Morale & Stock Price/ 2 In their study of 2004 stock market performance: The stock prices of 14 high morale companies increased an average of 16%, ...
Weekly market/sector overview(Performance from 15/06/09 to 19/06/09 )MarketCap.19 Jun 09(mln. local cur.)IndexDJS Telecom 217,81 2,4% 3,8% -7,4% -21,5% -7,2% 199,87 288,23 272.741Telecom Italia ord. 0,94 -2,8% -4,8% -17,9% -25,8% -18,0% ...
STOCK MARKET INDICES. Monthly Average. NOTE: Dots represent last reported daily close. Source: National Sources. June 29, 2009
Related posts:Social Interaction and Stock-Market Participation
High Morale Again Pays Off In Stock Market ...
thesensethatforeachoftheseclusters(sayC)theS&PclusterclosesttoCalsochoosesCasitsclosestcluster.AlloftheremainingclustersarenotnearestneighborsofanyS&Pcluster.Thehighqualityoftheclusteringobtainedusingderivativeshasveryinterestingimplications,sincetheperformanceanal-ysisformostofthetimesseriesdatastructuresassumesthatthesequencesaresmooth1,whichisclearlynotthecaseforthederivatives.Therefore,ourresultssuggestthatnewal-gorithmictechniquesshouldbedeveloped,tocapturethescenariosinwhichnon-smoothtimeseriesdataarepresent.2.SETUPDESCRIPTIONTheData.WehaveusedtheStandardandPoor500index(S&P)historicalstockdatapublishedathttp://kumo.swcp.com/stocks/.Thereareapproximately500stockswhichdailypriceuctuationsarerecordedoverthecourseofoneyear.Eachstockisasequenceofsomelengthd,whered252(thelatternumberisthenumberofdaysin1998whenthestockmarketwasoperational,butdcanbesmallerifthecompanyisremovedfromtheIndex).Weusedonlytheday'sopeningprice;thedataalsocontainstheclosingprice,andthelowandhighstockvaluationfortheday.ThedataalsocontainedtheocialS&Pclusteringinforma-tionwhichgroupsthedierentstocksintoindustrygroupsbasedontheirprimarybusinessfocus.Thisinformationwasalsousedinourexperiments,withtheassumptionthatitprovidesuswithabasisfora\ground-truth"withwhichwecancompareandratetheresultsofourunsupervisedcluster-ingalgorithm.Weabstractedthe102membersofthisS&Pclusteringinto62\superclusters"bycombiningcloselyre-latedonestogether,e.g.,\Automobiles"and\Auto(Parts)"or\Computers(Software)"with\Computers(Hardware)".FeatureSelection.Ourfeatureselectionapproachcon-sistsofthreemainsteps,depictedonthefollowingpicture:Dim reduction- Aggregation- Fourier Transform- PCA- none - first derivativeNormalization - global - piecewise - raw dataRepresentation - noneFigure1:Featureextractionprocess1.Representationchoice:inthisstepwemaptheoriginaltimeseriesintoapointind-dimensionalspace,wheredisclosetothelengthofthesequence.Weusetwotypesofmapping:identityandrstderivative(orFDforshort).Intherstcase,thewholesequenceisconsideredtobeone252-dimensionalpoint.Inthesecondcase,thei-thcoordinateofthederivativevector1E.g.,[1,7]approximateasequencebyremovingallbutfewelementsintheFourierrepresentationofasequence;thequalityofapproximationinthiscasereliesonthefactthathighfrequencycomponentofasignalhavelowamplitude,whichisclearlynotthecaseforthederivativesequence.isequaltothedierencebetweenthe(i+1)-thandi-thvalueofthesequence.Bothmappingsarenaturalinthecontextoftime-seriesdata.2.Normalization:inthisstepwedecideifandhowweshouldnormalizethevectors.Thestandardnormal-izationisdonebycomputingthemeanofthevectorco-ordinatesandsubtractingitfromallcoordinates(notethatinthiswaythemeanbecomesequalto0)andthendividingthevectorbyitsL2norm.Thisstepallowsustobringtogetherstockswhichfollowsimilartrendsbutarevalueddierently,e.g.,duetostocksplits(notethatourtimeseriesarenotadjustedforsplits).Wealsointroduceanovelnormalizationmethodwhichwecallpiecewisenormalization.Theideahereistosplitthesequenceintowindows,andperformnormalization(asdescribedabove)separatelywithineachwindow.Inthiswaywetakeintoaccountlocalsimilarities,asopposedtotheglobalsimilaritycapturedbythenor-malizationofthewholevector.3.Dimensionalityreduction:inthisstepweaimtore-ducethedimensionalityofthevectorspacewhilepre-serving(orperhapsevenimproving)thequalityoftherepresentation.OurrstdimensionalityreductiontechniqueisbasedonthePrincipalComponentAnal-ysis(PCA).PCAmapsvectorsxninad-dimensionalspace(x1;:::;xd)ontovectorszninanM-dimensionalspace,whereM<d.PCAndsdorthonormalba-sisvectorsui,calledalsoprincipalcomponents,andretainsonlyasubsetM<doftheseprincipalcom-ponentstorepresenttheprojectionsofvectorsxnintothelower-dimensionalspace.PCAexploitsthetech-niqueofSingularValueDecomposition(SVD),whichndstheeigenvaluesandeigenvectorsofthecovari-ancematrix=Xn(xnx)(xnx)Twherexisthemeanofallvectorsxn.Theprincipalcomponentsareshowntobetheeigenvectorscorre-spondingtotheMlargesteigenvaluesofandtheinputvectorsareprojectedontotheeigenvectorstogivethecomponentsofthetransformedvectorsznintheM-dimensionalspace.Oursecondtechnique,aggregation,isbasedontheassumptionthatlocaluctuationofthestock(say,withintheperiodof10days)isnotasimportantasitsglobalbehavior,andthereforethatwecanreplacea10dayperiodbytheaveragestockpriceduringthattime.Inparticular,wesplitthetimedomainintowindowsoflengthB(forB=5;10;20etc)andreplaceeachwindowbyitsaveragevalue.Clearly,thisdecreasesthedimensionalitybyafactorofB.OurthirdtechniqueisbasedontheFourierTransform(e.g.,see[1]forthedescription).Basically,weusedtruncatedspectralrepresentations,i.e.,werepresentedatime-seriesbyonlyafewofitslowestfrequencies.Untilnowwedescribedhowwecomparesequencesofiden-ticallength.Inordertocompareapairofsequencesofdierentlengths,wetakeonlytherelevantportionofthe
Related posts:TRUSTING ...
1. Introduction Today's new era of corporate governance requires higher levels of information disclosure and data integrity due to regulations such as the Sarbanes-Oxley (SOX), the Gramm-Leach- Bliley Act ...
EUGENE F. FAMA is the Theodore 0. Yntema Pro-fessor of Finance at the Graduate School of Busi-ness of the University of Chicago. His researchinterests encompass the broad areas of economics,finance, ...
where nobody else invests in the stock market (and it does not interact with ... stock market participation, and that forms the basis for our subsequent ...
Related posts:STOCK MARKET INDICES
TRUSTING ...
1IntroductionMostsocialsystemsinvolvecomplexinteractionsamongmanyindividuals.Muchofeconomicshopesbothtogreatlysimplifyhumanbehavior,andtodoitinsuchawaythataggregatemacrofeaturescanbeeasilycharacterized.Thesuccessofthistraditionalapproachtodescribehumanbehaviorhasonlyhadmixedsuccess.Oneareawherequestionsremainisnancewheremanyempiricalfeaturesaretroublingforexistingtheories.Severalofthefoundationsoftheeldareinastateofdisarray,andnew,radicallydifferenttheoriesareappearing.1Oneofthedirectionsthatresearchershavebeentakingistheuseofagent-basednancialmarkets.Thesebottom-upmodelsofnancialmarketsstartfromrstprincipalsofagentbehavior.Usingeithercomputational,ormoresophisticatedmathematicaltoolstheyareabletodescribemacrofeaturesemergingfromasoupofindividualinteract-ingstrategies.Sincetheearly1990'sI'vebeeninvolvedwithoneoftherstagent-basednancialmarketplatforms,TheSantaFeArticialStockMarket.Now,withnearlyadecadeofexperienceinlookingatnancialmarketsfromanagent-basedperspective,IwouldliketoturnmyattentionbacktotheSFImarket.Iwillexploresomeofthemarket'searlyhistory,andthedebatesanddesignquestionsthatwentintoitsdevelopment.FromamodernperspectivethereisstillmuchtolikeaboutwhattheSFImarketwastryingtodo.However,therearealsothingsthatIbelievemakeitalessthanperfecttoolformodernresearchonnancialmarkets.Muchofmymarketdesignphilosophystemsfromadesiretounderstandtheimpactofagentinteractionsandgrouplearningdynamicsinanancialsetting.Whileagent-basedmarketshavemanygoals,Iseetheirrstscienticuseasatoolforunderstandingthedynamicsinrelativelytraditionaleconomicmodels.Itisthesemodelsforwhicheconomistsofteninvoketheheroicassumptionofconvergencetorationalexpectationsequilibriumwhereagents'beliefsandbehaviorhaveconvergedtoaself-consistentworldview.Obviously,thiswouldbeaniceplacetogetto,butthedynamicsofthisjourneyarerarelyspelledout.Giventhatnancialmarketsappeartothriveondiverseopinionsandbehavior,arstleveltestofrationalexpectationsfromaheterogeneouslearningperspectivewasalwaysneeded.2ForthisreasonIhaveoftenusedtraditionaleconomicmodelswhichhavewelldenedhomogeneousequilibriaasthecoreformodelbuilding.Thiscertainlyisn'tnecessaryforanagent-basedapproach,butIthinkatthisstageadeeperunderstandingofdynamicsaroundwellstudiedequilibriaisessential.Theexistenceofanequilibriumbenchmarkalsoprovidesanimportanttestcaseformanyofthecomputationallearningtoolsinuse.Ifforsomecasestheycanactuallylearnconvergencetoanequilibriumthentheirabilitiestoperformpurposefulsearchthroughthesetsofrulesaregive1SeeHeaton&Korajczyk(2002)alongwitharticlesinthatvolumeforanintroductiontobehavioralnanceissues.AverynicesummaryoftheeldisShefrin(2000).MuchofthisworkisstillsubjecttocontroversyasshownbyRubenstein(2001).2SeeforSargent(1993)forfurtherthoughtsontheimportanceoflearningrationalexpectations.Also,Board(1994)providessomeimportantincitesintothetheoreticalcomputabilityofequilibria.1
Related posts:TRUSTING THE STOCK MARKET
STOCK MARKET INDICES
Who Gambles In The Stock Market?
Examination System in the Stock Market” which is scheduled on 12-13 January ... Stock market in Indonesia had achieved high performance in
Related posts:Social Interaction and Stock-Market Participation
TRUSTING THE STOCK MARKET
High ...
The decision to invest in stocks requires not only an assessment of the risk-return trade-ogiven the existing data, but also an act of faith (trust) thatthe data in our possession ...
1FINANCIAL LITERACY WORK READINESS ENTREPRENEURSHIP IntroductionOn October 9, 2007, the Dow Jones Industrial Index, a popular measure of the overall ...
WORKING PAPER SERIESNO 862 / FEBRUARY 2008In 2008 all ECBpublicationsfeature a motiftaken from the10 banknote.STOCK MARKET VOLATILITY ANDLEARNING1Klaus Adam2,Albert Marcet3and Juan Pablo Nicolini4This paper can be downloaded without charge fromhttp://www.ecb.europa.eu ...
