现金流量财务风险外文文献

Availableonlineatwww.sciencedirect.comScienceDirectHOSTEDBYReviewofDevelopmentFinance5(2015)24–33CashflowforecastforSouthAfricanfirmsOpoAvensityAAbstractThispapeoperfirmslistedexareportedresnvenSouthAfricaia.Inimprovetheappregressivemfirms©2014AfrireserJELclassificaKeywords:Cmode1.IntroductionGiventhatcashflowisthelife-bloodofafirm,accuratedeterminatfinancialdgoesbankrcialhealththesourceinvestors’flowsrecoroperation,flow,reflecdayoperatimportance∗CorresponE-mailadLuiz.Moutinh(K.K.Opong)Peerreviepredictionoffuturecashflowsareofextremeusefulnessandvalue.Therearetwoissuestobeconsideredwhenattemptingtohttp://dx.doi.o1879-9337/©ionofcashflowsenablesfirmstomakeimportantecisionsthatrelatetowhetherthefirmsurvivesorupt.Asameasureofafirm’sprofitabilityandfinan-,cashflowscouldprovidepotentialcluesaboutcompany’sabilitytopaydividedandthusattractinteresttoo.Therearethreecategoriesofcashdedinstatementofcashflow,i.e.cashflowsfromfinancingandinvestment,ofwhichoperatingcashtingtheabilityofthefirmtoengageinday-to-ionsanditscontinuityinbusiness,isofthemost.Forthemanagersoffirms,investorsoranalysts,dingauthor.dresses:Yun.Li@glasgow.ac.uk(Y.Li),o@glasgow.ac.uk(L.Moutinho),Kwaku.Opong@Glasgow.ac.uk,Y.Pang.1@research.gla.ac.uk(Y.Pang).wunderresponsibilityofAfricagrowthInstitute.predictafirm’scashflows.First,thevariablesthoseareuse-fulandinformativetocashforecastneedtobeidentifiedandincorporatedintotheforecastmodel.Secondly,thetypeandstructureofmodelstobeemployedintheforecastshouldbecarefullychosentoprovideamoreaccurateprediction.Thisstudyshedlightonbothissues,intendingtodemonstratetheprocedureofchoosingvariablesandmodelsformoreaccurateprediction.Thereareanumberofdifficultieswithcashflowprediction.Generallyspeaking,cashflowismorevolatilethanearningsandthushardertopredict.Thereisnouniformcashflowgeneratingprocessforthewholebusinessworldanddifferentcompaniesprovidedistinctpatternsofcashflows.Besides,duetothepopularityofcredittrade,afirm’srevenueandexpensesarenotequaltocashinflowandoutflowandthiscompoundstheproblemofaccuratecashflowprediction.Academicstudiesoncashflowpredictionrelyonpublicinformationasreflectedinafirm’sfinancialstatementforcashflowdata.Amongthevariablesthathavebeenfoundusefulnessincashflowrg/10.1016/j.rdf.2014.11.0012014AfricagrowthInstitute.ProductionandhostingbyElsevierB.V.Allrightsreserved.YunLia,LuizMoutinhob,KwakuK.aSchoolofEngineering,UniversityofGlasgow,UniversitybAdamSmithBusinessSchool,UniversityofGlasgow,UniverrappliesmodelsintheextantliteraturethathavebeenusedtoforecastontheJohannesburgStockExchange.Out-of-sampleperformanceisultsshowthatsomeaccrualterms,i.e.depreciationandchangesininfirmincontrasttothereportedresultsofstudiesinUSAandAustralmodels,accordingtotheout-of-sampleresults.Thepaperproposestheodelformulti-period-aheadpredictionofcashflowsforSouthAfricacagrowthInstitute.ProductionandhostingbyElsevierB.V.Allrightstion:G300;M410;M490ashflow;Movingaverage;Predictivemodels;Accruals;Vectorautoregressivengb,YangPangb,∗ue,G128QQ,UnitedKingdomvenue,G128QQ,UnitedKingdomatingcashflowstopredictthecashflowsofSouthAfricanminedforeachmodelandcomparedbetweenthem.Thetorydonotenhancecashflowpredictionfortheaverageclusionofmoreexplanatoryvariablesdoesnotnecessarilylicationofmovingaveragemodelinpaneldata,andvector.ved.l;DepreciationY.Lietal./ReviewofDevelopmentFinance5(2015)24–3325predictionincludeearnings,accrualtermssuchaschangesinaccountreceivableandpayableanddisaggregatedcashcompo-nents.EmpandinformisonismadisexpectedamodelwimeansrichismadebewithdiffereexplanatorywhereasthepaperpropaveragemoflowpredicsenseasitinfluencepmodelisapisproposedVARismolinearregreofSouthAThispaintroductiodiscussesfthepredictthedatafortheempiricinSection2.LiteratuCashflemployeesestedincaenablethemappreciatiovencydecisandemployissuesrelatinterestediwhentheyCashflotoearningsbringbetteralsosometinetcashinbasedoncrtobesettledtocustomeinformationingtoshortmeasurestdayfinanciworthofaallcashflothecashflowsarepaidoutasdividend.Asaresult,newsaboutcashflowcanpotentiallyhavesignificantimpactonafirm’stpricncludw,Kflowatcugseuntry,d,accble,arecruamakrkinon,proregaifferillinodeandfThiseats.Therecaomeentoin19ent.Tecaowcersiscorgoon-cohenjusteinOdcaed.3%,tudieng’stingdisclono,thehatinentdrerestncremsthaativeiricalstudiessuggestthatthesevariablesareusefulativeinpredictingcashflows.Inthispaper,acompar-ebetweendifferentsetsofvariablesaspredictors.Itthatthemorepredictorsthatareincluded,thebetterllperformbecausemoreinclusionofvariablesoftenerexploitationofinformation.Asecondcomparisontweenmodelswhichincludeexplanatoryvariablesntlags.Itisexpectedthatmodelswithmorelaggedvariablescouldprovidemoreaccuratepredictionreportedresultsinthispapersuggestotherwise.Thisosestwotypesofcashflowmodeling,i.e.movingdelandvectorautoregressive(VAR)modelforcashtion.Movingaveragemodelalsomakeseconomicmeasureshowanunexpectedcashflowshockcouldeoplemakingfutureprediction.Themovingaveragepliedtoone-period-aheadpredictionandVARmodelformulti-period-aheadprediction.Forthispurposerepowerfulandrelieslessondataavailabilitythanssion.Thesemodelsareappliedempiricallyondatafricafirms.perisorganizedasfollows:Section1providesthentothispaper.Section2reviewstheliteratureandactorsthatinfluencethepredictionofcashflowsandionmodelsutilizedinthestudy.Section3describesSouthAfricanfirms.Section4reportstheresultsofalanalysisandtheconclusionofthestudyisprovided5.rereviewowforecastisofinteresttoinvestors,creditors,andratingagenciesamongothers.Investorsareinter-shflowsasinputintotheirinvestmentmodelstotodecideonpayoffrelatingtodividendsandcapitalnoftheirinvestments.Creditorsareinterestedinsol-ionsrelatingtothefirmstheytransactbusinesswitheesareinterestedinjobsecurityandgoing-concerningtofirmstheyworkfor.Ratingagenciesarealsongoing-concernandafirm’sabilitytopayitsdebtsaredue.wcanbeconsideredascomplimentaryinformationsincecombinativeanalysisofbothquantitiesmightresultsthananalyzingearningsonitsown.Earnings,mesreferredtoasnetincome,arethesummationofcomeandnetcreditincome,thelatterofwhichisedittradeswithcustomersandisnotyetbutexpectedbycashinalaterperiod.Theamountofcreditgivenrscouldpotentiallybeoverlookedwithoutcashflowandthismaymisleadinvestorsabouttheriskrelat-ageofcashinthefirm.Inaddition,cashflowdirectlyheoperationalabilityofthefirmtomeetitsday-to-alcommitments.Inconventionalfinancetheory,thefirmistheoreticallyequaltothediscountedvalueofwsgeneratedduringthefirm’slifeassumingthatallmarkeinglyiDechoofcashsionthEarnininaccoinventomodelreceivawhichtermacmodelandwoheavily[BCN]disagghavedandWBCNmtively,result.tiontrrealitymeasunetincstatemissuedstatemfromthcashfltheirpdefinescostofThenoers.Wtheadilarly,methoexaminR2of41989sandZaestimausingTheydInsteaderrortcomporeportetermsahaveiimplieinforme.Alongwithearningsforecast,analystsareincreas-ingcashflowforecastintotheiranalysisandreports.othari,andWatts(1998)[DWK],proposedamodelwhichtheyderivedfromsalesandreachedaconclu-rrentearningsarethebestforecastoffuturecashflow.qualtocashflowplusaccrualsthatincludechangespayable,changesinaccountreceivable,changesinepreciationandamortizationandothers.IntheDKWruals,forsimplicity,includeonlychangesinaccountchangesininventoryandchangesinaccountpayable,equivalenttochangesinworkingcapitalwhilelonglssuchasdepreciationarenotconsidered.TheDKWesseveralstrictassumptionsaboutsalesprocessgcapitalcomponentsandtheirderivedmodelreliesthoseassumptions.Barth,Cram,andNelson(2001)posedamodifiedversionoftheDKWmodel.Theytetheaccrualsintocomponents,anticipatingthemtoentpersistenceinpredictingfuturecashflow.Lorekger(2010)comparedthepredictiveaccuracyofthel,intime-seriesandcross-sectionalanalysisrespec-oundthattime-seriesmodelgeneratesmoreaccurateresultisnotsurprisingsincecross-sectionalestima-allfirmsashomogeneous,whichishardlytrueinDKWandBCNmodelsusetheindirectmethodtoshflow,i.e.theycalculatecashflowcomponentfromandadjusttheresultswithaccrualterms.IntheUSA,fFinancialAccountingStandards(SFAS)No.9588allowedthedisclosureofdirectmethodcashflowherefore,cashflowsafter1988aredirectlyavailableshflowstatement.ChengandHollie(2008)partitionomponentsintocoreandnon-coreones,andanalyzedtenceforfuturecashflowdetermination.Thestudyecashflowcomponentsascashflowsfrom:sales,dssold,andoperatingandadministrativeexpenses.recashflowcomponentsareinterest,taxes,andoth-theseregressorsareappliedinapredictionmodel,dR2isslightlygreaterthantheBCNversion.Sim-rpurtandZang(2009),theissueofwhetherdirectshflowstatementenhancescashflowmodelingisTheirreportedcashflowforecastmodelhadadjustedalthoughtheirsamplewassmaller(comparedtopre-swhenSFASNo.95hadnotbeenpublished).Orpurtpaperexaminedwhetheritmakesadifferenceincashcomponentsusingindirectmethodcomparedtooseditemsdirectlyfromthestatementofcashflow.tdirectlycomparetheaccuracyofforecastmodels.yexaminethestatisticalsignificanceofarticulationsdefinedasthedifferencebetweenestimatedcashsanddisclosedonesintheirregressionmodel.Theirsultssuggestthatthecoefficientsofarticulationerroratisticallysignificantandthusthatarticulationerrorsentalinformationforcashflowforecast.Ithencetthedirectmethodforcashflowsdisclosureismoreinpredictingfuturecashflowthanindirectmethod26Y.Lietal./ReviewofDevelopmentFinance5(2015)24–33statementoSouthAfricwillbebasZang’spapSeveralaccuracyo(2008)andtocashfloaccrualscopositiveacwithhighergestafirmmanagers,(FrancisanbetterprivapredictionDiscretidecreasingetal.(2009agement,imanagemetunisticbeofdiscretiomeansforfore,manipabilityforcmayusedifirm’sfuturdiscretionafuturecashenhanceormotivationWhendiscretionanon-discretmakingtheAnumberdiscretionaandKotharcretionaryonrelevantdifferentactenceandtdependsonfinancialstrealactivitiSuchoperationaryexpThisstunumberofssuggesttheelsewhereaandoperatibasedonssetting.Thforfirms,ithecurrenteviewofcashflowmodelsanddiscussionofmodelsthisstudyhowelati(SACFdenotAPdhesisentiningarecat+1]theirectndsepredticalt+1]�INdprecomisleredicondandrecomp,the=α+++thecashminimenrederecnalentedidinBsuggcuracandNm38.49fcashflow.Thedirectmethodisnotapplicableforanfirms;thereforeinthispapercashdisaggregationedontheindirectmethoddescribedinOrpurtander.researchershaveexaminedfactorsthatinfluencethefcashflowprediction.Forexample,Brochetetal.Ogneva(2012)suggestthataccrualqualityrelateswshocksanddiscretionaryaccrualsmightreducentributiontocashflowprediction.Theyarguethatcrualsand/orhighcashflowvolatilityareassociatedpredictiveabilityofaccruals.Someresearcherssug-s’pastpoorperformancetendstobeimprovedbywhichcouldleadtohigherpredictedfuturecashflowdEason,2012)andthatboldforecastersmighthaveteinformation,sotheycanprovidemoreaccuratethanothers(PaeandYoon,2012).onaryaccrualsmaybeincomeincreasingorincome,dependingonmanagersmotivations.Badertscher)argueindetailabouttheimpactofearningman-ncludingdiscretionaryaccrualsandrealoperationntoncashflowprediction.Theyarguethatoppor-havioronthepartofmanagerscanleadtotheusenaryaccrualstohidethefirm’struesituationasamanagementtomaintaintheirequityvalue.There-ulatedstatementsusedasinputswouldreducetheirashflowprediction.Ontheotherhand,managementscretionaryaccrualstosignaltheirtrueviewofthee.Insuchasituation,amanagement’sstatementsofryaccrualscouldhaveabetterpredictiveabilityofflow.Therefore,whetherdiscretionaryaccrualscanadverselyaffectcashflowforecastdependsonthebehindthemanipulation,ifany,bymanagers.managersaremotivatedtomanipulatenon-ryinformationtosuittheirownselfishends,ionaryinformationcouldberevealingormasked,variablesmoreorlessinformativeinforecasting.ofresearchershavesuggestedmethodstoidentifyryaccruals[seeJones(1991),Dechowetal.(1995)ietal.(2005)amongothers].Apartfromsuchdis-problem,accrualwouldresultinvaryingcoefficientvariables.DechowandGe(2006)showthatundercruallevel,firmscashflowshavedifferentpersis-hecorrelationsbetweencashflowsandaccrualsalsoaccruallevel.AnotheroptionforManagerstodistortatementsinordertomeetreportinggoalsisthroughessuchasthosediscussedinRoychowdhury(2006).tionsincludepricediscounts,reductionofdiscre-endituresandsoon.dymakesanumberofcontributions.First,thescanttudiesoncashflowstudiesindevelopingeconomies,needformorestudiesintheareaastheresultsrenotdirectlytransferableduetodifferentbusinessonalenvironments.Secondly,applicationofmodelstudieselsewheremaynotsuitadevelopingcountryirdly,giventhenatureandimportanceofcashflowsnvestors,creditorsandpolicymakers,amongothers,studyisamajoradditiontotheextantliterature.2.1.RusedinDecbasicrCFt=where�ARdand�parentrepresoperatoffutuEt[CFBaruseddflowaferenttheoreEt[CFwheredefineonthemodelflowpBeyOrpurtnon-cothesec(2008)CFt+1wheresales,andadestpaythreeanon-coadditiostatemthususdefineresultstheacChengandBCR2ofetal.(1998)derivedcashflowforecastfromtheonship:LEt−�ARt)−(Pt−�APt),(1)enotescashflowfromoperation,SALEdenotessales,eschangesinaccountreceivable,Pdenotespurchaseenoteschangesinaccountpayable.Termsinthefirstrepresentscashinflowwhilethesecondparenthesisgcashoutflow.Withseveralassumptionsonfirm’sctivities,DKWmodelsuggeststhattheexpectationshflowsiscurrentearnings:=EARNt,(2)tal.(2001)arguedthatearnings,insteadofbeinglyasapredictor,shouldbedisaggregatedintocashveralaccrualtermsasthecomponentsprovidedif-ictivepowerforfuturecashflowsuggestedbytheirmodelbelow:=CFt+[1−(1−β)γ1γ2(1−π)α−1]�ARt+(1−β)�INVt−�APt,(3)Vdenoteschangesininventoryandothertermsareasviously.ComparedtoDKWmodel,theparametersponentsofearningsareallowedtodiffer.AsBCNssrestrictive,itoutperformstheDKWmodelincashtion(seeBarthetal.,2001).theseaccrualterms,ChengandHollie(2008)andZang(2009)disaggregatecashflowintocoreandomponentsinordertoexaminethepersistenceofonentsincashflowprediction.InChengandHolliemodelisspecifiedas:+β(CSALEt+CCOGt+COEt+CINTtCTAXt+COTHERSt)+β(�ARt+�INVt�APt+DEPRt+AMORTtOTHERSt)+εt+1,(4)termsinthefirstparenthesisarecashreceivedfromfromcostofgoodssold,cashrelatedtooperatingstrativeexpenses,cashusedtopaytax,cashforinter-t,andothercashcomponents,respectively.Thefirstfinedascorecashcomponentsandthelatterthreeareomponents.Thecashflowdisaggregationexploresinformationthatcouldbeobtainedfromcashflowunderdirectmethoddisclosure.ChengandHolliesaggregatedcashcomponentsandalsoaccrualtermsCNmodelaspredictorsforcashflow.TheempiricalestthatsuchcashflowdisaggregationdoesimproveyofBCNmodel,howevertheeffectisminor.InHollie’spaper,model(4)hasareportedR2of39.83%odel(withcashflowun-disaggregated)hasreported%.InU.S.,beforethepublicationofSFASNo.95,Y.Lietal./ReviewofDevelopmentFinance5(2015)24–3327statementofcashflowwasnotcompulsory,sothecashflowfromoperationnstatement.closureofcflowpredicwhichisdestimateduthatareactflow.SomementismoChuang,20setting).Foclosure,sudisaggregaAfricancasandZang(CFt=CFIN=(SALE−INTwherenetminuscashemployeesrespectivelEq.(5)meatively,usininterestanddirectlydisencebetwethesefourshownasthThispaflowpredicModelI:ModelII:AlltheparameterswiththeaModelIIfCashinflowModelIIingatingcashitemssuchpersistenceApartfrpaperalsoexogenousvariables(ARMAX)developedbyWhittle(1951)oxeispalmo=βεdeousparaandetereractowmovemThelean,aslues.le,thlaggtededdatationlyveisprtRtPtVt⎤⎥⎥⎥⎥⎥⎦Aativelethethroldasahet+p�Rt+p�Pt+p�Vt+0istobeheltweeofcaediswitceAtermemoeededtobeestimatedusingbalancesheetandincomeOrpurtandZang(2009)alsosuggestthatdirectdis-ashcomponentsprovidesincrementaleffectoncashtion.Theyintroduceaconceptofarticulationerror,efinedasthedifferencebetweencashcomponentssingbalancesheetandincomestatementandonesuallydisclosed(directmethod)instatementofcashstudiesreportthatthedirectmethodcashflowstate-reinformativeinpredictingcashflow(seeArthurand06andFarshadfarandMonem,2013inAustralianrfirmsthatdonotprovidedirectmethodcashdis-chincrementalbenefitcannotbeexploitedbutcashtioncanstillbedone.Inthispaper,wemodelSouthhflowsbydisaggregatingcashflowfollowingOrpurt2009):t−CFOUTt−INTt−TAXtt−�ARt)−(COGt+OEt+�INVt−�APt)t−TAXt+AEt,(5)operatingcashflowisequaltocashinflow(CFIN)outflow(CFOUTwhichispaidtosuppliersand)andinterestandtaxcashpayment(INTandTAXy).Thefirstandsecondbracketsinthesecondlineofsurestheestimatedcashinflowandoutflowrespec-ginformationinbalancesheetandincomestatement;taxpaymentincasharealsoincluded,whichareclosedindependentofthereportformat.Thediffer-endisclosedactualcashflowfromoperationandcomponentsistermedasarticulationerror(AE),efinaltermintheequation.perstartswithacomparisonoftwoempiricalcashtionmodelsfollowingOrpurtandZang(2009):CFt+1=β0+β1CFt+β2�ARt+β3�INVt+β4�APt+β5DDAt+β6INTt+β7TAXt+εt+1,(6)CFt+1=β0+β1CFINt+β2CFOUTt+β3AEt+β4DDAt+β5�ARt+β6�INVt+β7�APt+β8INTt+β9TAXt+εt+1,(7)variablesarethesameaspreviouslydefined.βsaretobeestimated.ModelIistheempiricalBCNmodeldditionofinterestandtaxpaymentaspredictors.urtherdisaggregatesthecashflowterminmodelI.andcashoutflowareestimatedbasedonEq.(5).principleissuperiortomodelIintwoways:disaggre-flowincorporatesinformationofotheraccountingassalesandallowscashcomponentstohavedifferent.omabovementionedlinearregressionmodels,thisintroduceautoregressivemovingaveragemodelwith(seeBandthgeneraas:CFt+1whereexogenmodelapplyparamInpcashflTheabment.variabdictionlagvaexampsecondforlondemantoricalpredicweappmodel⎡⎢⎢⎢⎢⎢⎣CF�A�A�INwhererespeccapturdictorswecouperiod⎡⎢⎢⎢⎢⎢⎢⎣1C�F�A�A�INwhereaheadIntsonbepowerexaminmodelintrodu(MA)providtal.,2008fordetails).Thisisatimeseriesmodelperattempttoadaptittopaneldataapplication.ThedelARMA(1,1)withexogenousvariablesiswritten0+β1CFt+BXt+β2εt+εt+1,(8)notesthemovingaverage(MA)termsandXdenotesvariables.Themovingaveragetermsdependsonthemeters,thereforetraditionalOLSmethoddoesnotweusemaximumlikelihoodestimators(MLE)forstimation.icalapplication,amulti-period-aheadpredictionofaybeofgreaterimportancethanasingleperiod.entionedmodelsaresimpletoadapttosuchrequire-usualwayistoadjustthelaglengthbetweentargetditspredictivevariables.Forone-period-aheadpre-discussedabove,theexplanatoryvariablestakeoneThus,consideringtwo-period-aheadpredictionforeexplanatoryvariablesshouldtakevaluesoftheir.Thedisadvantageofsuchmethodappearscriticalrmforecastwheredatacoveringalongperiodisbutnotalwaysavailable.Moreover,whenthehis-istoooutdated,theinformationtheycontainforpurposesmaybelimited.Todealwiththisproblem,ctorautoregressive(VAR)model(Sims,1980).VAResentedas(takeoriginalBCNmodelasexample):=A+B⎡⎢⎢⎢⎢⎢⎣CFt−1�ARt−1�APt−1�INVt−1⎤⎥⎥⎥⎥⎥⎦+et,(9)ndeare4×1vectorsofconstantsanderrortermsy,Bisa4×4parametermatrix.ThereforeVARcouldevolvementofnotonlycashflowbutalsootherpre-ughtime.OncetheparametersAandBareestimated,pplytheVARmodeloflagonetoforecastasmanyadasneeded.Theforecastequationiswrittenas:p⎤⎥⎥⎥⎥⎥⎥⎦=[10AB]p⎡⎢⎢⎢⎢⎢⎢⎣1CFt�ARt�APt�INVt⎤⎥⎥⎥⎥⎥⎥⎦,(10)a1×4vectorofzeros;pisthenumberofperiodsforecast.aterempiricalsection,wewillstartfromcompari-nmodelIandmodelIIexaminingtheincrementalshflowdisaggregationincashflowprediction.Alsostheeffectofincludingtwolagsofpredictorsonhinclusionofonlyonelagofpredictors.ThenweRMAXmodel.Withinclusionofmovingaverage,wecouldinthefirstplaceseewhetherthismodelreaccurateforecastandsecondlyshowthedirection28Y.Lietal./ReviewofDevelopmentFinance5(2015)24–33ofcashflowshock’seffectonnextperiodcashflowprediction.Finally,weapplyVARmodeltopredicttwo-year-aheadcashflow,incomonlydepen3.DataThisstuDatastreamareincludeable,sothearegatheregoodssold(SGA),deppayable(Aoperatingcpaidincaspayables,randarticulculatedma(SGA)arevariablesaordertosmnothaveavwholesamtions.TheofobservatablesarelaFormodelsampleconare623firlagperiods4.Empiri4.1.DescrTable1Itisworthassets.Thethesamplebyaveragetoestimateestimatednwiththedindirectlym4.2.LineaInthissmodelsdesmethods.T1995to200TheestimaandthetesareappliedTable1Descriptivestatistics.sllvariaflow,reethefacecaseemoFOUhefoionwipaymtthemre,Eqntmmodel.delIissetasthebenchmarkmodel.ModelIIdisaggre-ashintodifferentcomponents.Theresultsares