)−7.331⁎⁎⁎(−13.858)0.004⁎⁎⁎(5.624)−0.097⁎⁎⁎(−2.777)0.099⁎⁎(2.240)−0.037⁎⁎⁎(−5.7)0.599⁎⁎⁎(7.820)−2.617⁎⁎⁎(−6.183)0.357⁎⁎⁎(13.070)5,2030.150

highlystatisticallysigniﬁcant.Thissuggeststhatﬁrm-speciﬁcstockreturnvariationstrengthensthepositiverelationshipbetweenthelikelihoodofapositivechangeindividendsandtherecentabnormalstockperformance.Asimilarconclusioncanbedrawninrelationtotheprobabilityofdividendcuts(column(6)):thenegativeimpactofthepre-eventabnormalreturnonsuchprobabilitybecomeslarg-erwhenψincreases.Sinceﬁrm-speciﬁcstockreturnvariationmeasurestheextenttowhichprivateinformationisimpoundedinstockprices,theseﬁndingssuggestthatmanagersaremorelikelytoalterdividendlevelsfollowingabnormalchangesinstockvaluationswhenstockreturnsconveymoreprivateinformation.

InTable2,coefﬁcientestimatesforotherindependentvariablesgenerallyhavetheexpectedsigns.Highdividendyieldsmakedivi-dendcutsmorelikelyandreducetheprobabilityofdividendincreases.Incontrast,largeandproﬁtableﬁrmstendtoincreasedividendpaymentswhereasthelikelihoodofadividendcutisgreaterforsmallﬁrmswithlowproﬁts(seeﬁndingsforMarketcapitalizationandOperatingincome).Thereisanegative(positive)relationshipbetweenMarket-to-bookandthelikelihoodofdividendincreases(de-creases),implyingthatﬁrmswithbetterinvestmentopportunitiestendtopaylowerdividendsastheyvaluecashmore.

Table3replicatesthethreespeciﬁcationsofTable2usingOrdinaryLeastSquares(OLS)regressions.ThedependentvariableinTable3istherelativechangeinsuccessivequarterlydividends(Dividendchange).Asnotedabove,inthegreatmajorityofoccurrencesquarterlydividendsarenotchangedinoursample.Sinceweareprimarilyinterestedindividendincreasesanddecreasesfollowingabnormalchangesinstockprices,inourcontinuousmodelswedroptheobservationswithzerochangesinquarterlydividends.13Wetestwhetherourkeyﬁndingsaredrivenbythischoiceandestimateseveralregressionmodelsusingsamplesthatalsoincludeobservationswithzerochangesinquarterlydividends.Aneconometricchallengewefaceisduetothefactthatthedistributionofdividendchangeshasamasspointatzero.Atthesametime,dividendchangescanalsobenegative,eventhoughdividenddecreasesareveryinfrequent.We,therefore,relyontwodifferenttypesofTobitmodelstore-estimatethespec-iﬁcationsofTables3and5.IntheﬁrsttypeofTobitmodel,wesetthevariableDividendchangetozerowheneverthereisadividenddecrease.Inthesecond,observationswithdividenddecreasesaredropped.Hence,thedistributionsofthesamplesweuseintheTobitregressionshaveacornersolutionoutcomeatzero.ATobitestimationtechniqueisparticularlysuitableinthiscase.Overall,theestimatesoftheTobitmodelsprovideﬁndingsfortheinteractionsbetweenAbnormalreturnandthemeasuresofprivateinformationinstockpricesthatarequalitativelysimilartothosereportedinTables3and5.

1312A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table6

Privateinformationinstockprices,earningsannouncementreturn,anddividendchanges:MultinomialLogitmodels.ThetablepresentsMultinomialLogitregressionresultsforthelikelihoodofanincrease(columns(1),(3),(5),and(7))oradecrease(columns(2),(4),(6),and(8))inquarterlydividendpaymentsonthepre-eventabnormalreturn,ameasureofprivateinformationinstockprices,theinteractionbetweenthesetwovariables,theaverageabsolutemarket-adjustedreturngeneratedbyanearningsannouncement(ERC),andtheinteractionbetweenthisvariableandthelaggedabnormalreturn.Othercontrolvariables(notreportedinthetable)areDividendyield,Marketcapitalization,Debt,Cash,Market-to-book,Operatingincome,Dividendpremium,andtwo-digitSICindustrydummies.Thebaseoutcomeinthemodelisthezerochangeinquarterlydividendpayments.ThedependentvariableisbasedonthecontinuousvariableDividendchange.ERCistheaverageoftheabsolutemarket-adjustedreturnsaroundquarterlyearningsannouncements(fromday−1today+1)forthepreviouscalendaryear.DeﬁnitionsforalltheothervariablescanbefoundinTable1.z-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

DividendchangeN0(1)

Abnormalreturnψ

Abnormalreturn×ψIlliq

Abnormalreturn×Illiqγ

Abnormalreturn×γPIN

Abnormalreturn×PINERC

Abnormalreturn×ERCObservationsPseudo-R2−6.154⁎⁎⁎(−12.511)2.543(1.381)81,4070.0270

9.907⁎⁎⁎(8.042)6.005(1.217)81,4070.0270

−6.469⁎⁎⁎(−13.11)2.124(1.144)81,4070.0262

10.280⁎⁎⁎(8.041)7.751(1.556)81,4070.0262

−6.367⁎⁎⁎(−12.295)3.032(1.595)78,5200.02

10.155⁎⁎⁎(8.035)8.476⁎(1.6)78,5200.02

0.196(1.283)−0.075⁎⁎⁎(−5.661)0.112⁎⁎⁎(2.869)

Dividendchangeb0(2)−0.9⁎(−1.934)0.017(0.463)−0.344⁎⁎(−2.458)

DividendchangeN0(3)0.335⁎⁎(2.556)

Dividendchangeb0(4)−1.524⁎⁎⁎(−3.536)

DividendchangeN0(5)0.356⁎⁎⁎(2.71)

Dividendchangeb0(6)−1.828⁎⁎⁎(−4.176)

DividendchangeN0(7)−0.210(−0.608)

Dividendchangeb0(8)0.038(0.033)

0.003⁎⁎(2.176)0.010⁎⁎(2.507)

−0.005(−1.253)−0.034⁎⁎(−2.457)

−0.132(−1.46)0.125(0.2)

−0.566(−1.572)−2.531(−1.459)

0.045(0.121)3.624⁎⁎⁎(2.3)−8.663⁎⁎⁎(−9.331)1.580(0.505)34,0170.0379

−2.256⁎(−1.797)−8.214⁎(−1.732)7.883⁎⁎⁎(3.496)4.287(0.4)34,0170.0379

⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

ConﬁrmingtheempiricaloutcomesoftheMultinomialLogitregressions,incolumn(1)ofTable3weﬁndthatthecoefﬁcientonthevariableAbnormalreturnispositiveandstatisticallysigniﬁcant.Largerrecentabnormalstockreturnsleadtogreaterchangesinquarterlydividendpayments.Incolumn(2),wereportastatisticallysigniﬁcantnegativeimpactofψonDividendchange.Companieswithmoreprivateinformationinstockpricestendtochoosesmallerrelativechangesintheirquarterlydividends.Moreimportantly,weshowthatﬁrm-speciﬁcstockreturnvariationampliﬁestheeffectofthepre-eventabnormalreturnsoncurrentquarterlydividendchanges(seetheresultfortheinteractiontermincolumn(3)).Dividend-settingmanagersseemtoﬁndstockpricechangesmoremeaningfulandinformativewhenstockpricesaremorelikelytoimpoundandconveyprivateinformation.

OtherstatisticallysigniﬁcantﬁndingsfromTable3indicatethatquarterlydividendchangestendtodecreasewiththeDividendyield,theMarket-to-bookratio,and,surprisingly,Marketcapitalization.Ontheotherhand,companieswithgreatervaluesofthevariablesOperatingincomeandDividendpremiumchooselargerchangesinquarterlydividendpayments.4.3.Alternativemeasuresofprivateinformationinstockprices

Toinvestigatetherobustnessofourresults,weperformadditionaltests.Weconsiderasetofalternativemeasurescomprisingtheilliquidityratio(Illiq)byAmihud(2002),theprivateinformationtradingmeasure(γ)createdbyLlorenteetal.(2002),andtheprob-abilityofinformedtrading(PIN).Increasesinthevaluesofanyofthesemeasuresshouldbeassociatedwithmoreprivateinformationinstockprices.Onthewhole,thekeyﬁndings,althoughstatisticallyweakerthanwhenconsideringtheﬁrm-speciﬁcreturnvariation,arequalitativelyunaffectedbytheuseofthesealternativeprivateinformationmeasures.

EstimatesforseveralMultinomialLogitmodelscanbefoundinTable4.CoefﬁcientsonAbnormalreturnaregenerallystatisticallysigniﬁcantandtheyhavetheexpectedsigns.Theyarepositiveinregressionsofdividendincreasesandnegativeinmodelsfordividendcuts.InteractiontermsbetweenAbnormalreturnandprivateinformationmeasuresarepositivelyrelatedwiththelikelihoodofdividendincreases.Further,theserelationshipsarestatisticallysigniﬁcantatstandardlevelsexceptinthespeciﬁcationforγ.WhiletheresultsforpositivedividendchangesconﬁrmwhatisreportedinTable2,regressionﬁndingsfordividendcutsarelesssupportivesincetheyarestatisticallyinsigniﬁcant.

MovingontomodelswiththecontinuousDividendchangedependentvariable(Table5),weagainﬁndthatgreaterrecentabnormalchangesinstockpricesarefollowedbylargerrelativechangesinquarterlydividendpayments.Moreimportantly,thisrelationshipgets

A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–1713

Table7

Privateinformationinstockprices,earningsannouncementreturn,anddividendchanges:OLSmodels.ThetablepresentsOrdinaryLeastSquaresregressionresultsforthechangeinquarterlydividendpaymentsonthepre-eventabnormalreturn,ameasureofprivateinformationinstockprices,theinteractionbetweenthesetwovariables,theaverageabsolutemarket-adjustedreturngeneratedbyanearningsannouncement(ERC),andtheinteractionbetweenthisvariableandthelaggedabnor-malreturn.Othercontrolvariables(notreportedinthetable)areDividendyield,Marketcapitalization,Debt,Cash,Market-to-book,Operatingincome,Dividendpremium,andtwo-digitSICindustrydummies.Observationswithzerodividendchangesareexcludedfromthesample.ERCistheaverageoftheabsolutemarket-adjustedreturnsaroundquarterlyearningsannouncements(fromday−1today+1)forthepreviouscalendaryear.DeﬁnitionsforalltheothervariablescanbefoundinTable1.t-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

Dividendchange(1)

Abnormalreturnψ

Abnormalreturn×ψIlliq

Abnormalreturn×Illiqγ

Abnormalreturn×γPIN

Abnormalreturn×PINERC

Abnormalreturn×ERCObservationsAdjustedR20.069(0.572)1.230⁎⁎⁎(2.837)13,80.0

−0.071(−0.578)1.081⁎⁎(2.485)13,80.092

0.016(0.132)1.196⁎⁎⁎(2.625)13,3390.086

0.011(0.431)−0.010⁎⁎⁎(−3.827)0.026⁎⁎⁎(2.792)

(2)0.041⁎(1.711)

(3)0.060⁎⁎(2.375)

(4)−0.003(−0.039)

0.001⁎⁎⁎(5.4)0.003⁎⁎⁎(2.836)

0.008(0.463)0.162⁎(1.808)

0.322⁎⁎⁎(5.1)0.291(1.097)−0.271(−1.485)1.268(1.639)5,0080.155

⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

strongerwhenstockpricechangesaremoreinformative,i.e.,whenstockmarketvaluationsreﬂectandconveymoreprivateinformationthatmanagersmaynotpossess.Itis,infact,thecasethatthecoefﬁcientsonalltheinteractiontermsbetweenAbnormalreturnandmea-suresofprivateinformationinstockpricesarepositive.However,thecoefﬁcientonAbnormalreturn×PINisnotstatisticallysigniﬁcantpossiblyowingtothesmallersamplesize.

Afterfurtherinvestigatingourresearchquestionbyexploitingseveralalternativemeasuresofprivateinformationinstockprices,wecanstatethattheconclusionsdrawninrelationtotheﬁrm-speciﬁcstockreturnvariationhold.Pastabnormalstockreturnsexer-ciseafarmoresigniﬁcantimpactontheextentofcurrentdividendchangeswhentheyaremorelikelytobeinformativeformanagers.Towit,privateinformationinstockpricesstrengthenstheimpactofpastrevisionsinstockpricesoncurrentvariationsindividendpayments.

4.4.Managerialprivateinformation

Theempiricalﬁndingswehavehighlightedsofarsuggestthatmanagerslearnfromrecentstockpricechangesandobtainpreciousinformationthattheytakeintoaccountwhendecidingchangesinquarterlydividends.However,theprivateinformationmeasuresofthisstudydonotallowustodistinguishbetweeninformationthatisintheexclusivedomainofoutsideinvestorsandinformationthatissharedbyinsidersandoutsidersatthesametime.Abnormalincreasesinstockpricesmaybecausedbytradesfromoutsideinves-torsactinguponrecently-acquiredprivateinformation.Followingsuchpricechanges,managersmaygoaheadwithdividendin-creaseseitherbecausetheylearnnewinformationfromtherecentvariationsinstockpricesorbecausetheymoredirectlypossessthesameinformationthatoutsidersexploitedwhentradingstock.

Werecognizethatitisnotfeasibletopreciselyseparatetheprivateinformationthatisdirectlyandindependentlyacquiredbymanagersintheperiodprecedingadividendannouncementfromwhatislearnedfromrecentstockpricechanges.Nevertheless,inourempiricaltests,weattempttocontrolforthelevelofprivateinformationthatisheldbymanagers.Todoso,wefollowChenetal.(2007)andFresard(2012)andusethevariableearnings'surprise(ERC)asaproxyofmanagerialprivateinformation.Thisistheaverageoftheabsolutemarket-adjustedreturnsgeneratedbythefourquarterlyearningsannouncements(fromCompustat)thatprecedethecurrentannouncementofaquarterlydividendpayment.WeusetheCRSPvalue-weightedmarketreturnandthree-daywindowsaroundtheearningsannouncementstocomputethemarket-adjustedreturns.Intuitively,ERCshouldreﬂecttheextenttowhichinvestorsaresurprisedbyearningsannouncements.Ifmanagersdonotpossesssigniﬁcantprivateinformation

14A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table8

Privateinformationinstockprices,stockovervaluation,anddividendchanges:MultinomialLogitmodels.ThetablepresentsMultinomialLogitregressionresultsforthelike-lihoodofanincrease(columns(1),(3),(5),and(7))oradecrease(columns(2),(4),(6),and(8))inquarterlydividendpaymentsonthepre-eventabnormalreturn,amea-sureofprivateinformationinstockprices,theinteractionbetweenthesetwovariables,thelogoftheratiobetweenthemarketvalueandthe“fairvalue”ofequity(M/V),andtheinteractionbetweenthisvariableandthelaggedabnormalreturn.Othercontrolvariables(notreportedinthetable)areDividendyield,Marketcapitalization,Debt,Cash,Market-to-book,Operatingincome,Dividendpremium,andtwo-digitSICindustrydummies.Thebaseoutcomeinthemodelisthezerochangeinquarterlydividendpayments.ThedependentvariableisbasedonthecontinuousvariableDividendchange.DeﬁnitionsforthisandalltheothervariablesexceptM/VcanbefoundinTable1.M/VisestimatedfollowingRhodes-Kropfetal.(2005)andtheirthirdvaluationmodel.Thevariableisthesumoftheﬁrm-speciﬁcerrorandthetime-seriessectorerror.ThemarketvalueofequityusedtoﬁndM/VisthatattheendofthemeasurementperiodofAbnormalreturn.z-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

DividendchangeN0(1)

Abnormalreturnψ

Abnormalreturn×ψIlliq

Abnormalreturn×Illiqγ

Abnormalreturn×γPIN

Abnormalreturn×PINM/V

Abnormalreturn×M/VObservationsPseudo-R2−0.146⁎⁎⁎(−3.549)−0.317⁎⁎⁎(−3.949)98,5350.0227

−0.086(−0.703)0.800⁎⁎(2.566)98,5350.0227

−0.145⁎⁎⁎(−3.531)−0.299⁎⁎⁎(−3.685)98,5350.0220

−0.0(−0.5259)0.918⁎⁎⁎(2.968)98,5350.0220

−0.139⁎⁎⁎(−3.359)−0.367⁎⁎⁎(−4.471)94,4790.0225

−0.081(−0.665)0.902⁎⁎⁎(2.844)94,4790.0225

0.372⁎⁎⁎(4.034)−0.070⁎⁎⁎(−5.245)0.118⁎⁎⁎(3.239)

Dividendchangeb0(2)−0.569⁎(−1.772)0.011(0.34)−0.215⁎(−1.659)

DividendchangeN0(3)0.503⁎⁎⁎(7.8481)

Dividendchangeb0(4)−1.145⁎⁎⁎(−5.622)

DividendchangeN0(5)0.588⁎⁎⁎(10.694)

Dividendchangeb0(6)−1.016⁎⁎⁎(−5.834)

DividendchangeN0(7)0.286(1.024)

Dividendchangeb0(8)−0.125(−0.135)

−0.002(−1.607)0.010⁎⁎⁎(2.815)

0.007⁎⁎(2.466)0.010(1.019)

−0.244⁎⁎⁎(−2.826)0.257(0.657)

−0.486(−1.637)−1.525(−1.045)

−0.003(−0.008)2.219⁎(1.698)−0.071(−1.32)−0.550⁎⁎⁎(−4.285)35,3590.0327

−2.441⁎⁎(−2.016)−5.718(−1.209)−0.239(−1.357)0.285(0.504)35,3590.0327

⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

thatgivesthemalargeinformationaladvantageoveroutsiders,investors'reactionstoearningsannouncementsshouldbequitemuted.Incontrast,whenmanagersholdverysigniﬁcantprivateinformation,earningssurprisesshouldbequitesubstantial.

InordertocontrolformanagerialprivateinformationandfollowingtheapproachofChenetal.(2007)andFresard(2012),weaddthevariableERCandtheinteractionofthisvariablewithAbnormalreturntoallourmultivariatespeciﬁcations.14StartingfromtheMultinomialLogitmodelsofTable6,15weobserveastatisticallysigniﬁcantnegative(positive)effectofERContheprobabilityofadividendincrease(decrease).ThecoefﬁcientontheinteractionAbnormalreturn×ERCispositivebutinsigniﬁcantinmostmodels.Ex-ceptinmodelsforγ,consistentwiththeresultsfromTables2and4,coefﬁcientsontheinteractiontermsbetweenAbnormalreturnandthemeasuresofprivateinformationinstockpricesarepositiveandstatisticallysigniﬁcantatstandardlevelsinthespeciﬁcationsfordividendincreases.Theyarenegativeandsigniﬁcantindividendcutspeciﬁcations.

TheOLSmodelsreportedinTable7usethecontinuousvariableDividendchange.AftercontrollingforERC,theresultsconﬁrmthepositiverelationshipsbetweenDividendchangeandtheinteractiontermsofAbnormalreturnwiththemeasuresofprivateinformationinstockprices.Overall,ourevidenceshowsthatthepre-eventAbnormalreturnhasalargerinﬂuenceondividendchangedecisionsformanagersacquiringnovel,privateinformationfromstockprices.4.5.Stockmarketovervaluation

Inamarketwithwell-informed,rationalinvestors,unexpectedchangesinstockvaluationsaredrivenbyreliablenovelinforma-tiononfuturecashﬂowsanddiscountrates.Inourstudy,weessentiallyassumethatpastabnormalreturnsmayaffectmanagerial

ByincludingthisinteractiontermweessentiallytestwhethertheﬁndingsfortheinteractionsbetweenAbnormalreturnandthemeasuresofprivateinformationinstockpricesareactuallydrivenbytheavailabilityofprivateinformationbymanagersthatcouldbecorrelatedwithsuchmeasures.Inuntabulatedanalyses,wealsotestwhethertheﬁndingspresentedinthissectionarerobusttotheinclusionofinteractiontermsbetweenERCandthemeasuresofprivateinformationinstockprices.Wecanconﬁrmtherobustnessofourmainresults.Further,ifthetripleinteractiontermbetweenAbnormalreturn,privateinformationinstockprices,andERCisaddedtoourspeciﬁcations,thecoefﬁcientsonthistermarealwaysstatisticallyinsigniﬁcant.15Forthesakeofbrevity,inTables6–9wedonotreportcoefﬁcientsforthefullsetofcontrolvariables.FindingsfortheomittedvariablesarequalitativelysimilartothoseinTables2–4.

14A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–1715

Table9

Privateinformationinstockprices,stockovervaluation,anddividendchanges:OLSmodels.ThetablepresentsOrdinaryLeastSquaresregressionresultsforthechangeinquarterlydividendpaymentsonthepre-eventabnormalreturn,ameasureofprivateinformationinstockprices,theinteractionbetweenthesetwovariables,thelogoftheratiobetweenthemarketvalueandthe“fairvalue”ofequity(M/V),andtheinteractionbetweenthisvariableandthelaggedabnormalreturn.Othercontrolvariables(notreportedinthetable)areDividendyield,Marketcapitalization,Debt,Cash,Market-to-book,Operatingincome,Dividendpremium,andtwo-digitSICindustrydummies.Observationswithzerodividendchangesareexcludedfromthesample.M/VisestimatedfollowingRhodes-Kropfetal.(2005)andtheirthirdvaluationmodel.Thevariableisthesumoftheﬁrm-speciﬁcerrorandthetime-seriessectorerror.ThemarketvalueofequityusedtoﬁndM/VisattheendofthemeasurementperiodofAbnormalreturn.DeﬁnitionsforalltheothervariablescanbefoundinTable1.t-statisticsadjustedforclusteringbyﬁrmandcalendarmontharereportedinparentheses.

Dividendchange(1)

Abnormalreturnψ

Abnormalreturn×ψIlliq

Abnormalreturn×Illiqγ

Abnormalreturn×γPIN

Abnormalreturn×PINM/V

Abnormalreturn×M/VObservationsAdjustedR20.004(0.567)−0.053⁎⁎(−2.477)16,2400.095

0.009(1.079)−0.048⁎⁎(−2.254)16,2400.096

0.005(0.565)−0.061⁎⁎⁎(−2.742)15,5340.092

0.084⁎⁎⁎(4.858)−0.009⁎⁎⁎(−3.900)0.025⁎⁎⁎(3.030)

(2)0.106⁎⁎⁎(9.365)

(3)0.133⁎⁎⁎(13.430)

(4)0.109⁎⁎(2.074)

0.001⁎⁎⁎(4.825)0.002⁎⁎⁎(2.659)

−0.001(−0.093)0.196⁎⁎(2.394)

0.309⁎⁎⁎(5.114)0.112(0.429)0.015(1.314)−0.050(−1.193)5,2030.151

⁎⁎⁎Denotesstatisticalsigniﬁcanceatthe1%level.⁎⁎Denotesstatisticalsigniﬁcanceatthe5%level.⁎Denotesstatisticalsigniﬁcanceatthe10%level.

choicesonfuturedividendchangessinceabnormalreturnsareprimarilygeneratedbytradesfrominformedinvestors.Whilethisas-sumptionmaybeplausible,wecannotruleoutthepossibilitythattransactionsbypoorlyinformednoisetraderscouldalsoproducesigniﬁcanteffectsonstockmarketvaluations.Theseveralprivateinformationmeasuresweuseshouldhelpusdistinguishbetweeninformativeanduninformativepastchangesinstockprices.Theﬁndingswereportabovesuggestthatthesemeasuressuccessfullyachievethisgoalsincewereportastrongerrelationshipbetweenpastabnormalreturnsandcurrentdividendchangeswhenstockpricechangesconveymoreprivateinformation.

Nevertheless,todirectlycontrolforthepotentialeffectsofnoisetrading,weextendourbaselineanalysesbyconsideringamea-sureofstockovervaluation.WefollowthemethoddevisedbyRhodes–Kropfetal.(2005)tocomputesuchmeasure.LikeHertzelandLi(2010),werelyonthethirdvaluationmodelofRhodes-Kropfetal.(2005)thatconsidersbookvalueofequity,netincome,andmar-ketleverageratioasdeterminants.Ourovervaluationmeasure(M/V)isthelogarithmoftheratiobetweenthemarketvalueofaﬁrmandits“truevalue”.Theﬁttedestimateofthevaluationmodelsisconsideredtheﬁrm'struevalue.Themeasurereﬂectstwodifferentsourcesofovervaluationthatarebothrelevantinourstudy:theﬁrm-speciﬁcerrorandthetime-seriessectorerror.HighervaluesofM/Vindicatemoreovervaluation,andviceversa.ForaparticularDividendchangeevent,M/Viscomputedusingthemarketvalueoftheﬁrmattheendofthesixthdaybeforetheevent.Thedayischosenbecauseitprecedestheeventanditistheﬁnaldayofthemea-surementperiodforAbnormalreturn.16InTable8,were-estimatetheMultinomialLogitmodelsofTables2and4,addingM/VandtheinteractiontermbetweenAbnormalreturnandM/V.Mostcoefﬁcientsonthisinteractionarestatisticallysigniﬁcant.Thesignsofthecoefﬁcientssuggestthatﬁrmover-valuationweakensthepositive(negative)relationshipbetweenrecentabnormalreturnsandthelikelihoodofpositive(negative)changesindividendpayments.Thisevidenceseemstosupportthenotionthatmanagersarelesslikelytoconsiderpastreturnswheninvestorstendtomisvaluetheircompanyandareunreasonablyoptimisticaboutitsprospects.Inotherwords,managersseemtobeabletorecognizeanddisregardchangesinstockpricesthataredrivenbytransactionsfromnoisetraders.Moreimportant-ly,evenaftercontrollingforstockovervaluation,wecanstillconﬁrmtheresultsofTables2and4regardingtheinteractiontermsbetweenAbnormalreturnandourprivateinformationmeasures.

16Weobtainqualitativelysimilarresultsifweusethemarketvalueoftheﬁrmatthestartofthisperiod,i.e.,onthesixthdayafterthepreviousquarterlydividendannouncement.

16A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–17

Table9includesOLSregressionmodelswithM/VandAbnormalreturn×M/V.Coefﬁcientsforthisinteractiontermarealwaysneg-ativeandstatisticallydifferentfromzerointhreecasesoutoffour.Theseresultssupporttheconclusionswehavedrawnabove.More-over,byanalyzingthecoefﬁcientsontheinteractionsbetweenpre-eventabnormalreturnsandprivateinformationmeasures,wecanagainconﬁrmthemainﬁndingsofthisstudy:whendecidingquarterlydividendpayments,managersseemtogivemorecredencetotheinformationconveyedbypaststockpricechangesiftheyarelikelytobemoreinformative.5.Conclusion

Westudytheeffectofprivateinformationinstockpricesontherelationshipbetweenpastabnormalstockreturnsandcurrentchangesinquarterlydividends.Bothcashﬂowanddiscountratenewsonaﬁrmshouldleadinvestorstore-estimatethevalueoftheﬁrmandtradeitssharesaccordingly(e.g.,CampbellandShiller,1988;ChenandZhao,2009).Thus,abnormalchangesinthemar-ketvalueofaﬁrmarelikelytoreﬂectnovelinformationabouttheﬁrmandshouldbecorrelatedwithfuturevariationsindividendpolicy.Inlinewiththisargument,wereportthatrecentabnormalstockreturnshaveapositive(negative)impactonthelikelihoodofaquarterlydividendincrease(decrease).Further,thecurrentchangeinquarterlydividendsispositivelyrelatedtopastabnormalreturns.

Moreimportantly,weﬁndthatalltheserelationshipsarestrongerwhenstockreturnsaremorelikelytoconveyprivateinforma-tionthatisnewtomanagers.Asmeasuresofprivateinformationinstockpricesweconsiderﬁrm-speciﬁcstockreturnvariation,theilliquidityratiobyAmihud(2002),thetradingmeasuredevelopedbyLlorenteetal.(2002),andtheprobabilityofinformedtrading(PIN).Ourresultsarerobusttotheinclusionofaproxyformanagerialprivateinformationthatweusetocontrolforinformationcon-veyedbystockpricechangesthatisalreadypossessedbymanagers.Similarly,whenwecontrolforameasureofstockovervaluationonthemarket,wecanstillconﬁrmthemainﬁndingsofthepaper.Wealsoreportweakereffectsofpastreturnsoncurrentdividendchangeswhenstocksareovervalued.Asetofstandarddeterminantsofdividendchangesisalsoaddedtoallourspeciﬁcations.

Theﬁndingsreportedinthisstudysuggestthatnovelprivateinformationconveyedbystockreturnsisexploitedbymanagerswhendecidingthedividendpoliciesoftheircompanies.Ourstudy,therefore,showstheimportanceofprivateinformationinstockpricesasadeterminantofdividendpolicy.Moreover,consistentwithKauetal.(2008),ourempiricalevidencesupportstheideathatmanagers“listentothemarket”sincecorporateeventsandchangesincorporatepoliciesaresensitivetomarketreactions.Ourstudycontributestothegrowingliteraturehighlightingthenotionthatthestockmarketisnotsimplyasideshowbutithas,instead,asigniﬁcanteffectonrealeconomicactivitythroughtheinformationalroleofstockprices(Bondetal.,2012).Inparticular,thisstudyhighlightsthatveryfrequentandsigniﬁcantcorporateﬁnancialdecisionssuchasdividendpaymentsareaffectedbychang-esinﬁnancialmarketvariablesthroughaninformationalchannel.Acknowledgments

WethankJeffryNetter(theManagingEditor),ananonymousrefereeandseminarparticipantsatthe2014FinancialManagementAssociationmeetinginNashville,the2014AsianFinanceAssociationmeetinginBali,andthe2014TaiwanFinanceAssociationmeet-inginHsinchufortheirhelpfulcomments.However,theauthorsaresolelyresponsibleforanyremainingerrorsinthepaper.References

Allen,F.,Michaely,R.,2003.Payoutpolicy.Handb.Econ.Financ.1,337–429.

Amihud,Y.,2002.Illiquidityandstockreturns:cross-sectionandtime-serieseffects.J.Financ.Mark.5,31–56.Baker,M.,Wurgler,J.,2004.Acateringtheoryofdividends.J.Financ.59,1125–1165.

Bakke,T.E.,Whited,T.M.,2010.Whichfirmsfollowthemarket?Ananalysisofcorporateinvestmentdecisions.Rev.Financ.Stud.23,1941–1980.Benartzi,S.,Michaely,R.,Thaler,R.,1997.Dochangesindividendssignalthefutureorthepast?J.Financ.52,1007–1034.Bhattacharya,S.,1979.Imperfectinformation,dividendpolicy,and“thebirdinthehand”fallacy.BellJ.Econ.10,259–270.Blau,B.M.,Fuller,K.P.,2008.Flexibilityanddividends.J.Corp.Financ.14,133–152.

Bonaimé,A.A.,Hankins,K.W.,Harford,J.,2014.Financialflexibility,riskmanagement,andpayoutchoice.Rev.Financ.Stud.27,1074–1101.Bond,P.,Edmans,A.,Goldstein,I.,2012.Therealeffectsoffinancialmarkets.Annu.Rev.Financ.Econ.4,339–360.Brav,A.,Graham,J.R.,Harvey,C.R.,Michaely,R.,2005.Payoutpolicyinthe21stcentury.J.Financ.Econ.77,483–527.Brockman,P.,Yan,X.,2009.Blockownershipandfirm-specificinformation.J.Bank.Financ.33,308–316.

Brown,J.R.,Liang,N.,Weisbenner,S.,2007.Executivefinancialincentivesandpayoutpolicy:firmresponsestothe2003dividendtaxcut.J.Financ.62,1935–1965.Bulan,L.,Subramanian,N.,Tanlu,L.,2007.Onthetimingofdividendinitiations.Financ.Manag.36,31–65.

Campbell,J.Y.,Shiller,R.J.,1988.Thedividend–priceratioandexpectationsoffuturedividendsanddiscountfactors.Rev.Financ.Stud.1,195–228.

Chan,K.,Chan,Y.,2014.Priceinformativenessandstockreturnsynchronicity:evidencefromthepricingofseasonedequityofferings.J.Financ.Econ.114,36–53.Chan,K.,Hameed,A.,2006.Stockpricesynchronicityandanalystcoverageinemergingmarkets.J.Financ.Econ.80,115–147.

Charitou,A.,Lambertides,N.,Theodoulou,G.,2011.Dividendincreasesandinitiationsanddefaultriskinequityreturns.J.Financ.Quant.Anal.46,1521–1543.

Chemmanur,T.J.,He,J.,Hu,G.,Liu,H.,2010.Isdividendsmoothinguniversal?NewinsightsfromacomparativestudyofdividendpoliciesinHongKongandtheU.S.

J.Corp.Financ.16,413–430.

Chen,L.,Zhao,X.,2009.Returndecomposition.Rev.Financ.Stud.22,5213–5249.

Chen,Q.,Goldstein,I.,Jiang,W.,2007.Priceinformativenessandinvestmentsensitivitytostockprice.Rev.Financ.Stud.20,619–650.Chen,L.,Da,Z.,Zhao,X.,2013.Whatdrivesstockpricemovements?Rev.Financ.Stud.26,841–876.

Crawford,S.S.,Roulstone,D.T.,So,E.C.,2012.Analystinitiationsofcoverageandstockreturnsynchronicity.Account.Rev.87,1527–1553.

Dasgupta,S.,Gan,J.,Gao,N.,2010.Transparency,priceinformativeness,andstockreturnsynchronicity:theoryandevidence.J.Financ.Quant.Anal.45,11–1220.DeAngelo,H.,DeAngelo,L.,Skinner,D.J.,1992.Dividendsandlosses.J.Financ.47,1837–1863.

DeAngelo,H.,DeAngelo,L.,Skinner,D.J.,1996.Reversaloffortune.Dividendsignalingandthedisappearanceofsustainedearningsgrowth.J.Financ.Econ.40,341–371.DeAngelo,H.,DeAngelo,L.,Skinner,D.J.,2008.Corporatepayoutpolicy.Found.TrendsFinanc.3,95–287.

Defond,M.L.,Hung,M.,2004.Investorprotectionandcorporategovernance:evidencefromworldwideCEOturnover.J.Account.Res.42,269–312.

A.DeCesari,W.Huang-Meier/JournalofCorporateFinance35(2015)1–1717

Dow,J.,Gorton,G.,1997.Stockmarketefficiencyandeconomicefficiency:isthereaconnection?J.Financ.52,1087–1129.

Durnev,A.,Morck,R.,Yeung,B.,Zarowin,P.,2003.Doesgreaterfirm-specificreturnvariationmeanmoreorlessinformedstockpricing?J.Account.Res.41,797–836.Durnev,A.,Morck,R.,Yeung,B.,2004.Value-enhancingcapitalbudgetingandfirm-specificstockreturnvariation.J.Financ.59,65–105.Dyck,A.,Zingales,L.,2004.Privatebenefitsofcontrol:aninternationalcomparison.J.Financ.59,537–600.

Easley,D.,Hvidkjaer,S.,O'Hara,M.,2002.Isinformationriskadeterminantofassetreturns?J.Financ.57,2185–2221.

Fama,E.F.,French,K.R.,2001.Disappearingdividends:changingfirmcharacteristicsorlowerpropensitytopay?J.Financ.Econ.60,3–43.Faulkender,M.,Wang,R.,2006.Corporatefinancialpolicyandthevalueofcash.J.Financ.61,1957–1990.

Fernandes,N.,Ferreira,M.A.,2008.Doesinternationalcross-listingimprovetheinformationenvironment.J.Financ.Econ.88,216–244.Fernandes,N.,Ferreira,M.A.,2009.Insidertradinglawsandstockpriceinformativeness.Rev.Financ.Stud.22,1845–1887.Ferreira,M.A.,Laux,P.A.,2007.Corporategovernance,idiosyncraticrisk,andinformationflow.J.Financ.62,951–9.Ferreira,D.,Ferreira,M.A.,Raposo,C.C.,2011.Boardstructureandpriceinformativeness.J.Financ.Econ.99,523–545.Foucault,T.,Fresard,L.,2014.Learningfrompeers'stockpricesandcorporateinvestment.J.Financ.Econ.111,554–577.Fresard,L.,2012.Cashsavingsandstockpriceinformativeness.Eur.Finan.Rev.16,985–1012.

Grullon,G.,Michaely,R.,2002.Dividends,sharerepurchases,andthesubstitutionhypothesis.J.Financ.57,19–1684.Grullon,G.,Michaely,R.,Swaminathan,B.,2002.Aredividendchangesasignoffirmmaturity?J.Bus.75,387–424.

Grullon,G.,Michaely,R.,Benartzi,S.,Thaler,R.H.,2005.Dividendchangesdonotsignalchangesinfutureprofitability.J.Bus.78,1659–1682.Healy,P.M.,Palepu,K.G.,1988.Earningsinformationconveyedbydividendinitiationsandomissions.J.Financ.Econ.21,149–175.

Hertzel,M.G.,Li,Z.,2010.BehavioralandrationalexplanationsofstockpriceperformancearoundSEOs:evidencefromadecompositionofmarket-to-bookratios.

J.Financ.Quant.Anal.45,935–958.

Hutton,A.P.,Marcus,A.J.,Tehranian,H.,2009.Opaquefinancialreports,R2,andcrashrisk.J.Financ.Econ.94,67–86.Jensen,M.C.,1986.Agencycostsoffreecashflow,corporatefinance,andtakeovers.Am.Econ.Rev.76,323–329.Jensen,G.R.,Lundstrum,L.L.,Miller,R.E.,2010.Whatdodividendreductionssignal?J.Corp.Financ.16,736–747.Jin,L.,Myers,S.C.,2006.R2aroundtheworld:newtheoryandnewtests.J.Financ.Econ.79,257–292.

John,K.,Williams,J.,1985.Dividends,dilution,andtaxes:asignallingequilibrium.J.Financ.40,1053–1070.Kau,J.B.,Linck,J.S.,Rubin,P.H.,2008.Domanagerslistentothemarket?J.Corp.Financ.14,347–362.Kyle,A.S.,1985.Continuousauctionsandinsidertrading.Econometrica53,1315–1335.

LaPorta,R.L.,Lopez-de-Silanes,F.,Shleifer,A.,Vishny,R.W.,2000.Agencyproblemsanddividendpoliciesaroundtheworld.J.Financ.55,1–33.

Lee,D.W.,Liu,M.H.,2011.Doesmoreinformationinstockpriceleadtogreaterorsmalleridiosyncraticreturnvolatility?J.Bank.Financ.35,1563–1580.Li,W.,Lie,E.,2006.Dividendchangesandcateringincentives.J.Financ.Econ.80,293–308.

Lie,E.,2005.Financialflexibility,performance,andthecorporatepayoutchoice.J.Bus.78,2179–2202.

Llorente,G.,Michaely,R.,Saar,G.,Wang,J.,2002.Dynamicvolume–returnrelationofindividualstocks.Rev.Financ.Stud.15,1005–1047.Luo,Y.,2005.Doinsiderslearnfromoutsiders?Evidencefrommergersandacquisitions.J.Financ.60,1951–1982.Miller,M.H.,Modigliani,F.,1961.Dividendpolicy,growth,andthevaluationofshares.J.Bus.34,411–433.Miller,M.H.,Rock,K.,1985.Dividendpolicyunderasymmetricinformation.J.Financ.40,1031–1051.

Morck,R.,Yeung,B.,Yu,W.,2000.Theinformationcontentofstockmarkets:whydoemergingmarketshavesynchronousstockpricemovements?J.Financ.Econ.58,

215–260.

Nissim,D.,Ziv,A.,2001.Dividendchangesandfutureprofitability.J.Financ.56,2111–2133.

Piotroski,J.D.,Roulstone,D.T.,2004.Theinfluenceofanalysts,institutionalinvestors,andinsidersontheincorporationofmarket,industry,andfirm‐specificinforma-tionintostockprices.Account.Rev.79,1119–1151.

Rhodes-Kropf,M.,Robinson,D.T.,Viswanathan,S.,2005.Valuationwavesandmergeractivity:theempiricalevidence.J.Financ.Econ.77,561–603.Roll,R.,1988.R2.J.Financ.43,541–566.

Subrahmanyam,A.,Titman,S.,1999.Thegoing-publicdecisionandthedevelopmentoffinancialmarkets.J.Financ.54,1045–1082.

Watanabe,A.,Xu,Y.,Yao,T.,Yu,T.,2013.Theassetgrowtheffect:insightsfrominternationalequitymarkets.J.Financ.Econ.108,529–563.Wurgler,J.,2000.Financialmarketsandtheallocationofcapital.J.Financ.Econ.58,187–214.

Xing,X.,Anderson,R.,2011.Stockpricesynchronicityandpublicfirm-specificinformation.J.Financ.Mark.14,259–276.

因篇幅问题不能全部显示，请点此查看更多更全内容

查看全文