University of Florida Digital Collections Home
UFDC Home myUFDC Home  |   Help 
<%BANNER%>

Information Asymmetry in Direction and Volatility

Permanent Link: http://ufdc.ufl.edu/UFE0024812/00001

Material Information

Title: Information Asymmetry in Direction and Volatility Price Process and Transaction Level Analysis
Physical Description: 1 online resource (197 p.)
Language: english
Creator: Yoon, Joon
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: econometrics, market, markov, stochastic
Industrial and Systems Engineering -- Dissertations, Academic -- UF
Genre: Industrial and Systems Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation develops a structural methodology for equity pricing in a semi-strong efficient market and performs an empirical study supporting the need for such a methodology. Conventional price dynamics of equities and their related options assume markets to be strongly efficient. This hypothesis however is a severe simplification ignoring liquidity risk, adverse selection effect, and the difficulty of achieving informational efficiency. The methodology proposed in this thesis categorizes equity price dynamics into 3 sub-processes: fundamental value, duration, and National Best Bid and Offer (NBBO) revisions. This approach reveals how an informed trader's knowledge is eventually incorporated into equity and option prices. Market-makers are assumed to apply Bayesian Nash-equilibrium strategies to construct rational NBBO quotes to hedge against adverse selection risk as well as to provide quotes attractive to uninformed traders. The objective of the empirical study is to find evidence of market inefficiency in both fundamental value and intra-day trades. For the former, we focus on structural breaks in direction and volatility of equity prices when new key developments are announced. Conventional event-studies only classify three sub-event windows: pre-event, event, and post-event. We propose a change-point model that allows multiple regimes in association with pre-determined key events, with variable regime lengths depending on structural breaks in direction and volatility associated with abnormal returns. Our study confirms that adjustments to key developments may begin before and end after their announcements in several stages. Furthermore, the correlation between successive key developments is found to be statistically negligible, thus supporting the uniqueness of each. For the second issue, we examine intra-day trades and NBBO data upon the announcements of new key developments. Through various statistical models, we associate with trading and NBBO two distinct sets of decision factors, depending on whether we are in a stable or a transition regime. Estimates of trade duration, NBBO duration, and realized volatility are the most distinguishable between stable and transition regimes. Overall, $R^2$s are lower for transition regimes, showing that price evolution depends more on exogenous factors than in stable regimes. Coefficient estimates differ very little between the two types of regimes, with a few exceptions. Factors of traded or NBBO exchange have limited significance, with a few exceptions. A comparison between actual empirical data and simulated results, based on uninformed limit-order traders and market-makers following Bayesian Nash-equilibrium strategies, shows that the latter are not necessarily better, except that they lead to a less volatile price discovery.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Joon Yoon.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: AitSahlia, Farid.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024812:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024812/00001

Material Information

Title: Information Asymmetry in Direction and Volatility Price Process and Transaction Level Analysis
Physical Description: 1 online resource (197 p.)
Language: english
Creator: Yoon, Joon
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: econometrics, market, markov, stochastic
Industrial and Systems Engineering -- Dissertations, Academic -- UF
Genre: Industrial and Systems Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation develops a structural methodology for equity pricing in a semi-strong efficient market and performs an empirical study supporting the need for such a methodology. Conventional price dynamics of equities and their related options assume markets to be strongly efficient. This hypothesis however is a severe simplification ignoring liquidity risk, adverse selection effect, and the difficulty of achieving informational efficiency. The methodology proposed in this thesis categorizes equity price dynamics into 3 sub-processes: fundamental value, duration, and National Best Bid and Offer (NBBO) revisions. This approach reveals how an informed trader's knowledge is eventually incorporated into equity and option prices. Market-makers are assumed to apply Bayesian Nash-equilibrium strategies to construct rational NBBO quotes to hedge against adverse selection risk as well as to provide quotes attractive to uninformed traders. The objective of the empirical study is to find evidence of market inefficiency in both fundamental value and intra-day trades. For the former, we focus on structural breaks in direction and volatility of equity prices when new key developments are announced. Conventional event-studies only classify three sub-event windows: pre-event, event, and post-event. We propose a change-point model that allows multiple regimes in association with pre-determined key events, with variable regime lengths depending on structural breaks in direction and volatility associated with abnormal returns. Our study confirms that adjustments to key developments may begin before and end after their announcements in several stages. Furthermore, the correlation between successive key developments is found to be statistically negligible, thus supporting the uniqueness of each. For the second issue, we examine intra-day trades and NBBO data upon the announcements of new key developments. Through various statistical models, we associate with trading and NBBO two distinct sets of decision factors, depending on whether we are in a stable or a transition regime. Estimates of trade duration, NBBO duration, and realized volatility are the most distinguishable between stable and transition regimes. Overall, $R^2$s are lower for transition regimes, showing that price evolution depends more on exogenous factors than in stable regimes. Coefficient estimates differ very little between the two types of regimes, with a few exceptions. Factors of traded or NBBO exchange have limited significance, with a few exceptions. A comparison between actual empirical data and simulated results, based on uninformed limit-order traders and market-makers following Bayesian Nash-equilibrium strategies, shows that the latter are not necessarily better, except that they lead to a less volatile price discovery.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Joon Yoon.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: AitSahlia, Farid.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024812:00001


This item has the following downloads:


Full Text

PAGE 1

1

PAGE 2

2

PAGE 3

3

PAGE 4

IamdeeplyappreciatedtoallthehelpIhavereceivedinresearch.Mydeepestgratitudeistomyadvisor,Dr.FaridAitSahlia.Hegavemenotonlythetremendousfreedomtoexploreonmyownbutalsotheconstructivecriticismwhenmyresearchwentpointlessly.Hetaughtmehowexpressideasandexaminethefeasibility.Myco-advisors,Dr.Nimalendran,Dr.Karceski,Dr.Ghosh,andDr.Uryasevhavebeeninspiringmewithnewideas.WithouttheirgiantstepsontheirresearchareaIcouldnothavecouragetodigintosuchintricateelds. 4

PAGE 5

page ACKNOWLEDGMENTS ................................. 4 LISTOFTABLES ..................................... 7 LISTOFFIGURES .................................... 9 ABSTRACT ........................................ 11 CHAPTER 1INTRODUCTION .................................. 13 1.1FundamentalProblem ............................. 13 1.2AsymmetricInformationinEquityandOption ................ 15 1.3LiquidityMeasuresofEquityandOption ................... 20 1.4AnalysisonFundamentalValue:EventStudy ................ 25 1.5AnalysisofIntradayPriceProcesses ...................... 27 1.6Contributions .................................. 28 2EVENTSTUDY:ACHANGE-POINTMODELAPPROACH .......... 29 2.1Methodology .................................. 29 2.1.1AbnormalReturn ............................ 29 2.1.2EstimationofStructuralBreakPoints ................. 30 2.1.3EstimationofDirectionandVolatilityChanges ............ 33 2.1.4ModelSelectionviaBayesFactor ................... 35 2.2EmpiricalAnalysis ............................... 36 2.2.1Data ................................... 36 2.2.2AbnormalReturns ............................ 38 2.2.3KeyDevelopmentAnnouncements ................... 40 2.2.4EstimatingStructuralBreakPoints,Direction,andVolatility .... 42 2.2.5SelectingtheBestChange-PointModel ................ 44 2.2.6EstimatingtheLengthofEachRegime ................ 47 2.2.7StatisticalPropertiesinStructuralBreaks .............. 47 3ANALYSISOFINTRADAYPRICEPROCESS .................. 57 3.1AModelofEquityPriceDiscovery ...................... 57 3.1.1EvolutionofFundamentalValue .................... 57 3.1.2DurationsofTradesandNationalBestBidandOerRevisions ... 60 3.1.3EvolutionofBidandAskPrices .................... 64 3.1.4PerformanceMeasures ......................... 67 3.2EmpiricalAnalysis ............................... 69 3.2.1Data ................................... 69 3.2.2TheEmpiricalSpecication ...................... 84 5

PAGE 6

.............................. 89 3.2.4NBBOProcess .............................. 94 3.3SimulationResults ............................... 113 4CONCLUSION .................................... 120 4.1LimitationandFutureResearch ........................ 120 4.2Summary .................................... 122 APPENDIX ADERIVATIONOFEQUATIONS .......................... 124 A.1DerivationofEquation( 3{18 ) ......................... 124 A.2DerivationofEquation( 3{19 ) ......................... 125 A.3DerivationofEquation( 3{21 ),( 3{22 ) ..................... 126 BSUMMARYOFEXCHANGEFACTORS ..................... 127 B.1TradeSummaryofExchangeFactors ..................... 127 B.2NBBOSummaryofExchangeFactors ..................... 134 CFULLTABLESOFCHAPTER 2 .......................... 143 REFERENCES ....................................... 193 BIOGRAPHICALSKETCH ................................ 196 6

PAGE 7

Table page 1-1LiquidityDimensionsandTheirComponentsinEquityandOption ....... 24 2-1Direction,Volatility,andAR(1)TestforAbnormalReturnsbyIndustry .... 39 2-2Direction,Volatility,andAR(1)TestforAbnormalReturnsbyMarketCapitalizationandBook-to-MarketRatio .............................. 41 2-3AR(1)Test,StationarityTest,andDistributionFittingforInterarrivalTimesofKeyDevelopments ................................. 43 2-4SelectionoftheBestChange-PointModel:Example ................ 46 2-5TheBestChange-PointModels ........................... 46 2-6TheBestChange-PointModelforAllFirms .................... 48 2-7LengthsofRegimesintheBestChange-PointModels ............... 49 2-8DirectionandVolatilityDierencesinConsecutiveRegimes ........... 50 2-9CorrelationandAutocorrelationofDirectionsandVolatilities .......... 52 3-1TradeSummarybyExchange ............................ 71 3-2TradeSummarybyTradeTimeInterval ...................... 75 3-3TradeSummarybyIndustrySectors ........................ 77 3-4NBBOSummarybyAskandBidExchanges .................... 80 3-5NBBOSummarybyTradeTimeInterval ...................... 84 3-6NBBOSummarybyIndustrySector ........................ 85 3-8SummaryofRegressions ............................... 92 3-9SummaryofRegressions ............................... 103 3-10SummaryofSimulationResults ........................... 115 3-11SummaryofPerformanceofPricingStrategies ................... 119 B-1SummaryofExchangeFactors ............................ 127 B-2SummaryofExchangeFactors ............................ 134 C-1BestChange-pointModelsonAbnormalReturns ................. 144 C-2LengthsofRegimesinEachChange-PointModel ................. 152 7

PAGE 8

................... 159 C-4ConsecutiveDierencesinMLEofVolatilities ................... 165 C-5CorrelationandAutocorrelationofDirectionsandVolatilities .......... 170 8

PAGE 9

Figure page 1-1ASchematicModelofSemi-strongEcientMarket ................ 16 2-1TransitionDiagramofRegimeChange ....................... 30 2-2PosteriorProbabilitiesofThree-RegimeModelinanEventPeriod ........ 44 2-3MLEConvergenceofsandsinMCEM ..................... 45 3-1SamplePathsofFundamentalValue(left)and(ontheright)AssociatedDirection,Volatility,andPriceJumpsatKeyDevelopments ................. 59 3-2ARealizationofNBBOrevision ........................... 64 3-3RealizationsofNBBOPriceEvolutions ....................... 67 3-4EvolutionofMeanandStandardDeviationofTradeDurationinEventWindows 87 3-5EvolutionofMeanandStandardDeviationofRealizedVolatilityinEventWindows 87 3-6EvolutionofMeanandStandardDeviationofBid-AskSpreadinEventWindows 88 3-7EvolutionofMeanandStandardDeviationofBid-AskSpreadAfterTradeinEventWindows .................................... 88 3-8EvolutionofMeanandStandardDeviationofLog-ScalePriceinEventWindows 88 3-9EvolutionofMeanandStandardDeviationofTradeSizeinEventWindows .. 89 3-10EvolutionofMeanandStandardDeviationofR2inModellog(ITi) ....... 91 3-11EvolutionofMeanandStandardDeviationofR2inModelDi 92 3-12EvolutionofMeanandStandardDeviationofR2inModelA+iPi 92 3-13EvolutionofMeanandStandardDeviationofR2inModelPiB+i 93 3-14EvolutionofMeanandStandardDeviationofR2inModelPi 93 3-15EvolutionofMeanandStandardDeviationofR2inModelSTi 93 3-16EvolutionofRegressorsinModellog(ITi) ..................... 95 3-17EvolutionofRegressorsinModelDi 96 3-18EvolutionofRegressorsinModelA+iPi 97 3-19EvolutionofRegressorsinModelPiB+i 98 3-20EvolutionofRegressorsinModelPi 99 9

PAGE 10

100 3-22EvolutionofMeanandStandardDeviationofR2inModellog(INj) ....... 103 3-23EvolutionofMeanandStandardDeviationofR2inModelNDj 104 3-24EvolutionofMeanandStandardDeviationofR2inModelAjBj 104 3-25EvolutionofMeanandStandardDeviationofR2inModelBj 104 3-26EvolutionofMeanandStandardDeviationofR2inModelAj 104 3-27EvolutionofMeanandStandardDeviationofR2inModelSNBj 105 3-28EvolutionofMeanandStandardDeviationofR2inModelSNAj 105 3-29EvolutionofRegressorsinModellog(INj) ..................... 106 3-30EvolutionofRegressorsinModelNDj 107 3-31EvolutionofRegressorsinModelAjBj 108 3-32EvolutionofRegressorsinModelBj 109 3-33EvolutionofRegressorsinModelAj 110 3-34EvolutionofRegressorsinModelSNBj 111 3-35EvolutionofRegressorsinModelSNAj 112 10

PAGE 11

Thisdissertationdevelopsastructuralmethodologyforequitypricinginasemi-strongecientmarketandperformsanempiricalstudysupportingtheneedforsuchamethodology.Conventionalpricedynamicsofequitiesandtheirrelatedoptionsassumemarketstobestronglyecient.Thishypothesishoweverisaseveresimplicationignoringliquidityrisk,adverseselectioneect,andthedicultyofachievinginformationaleciency.Themethodologyproposedinthisthesiscategorizesequitypricedynamicsinto3sub-processes:fundamentalvalue,duration,andNationalBestBidandOer(NBBO)revisions.Thisapproachrevealshowaninformedtrader'sknowledgeiseventuallyincorporatedintoequityandoptionprices.Market-makersareassumedtoapplyBayesianNash-equilibriumstrategiestoconstructrationalNBBOquotestohedgeagainstadverseselectionriskaswellastoprovidequotesattractivetouninformedtraders. Theobjectiveoftheempiricalstudyistondevidenceofmarketineciencyinbothfundamentalvalueandintra-daytrades.Fortheformer,wefocusonstructuralbreaksindirectionandvolatilityofequitypriceswhennewkeydevelopmentsareannounced.Conventionalevent-studiesonlyclassifythreesub-eventwindows:pre-event,event,andpost-event.Weproposeachange-pointmodelthatallowsmultipleregimesinassociationwithpre-determinedkeyevents,withvariableregimelengthsdependingonstructuralbreaksindirectionandvolatilityassociatedwithabnormalreturns.Ourstudyconrmsthatadjustmentstokeydevelopmentsmaybeginbeforeandendafter 11

PAGE 12

Forthesecondissue,weexamineintra-daytradesandNBBOdataupontheannouncementsofnewkeydevelopments.Throughvariousstatisticalmodels,weassociatewithtradingandNBBOtwodistinctsetsofdecisionfactors,dependingonwhetherweareinastableoratransitionregime.Estimatesoftradeduration,NBBOduration,andrealizedvolatilityarethemostdistinguishablebetweenstableandtransitionregimes.Overall,R2sarelowerfortransitionregimes,showingthatpriceevolutiondependsmoreonexogenousfactorsthaninstableregimes.Coecientestimatesdierverylittlebetweenthetwotypesofregimes,withafewexceptions.FactorsoftradedorNBBOexchangehavelimitedsignicance,withafewexceptions. Acomparisonbetweenactualempiricaldataandsimulatedresults,basedonuninformedlimit-ordertradersandmarket-makersfollowingBayesianNash-equilibriumstrategies,showsthatthelatterarenotnecessarilybetter,exceptthattheyleadtoalessvolatilepricediscovery. 12

PAGE 13

Thischapterpresentsathoroughintroductionofthisdissertation.Section1.1denesandsetskeyassumptionsofthemarket,participants,andtheadverseselectioneect,andthenraisesthefundamentalproblem.Section1.2consistsofaliteraturereviewonrationalequityandoptionpricinginasemi-strongmarket,highlightingthelackofconsensusregardingoption-relatedadverseselection.Section1.3reviewstheliteratureonmarketmicrostructurerelatedtoinformedtradinginequitiesandoptionsandshowshowliquiditymeasurescaneectivelyrevealadverseselection.Section1.4introducesanalternativepricediscoveryprocessthatincorporatesliquiditymeasures,whichwillbeextensivelydevelopedinthenextchapter.Section1.5givesabriefoverviewoftheempiricalstudy,withSection1.6summarizingitscontributions. FamaandFrench ( 1993 ).Wealsoassumethatallinformedtradersplacemarket-orders,sincetheiradvantagedependsontiming,andalllimit-ordertradersareuninformed,withalltradesoccurringatmarketprices.Hencemarket-ordertraderscanbeinformedoruninformed,butalllimit-ordertraders,includingmarketmakers,areuninformed.Therearemanymarket-makersandpublic(uninformed)limit-ordertraders,whooerliquiditythroughholdinginventorybyimposingbid-askspreadtomarketordertraders.Tosimplifythemodel,weassumeanimaginarymonopolisticmarket-makerwhoalwaysoersequilibrium 13

PAGE 14

Firm-orindustry-specicinformationcumulativelyresetarm'sfundamentalvalueupontheoccurrenceofeachkeydevelopment,suchasanearningsannouncement,amergeroracquisition,anewresearchdiscovery,theappearanceofanewcompetitor,downsizingandsoon.Informationitselfisuncertainuntilitiseective.Thusitsimpactonexpectedreturns(direction)andstandarddeviation(volatility)makestradersmoreactivetoadjustthecurrentequityandoptionprices.Whenkeyeventshappen,directionalorvolatilityinformationisknownrstonlytosomeinformedtraders,thenismadepublic,followedbyitsreectionintheequityprice.Whetherknownonlytoinformedtradersormadetopublic,informationmaystillbeuncertain,butitsincorporationintotheequitypriceindicatesalesseningofthisuncertainty. Aninformedtradermaybenetfromheradvancedknowledgeinoneoftwoways.Ifshegetspositivedirectionalinformation,shetakesalongequitypositionandtakesabullishoptionstrategy,whichisgoinglongonthecallandshortontheput.Withnegativedirectionalinformation,shetakesashortequitypositionandabearishstrategy,goingshortonthecallandlongontheput.Ifshehaspriorknowledgeofaforthcominghighvolatility,shetakesalongstraddleposition,goinglongoncallandput.Ontheotherhand,herpriorknowledgeofaforthcominglowvolatilityleadshertoashortstraddleposition,shortonthecallandput.Oncetheinformationispubliclyavailable,bothinformedanduninformedtradershaverealizeddirectionalandvolatilityinformation.For

PAGE 15

Easley,O'Hara,andSrinivas ( 1998 ),thatisonewhotradesforliquidity-basedreasons,hencebuyingandsellingequityandoptionsatthesamerate,irrespectiveoftheinformationalcontentinthemarket.Amarket-makerrevisesbid-askquotesonthebasisofpubliclyavailableinformationandwhenmarketordersarrive.Weassumethereisnoherdingorblung:neitheruninformedtraderisherdedbyanymarkettrend,nordoesaninformedtrader\blu"uninformedtraderstoconcealinformation.Figure 1-1 isaschematicmodelillustratinghowdirectionalandvolatilityinformationowbetweenequityandoption,andhowamarket-makermeasuresprivateinformationandincorporatesitintoequityandoptionprices.2Amarket-makerfacesanadverseselectionproblemfromdirectional-andvolatility-informedtradersfromequityandoptiontrade.Sinceanoptionisaderivativesecurity,informationincorporatedintheunderlyingequityisreectedintheoptionpriceaswell.Inasemi-strongecientmarket,aninformedtradercantradeoptionstotakeadvantageofequity-relatedpriorinformation.Therefore,themainproblemiswhetheramarketmakercansetrationalpricesandspreadsofequityandoptionunderadverseselectionriskfromdirectional-andvolatility-informedtraders.Thesignicanceofthisproblemishighlightedinthereviewsectionbelow. Easley,O'Hara,andSrinivas ( 1998 )and CherianJ.A. ( 1998 ).

PAGE 16

ASchematicModelofSemi-strongEcientMarket 16

PAGE 17

Duetotheunprecedentedgrowthinoptionsmarketsandgiventheconsensusthatnancialmarketsarenotecientbecauseofinformationasymmetryamongtraders,agrowingnumberofresearchershavebeeninterestedintheconsequencesofinformationtradinginequityandoptionmarkets.However,theoreticalandempiricalstudiesaresplitonwhetherornotaninformedtraderwouldusethisadvantagetotradeoptions.OneschoolofthoughtsuggestsasocalledLeverageHypothesisandclaimsthataninformedtraderwouldfavoroptiontradingbecauseofanoption'shighleverage,itslowtransactioncosts,lessstringentmarginconditions,absenceofuptickruleforshorting,built-indownsideprotection,andtheopportunitytobetonvolatility.3Empiricalevidenceshowsthatoptionvolumecontainsinformationaboutfuturestockprices,4optionpricescontributeanaverageof17%totheequitypricediscovery5.optionvolumeisrelatedtopublicannouncement,6andoptionstrategieswithVegaexposureconveyinformationaboutfuturevolatility.7Evidenceofinformedtradinginoptionwasfoundinexaminingliquidityofoptionmoneyness.Researchersshowedthatanout-of-the-money(OTM)optionoersagreatleverageadvantage.Anat-the-money(ATM)optionprovidesvolatility Black ( 1975 ), Chakravarty,Gulen,andMayhew ( 2004 ), CherianJ.A. ( 1998 ), Back ( 1993 ),Busi-nesssnapshot10.2,p235of Hull ( 2000 ).4 ( 1998 )suggestedmultimarketsequentialtrademodel,Iextendedthemodeltoincorporatetradingstrategiesandtolinktospreaddecompositionmodelfrom Madhavan,Richardson,andRoomans ( 1997a ). Chen,Lung,andTay ( 2005 )usedoptiontradingvalueratiotoex-tractinformationinoptiontrading.5 ( 2004 )appliedHasbrouck'sinformationsharingapproach.6 ( 2006 )focusedonlinkingproxiesbetweeninformedtradingandnews.7 ( 2006 )focusedonoptionstrategiesthatarepackagedforvolatilitytrading.

PAGE 18

Empiricalevidenceshowsthatoptionbid-askspreadspositivelydependondelta-hedgingcostsandthebid-askspreadsofunderlyingequity.Furthermore,optiontradingvolumereducesbid-askspreads,whichisoppositethebid-askspreadofequity,andoptionliquidityhasaverysmalladverseselectioncomponent,becausetheoptionmarketusuallyhasbigdepth.Finally,optiontradingstrategieswithdeltaexposuredonotappeartocontaininformationaboutfuturereturns,whichimpliesoptionorderowanddirectionaltradingarenotquiterelated.11Evidencefrombothschoolsofthoughtrevealsthattheoptionmarketisavenueforinformedtrading.As Back ( 1993 )mentioned,theexistenceofanoptionimpliesthatricherclassofinformationsignalscanbereceived,howeverinformedtradingofoptionsdoesnothaveaclearadvantagefromanadverseselectionperspective. ( 2004 ), Chen,Lung,andTay ( 2005 ), Blasco,Corredor,andSanta-maria ( 2006 )givealmostthesamendings.9See Vijh ( 1990 ).10See ChoandEngle ( 1999 ).11See Vijh ( 1990 ), ChoandEngle ( 1999 ), Landsiedl ( 2005 ),and RudigerFahlenbrach ( 2006 ).

PAGE 19

Back ( 1993 )and Easley,O'Hara,andSrinivas ( 1998 ).13See BiaisandHillion ( 1994 )14See ChoandEngle ( 1999 )15See Dumas,Fleming,andWhaley ( 1996 )16See Ross ( 1976 )

PAGE 20

From Bagehot ( 1971 )'sessay,aresearcheldcalledmarketmicrostructure,whichexaminestheeectsofprivateinformationontheliquidityofanancialsecurity,hasthrived.Inanancialmarket,aprimarycauseofilliquidity,thecostofimmediacy,isadverseselection,whicharisesfromthepresenceofprivatelyinformedtraders.20Bagehotdescribedhowliquidityinamarketisinverselyrelatedtothespread.Thesmallestspreadamarketmakercanmaintainandstillsurviveisinverselyrelatedtoaverageowrateofnewinformationaectingthevalueoftheasset,andisdirectlyrelatedtothevolumeofliquidity-motivatedtransactions.Thereareinfactthreedimensionsofliquiditytocaptureadverseselection:(i)Tightness(orinverselybreadth)iscostofturningaroundatrade BiaisandHillion ( 1994 )18 ( 1998 )explainsself-fulllingprophecy.19See BiaisandHillion ( 1994 ).20See BrennanandSubrahmanyam ( 1996 ).

PAGE 21

GlostenandMilgrom ( 1985 )showhowbreath,orbid-askspread,becomeswiderastheadverseselectioneectishigher.Intheirsequentialtradingmodel,throughBayesianNashEquilibriumpricing,amarketmakerlearnsprivateinformationthroughtheowofbuy-initiatedorsell-initiatedtrades,andthentransfershislosstoanuninformedtraderthroughawidenedspread.Besidesadverseselectioncosts,thespreadincludesoperationandinventorycoststooerimmediacy.Howeverthesearetransitoryandareusuallyverysmallinacompetitivemarket. Inhisanalysisofacompetitiveauctionmarket, Kyle ( 1985 )showedhowmarketdepthisdeterminedbyinformedtrading,andhowaninformedtradersetsherstrategyaccordingly.Amarketmakergathersallordersforanauctionperiod,andsetarationalexpectationequilibriumprice.Thelatterisapricethatmaximizestheinformedtrader'sprotwhilethemarketmakerstaysrisk-neutralbytransferringlosstotheuninformedtrader.Sinceamarketmakercannotdistinguishwhethereachorderowcomesfromaninformedtraderoranuninformedtrader,hesetspricesproportionaltotheimbalanceoforders,whichindicatesthedegreeofinformedtrading.Theoptimalslopeoforderimbalance,whichshowshowmuchpriceschangetothedegreeoforderimbalance,iscalledthepriceimpact,inverselyrelatedtodepth.Ontheotherhand,aninformedtraderknowsthatthereisalimittotradeinindividualauctions,becausethepricewillgouptothetruevalueofherinformationifshetradesuptoalimit.Henceshestrategically\hides"in Black ( 1971 ),andorganizedby Kyle ( 1985 ).

PAGE 22

Thelastliquiditydimensionispricereversal,orresiliency.Accordingto Llorente,Michaely,Saar,andWang ( 2001 ),ifinformedtradingisrelativelyinsignicant,returnsaccompaniedbyhighvolumeshavetendencytoreversethemselvesinsubsequentperiods,becauseanecientmarketpullspricesbacktoequilibriumpricebytraderswhobelievethatcurrentpricesaretoofarfromthetrueprice.However,ifinformedtradingissignicant,asexpressedthroughhightradevolume,thenreturnsarelesslikelytoreverseandcancontinueinsubsequentperiods. Llorente,Michaely,Saar,andWang ( 2001 )suggeststhatpricereversalexpectationisnegativelyproportionaltopastreturnsandcurrentvolume,togetherwithuncertaintyofprice,dividends,andprivateinformation. Inadditiontoequityliquiditydimensionsandtheiradverseselectioneect,examinationofoptionliquiditydimensionsandtheiradverseselectioneecthasbecomeanimportantissueonthebasisofempiricalevidenceregardingincompletemarkets.Optionliquidityhasthreemorecharacteristicsthanequityliquidity.First,ithasmoneynessandtimetomaturity.Thesetwocontingenciestopriceandtimerangeallowtraderstochoosetobeexposedtoortobeprotectedfromriskimposedinthoseranges.Howevertheselimittheliquidityoftheoption,whichmakesitdicultfortheoptionmarketmakertondrationalprices.Second,byitsderivativenature,anoptionliquiditydependsontheliquidityofunderlyingequityandotherenvironmentalvariables.Theoptionprice'ssensitivitiesonthosevariablesarerepresentedbyGreekletterssuchasDelta,Gamma,Vega,Theta,andRhoinchapter15of Hull ( 2000 ).Hencetheadverseselectionimpactonoptionliquidityistransferredfromtheunderlyingequity,buttheoptionliquiditymayhaveitsownadverseselectioncomponenttransferringtotheunderlying.Sinceoptionisonlyavolatilityvehicle,theimpliedvolatilityforecastillustrateshowoptionaectstheliquidityoftheunderlyingasset.Third,optionliquiditycanbeaectedbyitsownhedgingstrategy.Anoptionwritermaychoosetohedgeshort 22

PAGE 23

Merton ( 1990 ).Theaskspreadofoptionisthecostofhedgingashortoption,whichistheexpectedvalueofaccumulatedhedgingcostsuptomaturity,thehedgingcostisthehalfspreadoftheunderlyingequity.Inotherwords,itisthemarketordercostforeverytradeofhedgingstrategy.Similarly,thebidspreadisthecostofhedgingalongoption,ortheexpectedvalueofaccumulatedhedgingcost,whichisthenegativeofthemarketordercostofhedging.Therefore,theGlosten-Milgrom'sBayesianNashEquilibriumcannotbedirectlyusedinthissetupbecausetheexpectedhedgingcostsalreadyaccountforadverseselectionintheunderlying.Moreover,ithasthemarket'sexpectedadverseselectionuptomaturityatallpossiblepriceevolutions.Itisinfacttheexpectedvalueofhedgingcostsinrisk-neutral(onlyneutralinasenseofpublicinformationrisk)probability. Thereisstrongempiricalevidencetosupporttherelationshipbetweenadverseselectionandbid-askspreads.AdverseselectioncostsaresignicantinOTMoption,sinceOTMoptionshavebiggerspreads,arelessfrequentlytraded,andhavebiggertradesizepertransactionthanATMoption,aslowpricelureinformedtradersintoseekleverageeects.HedgingcostissignicantinATMandOTMoptionsinlongdurations,sincemarketmakersshouldhedgetheiroptioninventory.Iftheunderlyingmarketishighlyliquid,thenthewholespreadisdeterminedbytheliquidityintheunderlyingsecuritymarketandhedgeratio.BecausethisisnotthecaseinthedierencebetweenATMandOTMoptionspreads,itisevidencethatamarketmaker(orequivalentlyoptionwriter) Martellini ( 2000 ).

PAGE 24

Kyle ( 1985 )'smodeltodirectionalinformationandoptionpricechangecouldnotbefound. Thelastdimensionispricereversal.Iwasnotabletondanyliteraturethatexaminespricereversaleectinoptionmarkets.Similartotheequityprice,anoptionpricemaytendtoreversewhenuninformativeshockoccur.AlthoughIexpectthatthedirectionaleectwillbemuchlessthroughhedging,uninformativevolatilityshockwillhaveasignicanteectonoptionpricereversal.Table1.1isasummaryofliquiditydimensionsandpropertiesinequityandoptionmarkets. Sofar,wehaveexaminedliquiditymeasuresofequityandoption.Nowamarketmakercanwatchthosemeasurestodetectanyinformedtradingpossibilitiesintheorder ChoandEngle ( 1999 ), BiaisandHillion ( 1994 ), Bollen,Stoll,andWhaley ( 2003 ),and Fontnou-velle,Fishe,andHarris ( 2003 ).24See Fontnouvelle,Fishe,andHarris ( 2003 ).25See Easley,O'Hara,andSrinivas ( 1998 ).26See ChoandEngle ( 1999 ).

PAGE 25

LiquidityDimensionsandTheirComponentsinEquityandOption Breadth(Bid-askspread)AdverseselectioncostHedgingcostOperatingcostAdverseselectioncostInventorycostOperatingcostCompetitioncost PriceimpactOrderimbalanceSignofordervolumeUncertaintyofuninformedvolumeNotmuchdirectionalimpactUncertaintyofpricedistributionvolatilityimpact PricereversalOrderowNotclearUncertaintyofpriceUncertaintyofprivateinformationUncertaintyofdividends 25

PAGE 26

InChapter 2 ,weinvestigatetheempiricalevidenceofstructuralbreaksinpricedynamicsofequitypriceswhennewkeydevelopmentsaectingfundamentalvalueareannounced.Weshowtheexistenceofatransitionregimethatcanbeattributedtoactivitiesperformedbyinformedtraders.Wefocusonstructuralbreaksduetofundamentalvaluechangesspecictoindividualrms,thusexcludingcasesofliquidityormarket-widechanges.Liquiditychangesnecessarilyinvolvemeasurementsofliquiditysuchasadverseselection,depth,andresiliency.Furthermore,theyrequireananalysisofintradaytradesandquotesdata.Market{orindustry{widechangescanberegardedsimilarlyifweconsiderthemarketorindustryportfolioasasingleequity.Fromanempiricalpointofview,amarket-widekeydevelopmentislikelytobeincorporatedinthereturnofthemarketportfolio.Thereforeitseectonaspecicequityreturncanbelteredoutbyusingabnormalreturnsrelativetomarketbenchmarks,suchasthosegeneratedbytheCAPMandtheFama-Frenchthree-factormodel. Themethodologyforthisresearchisbasedon Chib ( 1998 )'schange-pointmodel,whichhashadnumerousapplicationsintheanalysisofequitypricedynamics.Webrieydiscusstwopapersthatareclosesttothepresent. PastorandStambaugh ( 2001 )investigatewhetheralongreturnhistoryisusefulinestimatingthecurrentequitypremiumevenifthehistoricaldistributionhasexperiencedstructuralbreaks.Theirapplicationisbasedonachange-pointmodelonavalue-weightedportfolioofNYSE 26

PAGE 27

PastorandStambaugh ( 2001 )showthatachange-pointmodelenablesustouselonghistoricaldatawithhighprecision. LiuandMaheu ( 2008 )researchstructuralbreaksinrealizedvolatilityoftheS&P500indexfromJanuary1993toMarch2004.Becausetheinstabilityofthevolatilityprocesshasimportantimplicationsfordecisionsinriskmanagement,portfoliochoice,andderivativepricing,theseauthorsuserealizedvolatilityasitprovidesanaccurateestimateofexpostvolatilityandbecauseofitsabilitytobeincorporatedinabroadclassofcontinuous-timemodels.TheyuseaHAR-GARCH27modelforestimationandachange-pointmodeltodeterminestructuralbreaksinrealizedvolatility.Theyndstrongevidenceofstructuralbreaksinlog-realizedvolatilitybutweakerevidenceforchangesinboththeregressionparametersandvarianceoftheHAR-GARCHmodel.

PAGE 28

InChapter 3 ,weconstructbidandaskequitypricingstrategiesinasemi-strongecientmarketandevaluatetheirempiricalperformancewithtradeandquotedata.Throughthisanalysis,ourgoalistodetectpatternsdistinguishingtransitionfromstableregimes. 2 ,ourresearchfocusesonrm-specickeydevelopmentscontainedinannouncementdatesfromnancialmediaandrelatedabnormalreturndata. PastorandStambaugh ( 2001 )and LiuandMaheu ( 2008 )provideempiricalevidenceofanexpectedequitypremium,whichcorrespondstothedirectionofanindividualequity.Theirstudiesalsosupporttheexistenceofstructuralbreaksinrealizedvolatilityovertime.Insteadofanalyzingonelongtime-seriesastheydo,weconsiderseveral,witheachcontainedinatimewindowassociatedwithakeyevent.Ourmajorcontributionfromthissetupistonotonlyidentifysignicantstructuralbreaksaroundkeydevelopmentannouncements,butalsotoprovideanalternativetoconventionaleventstudies.Thisisduetothefactthatourmethodenablesustoestimatevariousregimelengths,detectnewdirectionsandvolatilities,andassessthecorrelationbetweenconsecutiveregimes. Anothercontributionofourresearchistoprovidesomegroundworktolinkfundamentalanalysis,liquidity,andpricedynamics.Thisisparticularlyimportantforpopularoptionpricingmodelsbasedonjump-diusionsandstochasticvolatility. InChapter 3 ,themethodologywedevelopisanextensionofthesequentialtradingmodelsof GlostenandMilgrom ( 1985 ), Easley ( 1996 ), CherianJ.A. ( 1998 ), Easley,O'Hara,andSrinivas ( 1998 ),tonameafew.Themaincontributionofthispaperisthatourmethodologyextendsthosesequentialtradingmodelsintoacontinuoustimeframework,whichcansimulaterealpricedynamicsdirectlycomparablewithspread 28

PAGE 29

Roll ( 1984 ), George,Kaul,andNimalendran ( 1994 ), Madhavan,Richardson,andRoomans ( 1997b ).Althoughourempiricalstudydoesnotfollowtheframeworkofspreaddecompositionmodels,itfocusesonhowintradaytradesandNBBOdatamayactdierentlybetweentransitionandstableregimes.Furthermore,thestudyprovidesparameterstosimulatestrategiesofbidandaskpricing. 29

PAGE 30

Thischapterisorganizedasfollows.InSection 2.1 ,wepresentanapplicationof Chib ( 1996 )'schange-pointmodel.Section 2.2 containsourempiricalresults. Chib ( 1998 )'schange-pointmodel.OurmodelestimatesthelengthofeachregimebyGibbssampling,ndsmaximumlikelihoodestimatorsofdirectionandvolatilityforeachregimebytheMonte-Carloexpectation-maximization(MCEM)method,andproducesthemarginallikelihoodofeachmodel(associatedwithaspecicnumberofregimes)tondaBayesfactorthatevaluatesthemodel'sperformancerelativetotheothers.Thestatisticalmethodsusedtoanalyzeestimatesofthechange-pointmodelaredescribedmorefullyinSection 2.2 FamaandFrench ( 1993 )togeneratenormalreturns.Thepredictionsforthelatterarethevalue-weightedmarketreturn(denotedrn),theaveragereturnonasmallmarket-capitalizationportfoliominustheaveragereturnonalargemarket-capitalizationportfolio(SMB),andtheaveragereturnonahighbook-to-marketportfoliominustheaveragereturnonalowbook-to-marketportfolio(HML).Abnormalreturn,anktforrmnonitskthkeydevelopmentattimetisdenedasthedierencebetweenrnktandpredictedequityreturn,^rnktbasedontheFama-Frenchthree-factormodel: 30

PAGE 31

TransitionDiagramofRegimeChange where ^rnkt:=^n+^nrmkt+^snSMBkt+^hnHMLkt;(2{2) wherermkt,SMBmkt,andHMLmktare,respectively,thevaluesofrm,SMB,andHMLfortthdateofkeydevelopmentk.Theregressionestimates^n,^n,^sn,and^hnaregeneratedbytheEVENTUSsoftware,throughaWRDSinterface Cowan ( 2007 ).Theyarebasedon\normal"returndataconsistingofobservationsoutsideeventwindows(eachoflengthT).TheestimationmethodwechoseisthatofEGARCH(1,1),whichaccountsforthepossiblecorrelationbetweenchangingreturnsandvolatilityaswellasmodelstochasticvolatility.Sincethereareusuallyatmostonekeyeventonanyday,ourentirestudyisbasedondailydata. Letf(atjs;2s)denotethedensityoftheabnormalreturnwhentheregimestattimetiss.ThevariablestisassumedtofollowaMarkovchain,st+1canstaythecurrentvalueofstorjumptothatofst+1.Figure 2-1 showsatransitiondiagramofregimechanges, 31

PAGE 32

Let=(s;2s)fors21;:::;S.Forourmodelweneedtoestimatethejointposteriordistributionoftheparameterset=s;1sSandtransitionprobabilitymatrixP,denoted(;PjAT),whereAT=a1;a2;:::;aT,aswellastheposteriorprobabilitiesp(stjAT),1tT,fortheregimesetST=s1;s2;:::;sT.Sincethelatterisunobservable,weadopttheBayesianparadigmbyspecifyingconjugatepriors.Moreexplicitly,usinggivenhyper-parameterstobedetailedlater,weemploythefollowingdistributions: ByBayestheorem,(;PjAT)/(;P)f(ATj;P).Sincedependsontheunobservedregimesfs1;s2;:::;sTg,weaugmenttheposteriordensityto(ST;;PjAT).WethenuseaGibbssamplertogenerateposteriorrealizationsoftheparameters,(P;;ST)asfollows: 1. GenerateP(PjST) 2. Generate(jST;AT) 3. GenerateST(STj;AT;P) 32

PAGE 33

wherem:=mss,m=0 mas,m=0+m,m=0+m,2m=20+(m1)S2s+0 whereSt:=fst;:::;sTg.Weomitthe(sTjAT;;P)and(s1jAT;S2;;P)termsbecausep(sT=S)=1andp(s1=1)=1byconstruction.ThengeneratingposteriordistributionofSTreducestogeneratingf(stjAT;St+1;;P)g,t=T1;:::;2,recursivelytakingtheirproduct. Chib ( 1996 )showedthat wherest+1isknownfromthepreviousiteration.GivenST=sT=S,sT1;sT2;:::;s2aregeneratedthrough(2.8)asfollows: 33

PAGE 34

Asshownin(2.7),therighthandsideof(2.6)requiresnding(st=kjAt;;P)and(st+1jst;P).ThelattercanbefoundthroughtheconditionalMarkovtransitionprobabilitygivenSTfromStep1.Tondtheformer,weconsideranotherrecursivecalculation.Becauseweknow(s1=1jA1;;P)=p(s1=1)=1,wecancalculate where Byconstruction,wehave(s1=1jA0;;P)=p(s1=1)=1. ThroughtheGibbssampler,wecansimulatetheposteriordistributionofregimesas: whereMisthenumberofiterationsoftheGibbssamplerandsuperscriptisusedforsampleidentication. 34

PAGE 35

WendthemaximumlikelihoodestimatorsofdirectionandvolatilityofaregimebyapplyingtheMonteCarloexpectation-maximization(MCEM)algorithm,sincethelikelihoodfunctionisintractablewithunobservableregime RobertandCasella ( 2005 ).TheMCEMalgorithmproceedsasfollows: Ineachiteration(i),i=1;:::;IwegenerateMsampleregimesetsS(1)T;S(2)T;:::;S(M)Twhereform=1;:::;M,S(m)Tisgeneratedaccordingto(STjAT;(i)).Wethenobtainthefollowingexpectationestimate: Thesecondtermf(S(m)TjP)isxedbecausef(s(m)1=1jP)=f(s(m)T=TjP)=1and ^pss=PMm=1m(m)ss wherem(m)ssistheexpectednumberofdatesthesystemremainsinregimesduringthemthiteration. =argmax1 Itisnowstraightforwardtondthemaximumlikelihoodestimatorbecauseweknowtheposteriordistributionsofdirectionandvolatility.TheMLEofmeandirectionforthe 35

PAGE 36

bythevirtueofthefactofthenormalposteriordistributionofthedirectionparameter.Similarly,giventhatgammaposteriordistributionofthevarianceofabnormalreturnthevarianceforthesthregimeis: wherea(m)sisdenedin(2.15).Forthesimulation,thesamplesizeMandIaresetlargeenoughsothedierencebetweentwosuccessiveestimateisnegligible. Chib ( 1996 ),themarginallikelihoodforabnormalreturnsinachange-pointmodelMrwithrregimescanbeexpressedas: wherethedensitiesareaspreviouslydenedwithnowanexplicitreference.Theoreticallysettinganyvaluesinf;Pgdoesnotaecttheresult,butpracticallyweuseMLEestimatesfromtheprevioussection.WiththismarginallikelihoodwecancomparetwomodelsMrandMsbyusingtheBayesfactordenedas: orsimplytakeadierencebytakinglogforbothmarginallikelihoodsas: lnBrs:=lnm(ATjMr)lnm(ATjMs):(2{19) 36

PAGE 37

Jereys ( 1961 ),alargevalueoflnBrsindicatesthatthedatasupportsMroverMs.Wecanfurtherexpressthelog-scalemarginallikelihoodasfollows: lnm(AT)=lnf(ATj)+ln()+ln(P)ln(jAT)ln(PjAT;):(2{20) NotethatwhenthereisnoriskofconfusionwesuppressthemodeltermMinournotation.Thersttermontherightsidecanbemademoreexplicitasfollows: lnf(ATj)=TXt=1lnf(atj)=TXt=1lnSXs=1f(atj;st=s)(st=sjAt1;)!;(2{21) where(st=sjAt1;)isgivenin(2.8).Thesecondandthirdtermsin(2.20)canbeeasilyfoundbypluggingMLEestimatesofdirectionandvolatility.Forfourthterm,weagainusesimulationestimates: whereS(g)Tisthegthdrawfrom(STjAT),asdescribedattheendofSection2.1.1.Thefourthtermissimilarlyestimated: whereS(g)Tisthegthdrawfrom(STjAT;),asdescribedabove.Ineachcase,wehave 2.2.1Data 37

PAGE 38

Basedonthesampledrms,weused ReutersKnowledge ( 2006 )tosearchfor'High'signicantdevelopmentsreportedfrom2004to2006.Asaresult,273rmshaveononeormoresignicantdevelopments,26rmshadmorethan30signicantdevelopmentsfromatotalof4,114developments.TheReutersKnowledgewebsitegivesthespecictimeofeachannouncementandsomeexpresstheinterarrivaltimeofkeydevelopmentsinminutesforeachrm. EventstudydataareobtainedfromEVENTUSatWRDS.Eacheventdateisselectedfromakeydevelopment,andbasicFama-FrenchDailyeventstudyisperformedwiththissoftware.TheindexforthemarketfactorisCRSPvalue-weightedreturn. Foreverykeydevelopment,61daysofabnormalreturnsarecollected:30beforetheeventdate(asreportedinEVENTUS),30after,andthereturnontheeventdate.Thecorrespondingeventwindowsaretheintegerscountingthedaysbeforeandaftertheeventdateintheinterval[30;30],with0correspondingtotheeventdates.Associatedsub-intervalsarethepre-eventwindow[30;2],eventwindow[1;0],andpost-eventwindow[1;30].ForetheFama-Frenchparameterestimationperiod,themostonecanusein3255daysintheyears2004-2006,fromwhichthelast46tradingdaysareremoved.Fromwhatremains,setsof61daysareremovedwheneveranassociatedeventoccurs. AllmethodologyandbackgrounddataprocessingareimplementedthroughthestatisticalsoftwareRandtheMySQLdatabase.InsomecasesRproducesvalueofzeroforthenormaldensity.Theyarereplacedbythenumber1010.Furthermore,caseswherethelog-likelihoodfunctionyieldsverylargenegativevaluesaredeleted. 38

PAGE 39

Chib ( 1998 ),theprioreectofchoosingdierentexpecteddurationsforeachmodelisminimalinpractice. TheGibbssamplertondbreak-pointsisiterated500timesafter150transientiterations,andtheMCEMalgorithmtondmaximumlikelihoodestimatorsisiterated150times,witheachiterationtaking50,100and200sub-iterationstondthesamplemeanin(2.15)andvariancein(2.16). Campbell,Lo,andMacKinlay ( 1997 ).Aweaker,butstillindicativeoftheindependenceassumptionistherejectionofanAR(1)modelforthetimeseriesofabnormalreturnsaswedonext.Inthiscase,thenullhypothesisstatesthat IfisinsignicantorthecorrespondingR2isnearzero,ourindependenceassumptionisempiricallysound.Table 2-1 reportssummariesofkeydevelopmentsandcorrespondingabnormalreturnscategorizedbyindustrysectors.Wecanseethatthestandarddeviation 39

PAGE 40

Direction,Volatility,andAR(1)TestforAbnormalReturnsbyIndustry FirmsEventsMarketCap.BTMEventAbnormalReturn^ Materials2814.18.520.47(30;2)0.0311.809-0.04810.0065(1;0)0.1193.138-7.42930.0000(1;30)0.0281.8420.0023 Industrials2313.39.110.41(30;2)0.0052.085-0.03010.0074(1;0)0.074.087-4.07410.0000(1;30)-0.012.1370.0009 ConsumerDiscretionary2617.77.080.42(30;2)0.0251.711-0.01900.0060(1;0)0.2193.345-3.15380.0016(1;30)0.0211.8060.0004 ConsumerStaples2514.727.690.43(30;2)-0.0061.7840.00410.0067(1;0)-0.0023.6680.60660.5441(1;30)0.0431.8750.0000 HealthCare2521.218.330.43(30;2)-0.0072.283-0.01720.0056(1;0)-0.1654.195-3.06890.0022(1;30)-0.022.4460.0003 Financials268.54.950.41(30;2)-0.0131.5070.02770.0087(1;0)0.1712.3133.19930.0014(1;30)0.0031.6530.0008 InformationTechnology2517.86.820.47(30;2)-0.0222.418-0.03730.0061(1;0)0.1413.504-6.10050.0000(1;30)-0.0072.6330.0014 TelecommunicationServices2018.752.970.49(30;2)0.0111.8720.03410.0067(1;0)0.0082.5325.12370.0000(1;30)0.0311.9290.0012 Utilities2018.79.430.46(30;2)-0.0080.968-0.01250.0067(1;0)0.0881.534-1.87930.0602(1;30)0.0110.9820.0002

PAGE 41

2-2 presentsthesamedatabutcategorizedbybig/smallmarketcapitalization,andvalue/mid/growthbook-to-marketratio.Thesmallmarketcapitalizationshowsabout10timesbiggerjumponaveragesabnormalreturns.Theabnormalreturnsinthevaluebook-to-marketcategorydrop0:04%onaverage,whereasforthegrowthcategorytheyjump0:07%onaverage.AsinTable 2-1 ,the^andR2valuesareindicativeoflowornoserialcorrelationforeverycategory. log(ki)=+log(k(i1))+i;i=1;:::;K;(2{26) whereiiswhitenoise.Weuselog-scaleinterarrivaltimessincetheyonlyallowpositivevalues.ThestationarityofinterarrivalcanbeassessedwiththeKwiatkowski-Phillips-Schmidt-Shin(KPSS)test Kwiatkowski,Phillips,Schmidt,andShin ( 1992 ).The 41

PAGE 42

Direction,Volatility,andAR(1)TestforAbnormalReturnsbyMarketCapitalizationandBook-to-MarketRatio Thistablepresentssummarystatisticsforabnormalreturnsfromatotal237rms,categorizedbybig/smallmarketcapitalizationandValue/Mid/Growthbook-to-marketratiosfrom2004to2006,followingthecategorizationruleinSection2.2.1.ColumnlabelsareidenticaltoTable 2-1 FirmsEventsMarketCap.BTMEventAbnormalReturn^ Small1019.81.080.45(30;2)0.0032.23-0.01840.0041(1;0)0.1613.889-4.48890.0000(1;30)0.022.3340.0003 Value8214.97.940.71(30;2)0.0061.8-0.03220.0037(1;0)-0.0423.304-8.75140.0000(1;30)-0.0121.9290.0010 Mid8316.426.710.4(30;2)0.0011.736-0.02490.0035(1;0)0.133.025-7.10850.0000(1;30)0.0261.7820.0006 Growth7214.615.170.19(30;2)0.0052.240.00540.0040(1;0)0.0723.6371.37070.1705(1;30)0.0032.3710.0000

PAGE 43

Table 2-3 reportsresultsonautocorrelation,stationarity,distributiontforkeydevelopmentinterarrivaltimescategorizedbyindustrysectors,marketcapitalization,andbook-to-marketratios.Overallestimatesof^areverysmall,nearzeroforallrmsandtheR2ofthemodelshowstheeectof^tobeweak.Hencewecanconcludethattheinterarrivaltimeseriesthroughoutallthecategoriesarenotlikelytobeseriallycorrelated.Theresultallowstoassumethatinterarrivaltimesaredistributedindependently.IntheKPSStest,wecompareKPSSlevelandKPSSvalueasinTable1atpage166in Kwiatkowski,Phillips,Schmidt,andShin ( 1992 ).Ifweconsidercategoriesaccordingtoindustryandmarketcapitalization,thenweseethatKPSSlevelvaries.However,forcategoriesalongbook-to-marketratiosshowssubstantialstationarityby5%criticallevel.Theseresultsimplythatthebook-to-marketratioofarmisabetterindicatorofnewkeyarrivalthanothertwofactors.FromtheKStest,wendthattheDstatisticsofexponentialdistributionareclosertozerothanthoseofthenormaldistribution,Hencewecanarmthatinterarrivaltimedistributionofkeydevelopmentfollowsindependentexponentialdistribution,andthatdistributionislikelytohaveastationaryparameter. 2-2 showsthemarginalposteriorprobabilityPssgivenfanktgduringaneventperiod.TheeventdateisApril6th,2006,andthermisJ.C.Penny.Wehavetwokeydevelopmentsonthatday:oneisthatJ.C.PenneyreiteratedQ1EPSguidanceandtheotheristhatS&PraisedJ.C.Penney'screditrating.Clearly,theRegime2probabilitiesshowthatinformationmayhavebeenknowntoinvestorsabout15daysbeforetheannouncement.ThemostrecentkeydevelopmentaboutJ.C.Pennywas 43

PAGE 44

AR(1)Test,StationarityTest,andDistributionFittingforInterarrivalTimesofKeyDevelopments 2-1 and 2-2 .TheestimatefortheinterceptoftheAR(1)modelisomitted. ^KPSSKS-EKS-N Energy1620.190.063.170.000.060.120.10.080.310.180.00Materials3530.220.063.860.000.040.190.10.070.100.190.00Industrials3390.200.053.890.000.040.690.00.070.050.200.00ConsumerDiscretionary4930.240.054.950.000.050.350.10.070.010.150.00ConsumerStaples3470.170.053.190.000.030.220.10.070.050.160.00HealthCare5660.220.045.240.000.050.230.10.070.010.210.00Financials2190.230.064.060.000.070.580.00.080.140.190.00InformationTechnology4280.230.045.160.000.060.510.00.100.000.100.00TelecommunicationServices4070.260.046.720.000.101.090.00.100.000.220.00Utilities3800.200.053.990.000.040.100.10.080.020.110.00 Big27130.250.0214.380.000.070.170.10.050.000.170.00Small9810.170.035.860.000.030.120.10.090.000.180.00 Value12900.260.0310.490.000.080.820.00.050.000.180.00Mid13470.230.038.670.000.050.780.00.060.000.190.00Growth10570.240.039.030.000.070.520.00.060.000.170.00

PAGE 45

PosteriorProbabilitiesofThree-RegimeModelinanEventPeriod announcedonMarch2nd,morethanamonthprior.Thereforeitisnotlikelythatregime2isaresultofthepreviousnews.Figure 2-3 presentsasetofconvergencegraphsofMLEestimatesforfsgandfsgfromMCEM,withthesamedataasinFigure 2-2 ,Therstrowplotsfsgandsecondrowplotsfsg,fors=1;2;3,consecutively.InthiscasewecaneasilynoticethatMLEestimatesconvergequickly.Weidentifythatinregime2,directionisnegativeandeventhoughobviouslygoodnewswasannounced,directioninregime3isalmost50%lowerthanthatinregime1.Volatilityinregimes1and2areverysimilar.However,inregime3,itjumpsupto80%thevalueinregimes1and2.Thisresultimpliesthatthemarketovervaluedtheequitybeforetheannouncementandthatthepost-announcementperiodshowshighervolatilitywithadropindirectioncomparedtothepre-announcementperiod. 45

PAGE 46

MLEConvergenceofsandsinMCEM whichhasthehighestBayesfactorrelativetoalltheothers.Thisapproachisempiricalandisonlymeanttosuggestgood,plausibleregime-switchingmodels.Togetstatisticallymeaningfulresults,werestrictourdatatothe30mostrecentkeydevelopmentsandtheirassociatedabnormaldatafor26rmsintheyear2004to2006. LetBrsdenotetheBayesfactorofamodelwithrregimesrelativetoanotherwithsregimes.Achange-pointmodelwithrregimewillbethebestifitattainsthelargestvalueoflnBrsforallvaluesofsinf1;2;3;4g.Sincethiscomparisonsmustaccountforall30eventsinvolved,thecriterionisinfacttheaverageoveralltheseevents,denoted lnBrs. Jereys ( 1961 )suggeststhatamodelMrwithrregimesmustsatisfy lnBrs>2forallsinordertobeconsideredamongthebest. Table 2-4 reportstwoexamplesofselectingthebestchange-pointmodel.FirmAA,Alcoa,hasfourcompetingchange-pointmodelsfrom1regimeto4regimes.ObservingthesamplemeanofBayesianfactor lnBrs>2on30keydevelopments,wecanidentify 46

PAGE 47

SelectionoftheBestChange-PointModel:Example ThetablereportsGibbssamplerresultdescribedinSection2.1.2.26rmshaving30ormorekeydevelopmentareselected,andlnBrsiscalculatedbychange-pointmodelhavingrregimes.Therstfourcolumnsshow lnBrs(>2)tocompareMrwithMs.Thenextfourcolumnsshow95%uppercondencelimitfromone-sidedttest. ComparedModelUpperCondenceLimit TickerModelM1M2M3M4M1M2M3M4 JCPM10.0030.6328.7112.44NA67.5957.2230.92M2-30.630.00-1.92-18.196.32NA33.2323.93M3-28.711.920.00-16.27-0.2137.07NA11.13M4-12.4418.1916.270.006.0360.3143.68NA Table2-5. TheBestChange-PointModels Numberforeachtickeristhebestnumberofregimesinchange-pointmodel,whichisselectedbythebiggestsumofpvalues TickerMrTickerMrTickerMrTickerMr thatM4hasthebiggest lnB4soveralls=1;:::;4.Thereforeweselect4-regimechange-pointmodelforAlcoa.ThesecondexampleofJ.C.Pennyalsoshowsthatover30samples,one-regimemodelispreferred.FollowingthewayintroducedinTable 2-4 ,Table 2-5 reportsthebestchangepointmodelsbyallrms.Among26rms,14rmspreferone-regimemodel,hencetheydoesnotshowspecicpatternofregimenumbers.Theresultmayimplythatoverahalfofallsamplermsmayhaveexperiencednosignicantimpactinabnormalreturnfromanykeydevelopment.Otherimplicationisthatnoonechange-pointmodelisdominantthroughout30events.Forexample,someeventsprefer3-regimemodel,andothereventsprefer4-regimemodel,sothat1-regimemodelmay 47

PAGE 48

2-6 showsthatonlyoneregimemodelispreferredwhenwetakesamplefromallrms.Oneinterpretationofthisresultispossible.Equitymarkethasaconstantowofnewdevelopmentfromvariousrms.Therefore,ifwetakeaverageonabnormalreturnofallrms,approximatedabnormalreturnofmarketisnotsensitivetoindividualrm'skeydevelopmentarrival. 2-2 ,regime2islikelytobeginabout15daysbeforethenewsisannounced,andregime3beginsrightafterthenewsisannounced.Wecansuspectupto15daysbeforetheannouncements,priceadjustmentsinthemarketmayhavealreadystarted.Table 2-7 reportslengthsofregimesinthebestchange-pointmodels,excludingone-regimemodel.AtM2,therstregimelengthisalwayslessthan30days,implyingtherstregimeisnishedbeforeannouncement.Itmaybeinterpretedthatatthesecondregimekeydevelopmentisincorporatedinequitypricebeforetheannouncement.AtM3,thesecondregimetendstoendrightbeforetheannouncementdate.AtM4,3rdregimehasannouncementdateinmostcases.Theseobservationsindicatethatpriceisnotadjustedinstantlyateventdate,butgraduallyadjusted,eitherbyreachingconsensusatnewpricelevelorbyadverseselectioneectcausedbyinformedtrader.Therefore,wedenotearegimehavingannouncementdateisatransitionregime. 48

PAGE 49

TheBestChange-PointModelforAllFirms ThetablereportsGibbssamplerresultdescribedinSection2.1.2.26rmshaving30ormorekeydevelopmentareselected,andlnBrsiscalculatedbychange-pointmodelhavingrregimesfromall26rms.Nextfourcolumnsshow lnBrs>2tocompareMrwithMs.Thenextfourcolumnsshow95%uppercondencelimitfromone-sidedttest.Thelastfourcolumnsshowpvaluesdeterminingwhethernullhypothesisisrejected. ComparedModelUpperCondenceLimitpvalue ModelM1M2M3M4M1M2M3M4M1M2M3M4

PAGE 50

LengthsofRegimesintheBestChange-PointModels Tablereportssampleaverageofmssanditscondencelimitssampledfrom30keydevelopmentsbyeachrms.mssisthenumberofone-steptransitionsstayingthesameregimes.Thisvaluecanbeinterpretedasthenumberofdaysinregimes.Firsttwocolumnsindicatetickersymbolsandchange-pointmodelhavingrregimes.Thenextfourcolumnsoftherstrowshowstheaveragelengthofregimes,^mssinaneventperiodof61days.Thesecondrowshows95%condencelimitoftwo-sidedttest. TickerMr^m11^m22^m33^m44 2-3 ,inthatvolatilityintransitionregimemaybehigherthanthatinstableregime.Table 2-8 reportssamplemeansandWelch 50

PAGE 51

DirectionandVolatilityDierencesinConsecutiveRegimes 2-5 ,excludingrmspreferoneregimemodel.FirstpanelshowsthesamplemeanofMLEofdirections,^s,on30samplesbyeachregime,and95%condencelimitandpvaluefromWelchtwosamplet-test.Nullhypothesisonthetestisthatthedierencebetweenmeansofthedirectionsonconsecutiveregimesarezero.ThesecondpanelpresentssamplemeanofMLEvolatilities^s,andt-testresults.^sand^sareestimatedbyMCEMdescribedinSection2.1.3. WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr ^1

PAGE 52

Anotherissueisthepossibilityofrelationsamongdirectionandvolatilityoncurrentandpreviousregimeswhenstructuralbreakoccurs.Asimplepairoflinearautoregressivepaneldatamodelscanbeusedtoestimatetherelations: ^j=10+11^i+12^i+13^j+1j; ^j=20+21^i+22^i+23^j+2j; wherei=1;:::;S1,j=2;:::;S,andtheindexofeventsamplenkaresuppressed.Table C-5 showsautoregressivepaneldatamodelestimationresultsontheprevailingchange-pointmodels. 52

PAGE 53

WereportOLSestimatesofsimpleautoregressivepaneldatabetweendirectionandvolatilityofcurrentregimeandofthelastregime.Nullhypothesesoncoecientsares=0,s=1;:::;S.Explanatoryvariableisspeciedatcolumn^=^,andcomparingpairofregimesis(i;j),jiscurrentregimeandiispastregime.Superscriptsdistinguishexplanatoryvariablesin(2.28)and(2.29)aresuppressed. ^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 BIIB3(1;2)^-0.13080.5223-0.32948.4220.00590.8230.81230.00960.2365^0.01070.00230.09590.027400.98650.21930.00960.3512(2;3)^00.0224-0.09760.42610.99430.18960.66250.26090.0725^0.0131-0.02680.0210.113400.00090.85580.26090.4035 Continuedonnextpage

PAGE 54

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 CAT3(1;2)^-0.0157-0.03710.20231.02980.11140.92960.71450.15050.0823^0.0106-0.0988-0.22980.075600.38280.11670.15050.1682(2;3)^0.00130.0005-0.08810.0280.73190.99390.71040.92180.006^0.00970.02180.02160.013500.64140.89550.92180.0115 CELG4(1;2)^-0.01570.13412.8711-0.39830.2710.80220.010.37710.2377^0.01290.43130.4765-0.07560.0320.05540.35330.37710.1649(2;3)^0.026-0.31910.1685-1.08870.16510.17880.74690.18150.1033^0.0175-0.1189-0.0052-0.062100.03110.96650.18150.1919(3;4)^-0.01670.0450.14520.65380.0260.51180.59490.00930.24^0.0213-0.0265-0.1350.356200.60050.50240.00930.2404 Continuedonnextpage

PAGE 55

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 KFT2(1;2)^-0.00450.01950.0140.52520.00010.37160.833300.5771^0.0091-0.0364-0.0611.093100.24510.522300.5903 PTRY4(1;2)^-0.00710.07780.6055-0.13460.73340.87810.52820.88420.0201^0.0157-0.0276-0.1959-0.006200.79930.33830.88420.0423(2;3)^0.0125-0.0921.2469-1.1230.52460.69510.24620.07670.1731^0.0175-0.086-0.0632-0.10290.00110.21990.84810.07670.1713(3;4)^0.0248-0.0406-0.3579-1.026900.30080.01020.00020.477^0.0201-0.0048-0.2149-0.412300.84740.01560.00020.473 TMO4(1;2)^-0.0084-0.195-1.29891.17270.42010.78380.2180.11490.1537^0.00880.0338-0.02350.07920.00030.85480.93260.11490.1021 Continuedonnextpage

PAGE 56

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 UNH2(1;2)^-0.00640.05710.16330.2490.03190.61550.19260.13020.1264^0.01480.1012-0.15720.344700.44760.28890.13020.1882 VZ3(1;2)^0.0069-0.0979-0.2916-0.44440.03130.76480.43260.1960.1065^0.00570.03770.1338-0.14270.0010.8390.5260.1960.0964(2;3)^-0.0008-0.03440.06810.16750.77430.79010.77210.54820.0171^0.00760.0611-0.21210.083800.50260.1950.54820.1109 WYE2(1;2)^-0.0041-0.04680.07120.30160.01430.67240.53350.01140.2363^0.0117-0.0248-0.06620.735700.88610.71140.01140.2273

PAGE 57

Fromtheattempttoanalyzestatisticalpropertiesfromthesampleofeventperiods,wefoundthatdirectionandvolatilitymightbedrawnfromthesamedistribution,andtheyareseeminglyuncorrelatedbetweenregimes.Thesefacts,andthendingfromSection2.2.2,provideajusticationofdesigningstochasticprocessofkeydevelopmentarrivalwithindependentandidenticallydistributeddirectionandvolatilityrandomvariables,hencewecandeneitasaMartingaleprocess. Wecanthinkoftwopossibleissuesfromtheseresults.Oneisthatapossibilityofimpactsofsamplekeydevelopmentsaresimilarenoughtobesummarizedasonedistribution.Ifkeydevelopmentaboutsomecataclysmicdamagetothermexistsbutisnotincludedourmodel,themodelparametersestimatedbygivensamplecannotbesaidtorepresentallpossiblekeydevelopment.Theotherissueisthatthelackofstatisticalpropertyisduetosmallsamplesize.Bothissuesrequirelongertimehorizonwithvarioustypesofkeydevelopments,anditmaybesubjecttothefutureresearch. 57

PAGE 58

Thischapterisorganizedasfollows.InSection 3.1 ,weconstructthebidandaskpricingstrategy.Section 3.2 illustratestheempiricalanalysisontradeandquotedata.Section 3.3 showssimulationresultscalculatedbythemodelinSection 3.1 3.1.1EvolutionofFundamentalValue Theannouncementofakeydevelopmentcreatesatransitiontimeinterval,whichwecalltransitionregime.Informedtradersacquireinformationearlierthanthepublicandtheirtradescauseadverseselectioneectsonuninformedtradersuntiltheinformationisfullyincorporatedintotheequityprice.Itmaytakeareasonableamountoftimetointerprettheimplicationofakeydevelopmentwithpricesadjustingtonewlevelsofdirectionandvolatility.Wecallthistimeintervalatransitionregime.Whenthetransitionregimeends,weassumenewlevelofdirectionandvolatilitythatpersistuntilthenextkeydevelopmentoccurs,whichagainbreaksthefundamentalvaluestructurallyfromitspast.Welabelthisintervalstableregime. Thefundamentalvalueprocessinthepresentworkisdenedasapairofprocessesoffundamentalvaluecompoundingdirectionsandvolatilitiescontinuously.Weassume 58

PAGE 59

Followingtheannouncementofthenthkeydevelopment,theresultingchangeinfundamentalvalueleadstovaluesSnandSn2,ofdirectionandvolatility,respectively,forthenextregimedrawnfromthedistributionbelow: Theabovedistributionsarealsoassumedtobeindependentacrossregimes.ApricejumpJnmayalsooccurwhenthereisagapbetweenitsmarketexpectationandthepricefollowingthekeydevelopment.WeparameterizethejumpwiththedirectionTnandvolatilityTn2forthetransitionregimeas where Wefurtherassumethattheabovedistributionsareindependentacrossregimes.Theparametersforstableregime(S;S2;S;S)andforthetransitionregime(T;T2;T;T)canbeobtainedthroughtheofeventstudymethodologywithhistoricalpricedata.DirectionalandvolatilityinformedtradersobtaintheactualvalueSnandSn2,respectively,whereasuninformedtradersandmarketmakersonlyknowtheircorrespondingdistributionsgivenabove. ThekeydevelopmentarrivalprocessI(t)isofPoissontypewithconstantrateI,andisknowntoinformedtraders.Thedurationofthenthtransitionregime,LIn,followsanexponentialdistributionwithconstantrateL.WithTInmeasuring(indays)thelength(orduration)ofthenthkeydevelopment,thecorrespondingalternatingstableand 59

PAGE 60

SamplePathsofFundamentalValue(left)and(ontheright)AssociatedDirection,Volatility,andPriceJumpsatKeyDevelopments transitionregimeintervalsaredenedrespectivelyas: and Obviously,thebeginningofatransitionregimecoincideswithastableregime'sending. Informedtradersestimatethelog-scalereturndirectionofanequity,wI(t),bycontinuouslycompoundingSn,anditsvolatilityvI(t)bycontinuouslycompoundingSn2: Figure 3-1 displayssamplepathsofafundamentalvalueprocessandassociateddirection,volatility,andpricejumpprocessesatkeydevelopments.Theleftsidegraphshowsa1-yearevolutionoffundamentalvaluewith3standarddeviationslimits.Directionaltradersbelievethatthepricewillcompoundataconstantdirectionuntilthearrivalof 60

PAGE 61

Oneofthekeyassumptionsinthispaperisthatprivateinformationcontainseitherdirectionorvolatilityaboutthenextstableregime,notacertainpricelevel.Thereforeunlikein Kyle ( 1985 )and Easley ( 1996 ),informedtradersdonottradefrominformationonapricelevel.Instead,theytradedynamicallyfollowingmarketaction.Volatilitytraderscannottakeanyadvantageinequitytradingwithoutreferringtoassociatedoptionprices.Hence,weruleouttheirroleinthissetup.Asweassumethatallinformedtradersplacemarketorders,theyhavenoabilitytochoosethesizeofthetrade.Insteadtheycantrademoreintensivelytoaccumulatethemaximumprot.Thereforethestrategyofmarket-ordertradersandthequoterevisionsofmarket-makersarebasedonthetradingintensity.WhenNBBOisrevisedbyeitheratrade,publicinformation,orkeydevelopmentannouncement,allthemarketparticipantsinitializetheirstrategyonthe 61

PAGE 62

Adirectionaltrader'straderateisproportionaltothespreadswI(t)a(t)andb(t)wI(t)asfollows: wherex+;xrefertomax(x;0);min(x;0)respectively.DB(t)andDB(t)arethetradingratesofdirectionaltraderforbuyingandsellingattimet,respectively,andcDBandcDSarecoecientsrepresentingsensitivitiestothosespreads.Basedonthisrate,directionaltradersplaceordersbyfollowingthedurationdistributionsas Alltradeandquotedurationsaremeasuredinseconds. Anuninformedtrader'straderateisbasedonarelationshipbetweenthebuyingandsellingdurationsTB;TS,andthespreadsa(t)w(t)andw(t)b(t).Inestimatingtradedirection,wemayhavecaseswherewecannotdetermineasideoftradewhenthebidoraskpricedoesnotchangeafteratrade,oriftwoormoresubsequenttradesoccuratthesameprice.Wecallthiscaseasindeterminatetrade,withdurationTI.EmpiricalidenticationoftradingdirectionB;I;SismentionedinSection 3.2.2 .Log-scaledurationlogTandthespreadsa(t)w(t)orw(t)b(t)canbeexpressedaslinearmodelssuchas:logTB:=cUB0+cUB1(a(t)w(t))+B;logTI:=cUI0+I;logTS:=cUS0+cUS1(w(t)b(t))+S;

PAGE 63

3.3 .RatesofuninformedtraderbytradedirectionsareUB(t)=1=exp(UB(t)+UB2=2);UI(t)=1=exp(UI(t)+UI2=2);US(t)=1=exp(US(t)+US2=2); AmarketmakerrevisesNBBOquotesinthreecases:afteratrade,newpublicmarketinformation,orannouncementofakeydevelopment.Whenatradeoccurs,amarketmakersetsthenextbidandaskquotesincorporatinganyinformationrevealedbythelasttrade.Accordingto GlostenandMilgrom ( 1985 ),amarketmakersetsbid-askspreadtohedgeagainsttheadverseselectionriskfrominformedtrades.NewpublicinformationrevisesNBBOquotesinarealmarketbecausemultiplemarketmakersandpubliclimit-ordertradersconstantlyupdateNBBOquotesforliquidityorfromindustry-wideinformationoracompetitor'skeydevelopment.Whenakeydevelopmentisannounced,amarketmakershouldreviseNBBOquotestoincorporatethenews;otherwiseitcancauseasignicantadverseselectionriskforanytraderagainstthemarketmaker. 63

PAGE 64

WealsoassumetherateofNBBOrevisionwithouttradeorkeydevelopment,NWOTisconstant.ItcanbefoundempiricallybysubtractingtraderatefromthetotalNBBOrate,ignoringtherateofkeydevelopmentdurationbecauseitisalotlongerthenatradeoranNBBOrevision. Fromthememorylesspropertyoftheexponentialdistribution,thenextarrivalmayhavethesmallestdurationsofar.DenotebyNBBO(t)thecountingprocessofNBBOrevisions.ThenthedurationofnextNBBOrevisionis Fortheremainderofthispaper,wesuppressthesubscriptNBBO(t)+1.ThentherateofNBBOarrivalattimetis Finally,letX(t)beacauseofNBBOrevisionattimet,thenX(t)istheoneoffUB,UI,US,DB,DS,NWOT,Ig.ThentheprobabilitythatthenextNBBOrevisioniscausedbyX(t)=xis 3-2 presentssamplecountprocessesofeachtypeofNBBOrevisionsontheleftandshowsdurationdensitiesforeachtypeontheright.Thetimehorizonis30minutes. 64

PAGE 65

ARealizationofNBBOrevision 3.1.2 ,therevisionofbidandaskpriceswithouttradeincorporatespublicinformationorannouncementofkeydevelopments.Inthiscaseweassumethatadverseselectionriskisresolvedbynewinformation,sothatthemarketmakerrevisesbid-askspreadwithahistoricalaverage.NBBOrevisionsasaresultoftrade,however,containadverseselectionriskthattradersmaybemoreinformedthanmarketmakers.Thereforemarketmakerrevisebidandaskpricesaswellasbid-askspreadstocompensateforthepotentiallosstoinformedtraders. Wesuggestthreepossiblebid-askspreadsettingstrategies.First,aconstantbid-askspreadstrategytosimplykeepthebid-askspreadconstantandonlychangeoccursatnextbidandprices.Second,aBayesianNashequilibriumstrategywithpartialmomentthatisadirectextensionof GlostenandMilgrom ( 1985 ),whereamarketmakercalculatestheexpectedvalueoffundamentalvaluegiventhenexttradedirection.Estimationofpartialmomentsisbasedonhistoricaldurationofkeydevelopment,direction,andvolatilitydata.Third,aBayesianNashequilibriumstrategywithorderimbalancethatincorporates 65

PAGE 66

SupposethenextNBBOrevisionoccursattimetandthecurrenttimeist0.DenotebyD(t)thejumpprocessoftradedirectiongiventhattradeoccursattimet.ThenDi=8>>>>>><>>>>>>:1;Ifbuy-initiatedtradeoccurs,0;Iftradedirectioncannotbedetermined,1;Ifsell-initiatedtradeoccurs. WerecallthattheestimationoftradingdirectionisaddressedinSection 3.2.2 Thenthelasttradedpriceisw(t)=8>>>>>><>>>>>>:a(t0);IfD(t)=1;a(t0)+b(t0) 2;IfD(t)=0;orNBBOrevisionwithouttradeoccurs,b(t0);IfD(t)=1 IfanNBBOrevisioniscausedbynthannouncementofkeydevelopment,thenw(t)=w(t0)+Jn: 66

PAGE 67

ABayesianNashequilibriumpricestrategywithpartialmomentsapproximatesaninformedtrader'sbuying,DB(t)andsellingDS(t)as Unfortunately,amarketmakerhaslittlemeantoestimatethecoecientscDBandcDSdirectlyfromtradeandquotedata.Instead,themarketmakerperformsasimulationofthepricedynamicstosearchthosecoecientsoptimizingperformancemeasuresdescribedinSection 3.1.4 .Belowaretheformulasoftwospreads: 67

PAGE 68

RealizationsofNBBOPriceEvolutions Then, (3{21) 3{18 ),Eq.( 3{19 ),Eq.( 3{21 ),andEq.( 3{22 )aregiveninAppendix??.Figure 3-3 presentsthreesamplepathsofNBBOpriceevolutionsfollowingthethreestrategies.Theupperpanelshowsthepriceevolutionofask,lasttraded,andbidprices.Thelowerpanelshowstheorderimbalanceofeachstrategy. 68

PAGE 69

NBBO(T)NBBO(T)Xn=1(w(Sn)wI(Sn)); NBBO(T)1NBBO(T)Xn=1(jw(Sn)wI(Sn)jME(T))2; whereSnisthesumofdurationsofNBBOrevisionsdenedas Amarketmaker'snetprotsconsistsofprotsfromuninformedtradesandlossesfrominformedtrades.Amarketmakermakesaprotfrombid-askspreadsforeachsharewhenaround-triptradehappens.WeextractthetradecountingprocessT(t)fromNBBO(t).T(t)consistsofoneoffUB(t);US(t)DB(t);DS(t)g,whichareprocessescountingthenumbersofuninformedbuy,uninformedsell,directionalbuy,anddirectionalselltradesuptotimet,respectively.Amarketmaker'sprotPM(T)uptotimeTcanbeexpressedas DecisionvariablesforselectingbidandaskpricingstrategiesarethecoecientscDBandcDSthatweuseincalculatingthebid-askspreadintheBayesianNashequilibrium 69

PAGE 70

3.2.1Data KeydevelopmentsdataforselectedrmsareobtainedfromThomsonReuters''High'signicantdevelopmentreportsduringtheyear2004-2006.273rmshaveoneormoresignicantdevelopments,26rmshavemorethan30signicantdevelopmentsfromtotal4114signicantdevelopments. IntradayquoteandtradedataareacquiredfromconsolidatedtradeandquotedatafromTAQdatabaseforselectedrmsin2006.SinceconsolidatedquotedatasuppliedbyTAQretainwholelimitorderbookdata,weextractNBBOdatabyrunningPerlscriptontheWRDSUNIXserver.FromtheresultingdataweonlyselectdistinctNBBOquotes.WelterconsolidatedtradedatabyselectingonlythoseindicatorCORRequaltoeither0;1or2,toruleoutcanceledtransactionaswellastradeswithsize0.Incaseofmultipletradesorquotesinasecond,wedistinguishthoserecodesbyaddingnumbersgeneratedfromuniformdistributionintheinterval(0;1).Forexample,ifthreetradesorquotes 70

PAGE 71

Todeterminethetradedirection,D(t)ofeachtrade,weuseanalgorithmof Ellis,Michaely,andO'Hara ( 2000 ).First,wesplitbidandaskquotesinNBBOandinterweavewithtradedata,orderingbytime.ThenwelagtheNBBOdataby1secondtoadjustwiththetimelagoftradereporting.Second,weperformpricematchingfromeachtrade.IfthemostrecentNBBOrecordbeforeatradehasanaskpriceidenticaltothattrade,thensetdirectionofthetradeto1.IfthemostrecentNBBOrecordhasbidpriceidenticaltothattrade,thensetdirectionofthetradeto-1.Third,iftradepricesandNBBOpricesdonotmatchthenweperformatick-testbysimplysettingdirectionto1whencurrenttradepriceishigherthanthepreviouslytradedpriceand-1otherwise.Ifneitherpricematchingnortick-testdeterminesatradedirection,thenwesetthedirectionto0,implyingthatthedirectionisindeterminate.FromTable 3-1 toTable 3-3 ,wesummarizethetradedatabyexchanges,tradetimeintervals,andindustrysectors,respectively.InTables 3-4 to 3-6 ,wesummarizeNBBOdatabyexchanges,tradetimeintervals,andindustrysectors. Rosenthal ( 2008 ),andisaconsequenceofthepropertyofthePoissonprocessusedinmodelingNBBO(t).

PAGE 72

ThistablesummarizesconsolidatedtradedataincludingdirectionanddurationfromTAQin2006from293rmsintheRussell3000set.Thedataarecategorizedbyexchanges.DuetothedataerrorsinTAQ,weexcludedtradedataofsomermswithexchangecode1and8,whichisabout882,590trades.Exchangecodes:`A':AmericanStockExchange,`B':BostonStockExchange,`I':ISE,`M':MidwestExchange(Chicago),`D'NASDADFandTRF,`Q'and`T'NASDAQ-NMSStockMarket,`N'NewYorkStockExchange,`C'NSX,`P'PacicExchange,`X':PhiladelphiaExchange.During2006,NASDAQexchangecodehasbeenchangedfrom\T"to\Q".Wepresentsummaryoftwocodesseparately.Price,spreads,andtradedurationaremeasuredbylog-scale.B-Aisanaveragepercentagebid-askspread.A-LTisanaveragepercentagespreadbetweenaskpriceandlasttradedprice,andLT-Bisanaveragepercentagespreadbetweenlasttradepriceandbidprice. ExchangeFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeCodeDir.atTrd.atTrd.afterTrd.afterTrd.Dur. A16Total6569315273.2040.00080.00180.00080.00122.56Buy3383155103.2060.00010.00190.00020.00162.59Indet.583315383.2610.00130.00480.00080.00162.14Sell2602855463.1880.00170.00100.00160.00062.60 Continuedonnextpage

PAGE 73

ExchangeFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeCodeDir.atTrd.atTrd.afterTrd.afterTrd.Dur. B235Total4745273983.6920.00050.00080.00030.0003-0.03Buy1936894093.6800.00000.00060.00000.00040.10Indet.644643473.7270.00230.00380.00050.0006-1.17Sell2163744023.6930.00040.00000.00040.00000.20 I51Total6931963.1020.00040.00040.00040.0004-0.99Buy2891863.0170.00000.00070.00000.0007-0.90Indet.963143.4130.00050.00040.00030.0004-1.08Sell3081683.0840.00070.00000.00070.0000-1.04 M219Total9475539343.6060.00100.00220.00040.00050.14Buy41814210723.5930.00000.00080.00000.00070.37Indet.1342876343.6680.00470.01320.00100.0014-0.96Sell3951248893.6000.00080.00010.00070.00000.27 D273Total256782084583.4850.00040.00070.00050.0007-0.69Buy124077514583.4810.00010.00070.00020.0008-0.64Indet.21005325363.5260.00140.00420.00120.0019-1.27 Continuedonnextpage

PAGE 74

ExchangeFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeCodeDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Sell111699254433.4830.00070.00010.00070.0003-0.63 Q115Total292314982913.1950.00040.00050.00050.0005-1.41Buy124467252883.2060.00000.00090.00020.0008-1.24Indet.50270583363.1340.00030.00030.00050.0005-2.28Sell117577152763.2100.00090.00000.00080.0002-1.22 T257Total342711103943.3150.00050.00060.00060.0007-0.94Buy153274934013.3170.00000.00100.00020.0011-0.81Indet.46878013923.2980.00070.00110.00080.0010-1.92Sell142558163873.3180.00100.00010.00100.0003-0.76 N148Total624699025913.6580.00030.00040.00030.00030.70Buy308249545903.6540.00000.00060.00010.00050.71Indet.42923727683.7280.00040.00060.00030.00040.30Sell273525765653.6520.00050.00010.00050.00010.74 C256Total30737762783.2880.00040.00040.00040.0004-1.34Buy12938902823.2990.00000.00080.00010.0007-1.21 Continuedonnextpage

PAGE 75

ExchangeFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeCodeDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Indet.5180722873.2510.00030.00030.00030.0003-1.99Sell12618142703.2920.00080.00000.00070.0001-1.22 P273Total389882222433.3530.00040.00040.00040.0004-1.28Buy167714332413.3770.00000.00080.00010.0008-1.14Indet.57798582733.2950.00030.00030.00030.0003-2.11Sell164369312343.3480.00080.00000.00070.0001-1.13 X99Total2001502153.6660.00020.00030.00020.00031.01Buy1117543413.678-0.00030.0007-0.00020.00070.94Indet.149363613.7190.00030.00030.00030.00031.13Sell735467533.6360.0009-0.00040.0009-0.00031.10

PAGE 76

ThistablesummarizesconsolidatedtradedataincludingdirectionanddurationfromTAQin2006from293rmsinRussell3000.Thedataiscategorizedbytradetimeintervals. TradeFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeIntervalDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Continuedonnextpage

PAGE 77

TradeFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeIntervalDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Buy267589443993.4550.00000.00070.00010.0006-0.16Indet.63072193583.3630.00020.00020.00030.0003-1.40Sell248246873793.4440.00070.00000.00060.0001-0.17

PAGE 78

ThistablesummarizesconsolidatedtradedataincludingdirectionanddurationfromTAQin2006from293rmsinRussell3000.ThedataiscategorizedbyindustrysectorclassiedbyGICS. IndustryFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeSectorDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Consumer29Total220276423193.6370.00030.00040.00040.0006-0.45DiscretionaryBuy99725583263.6450.00000.00070.00020.0007-0.30Indet.28314463143.6140.00050.00080.00050.0007-1.50Sell92236383123.6350.00060.00010.00060.0003-0.29 Consumer26Total175327354723.4750.00040.00060.00050.0006-0.10StaplesBuy85166224783.4430.00000.00090.00010.0009-0.01Indet.15163184953.6190.00070.00140.00070.0009-0.98Sell74997954603.4830.00090.00000.00080.0002-0.02 Energy27Total165417904163.6980.00040.00070.00040.00060.10Buy80839304183.6940.00000.00080.00020.00070.15Indet.14484894723.7440.00100.00290.00080.0014-0.53Sell70093714043.6920.00070.00010.00070.00020.17 Financials29Total152477513273.4470.00040.00050.00040.0005-0.37 Continuedonnextpage

PAGE 79

IndustryFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeSectorDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Buy68233003313.5000.00000.00080.00020.0008-0.16Indet.20623873243.2360.00050.00090.00060.0007-1.54Sell63620643243.4590.00070.00010.00070.0002-0.22 Health30Total253395264223.4490.00040.00050.00050.0005-0.63CareBuy114985004263.4520.00000.00080.00020.0008-0.53Indet.27818464253.4530.00050.00080.00070.0007-1.41Sell110591804173.4440.00080.00000.00080.0002-0.54 Industrials29Total161426423323.5200.00050.00050.00050.0005-0.01Buy75253633323.5440.00000.00080.00020.00080.08Indet.15954293703.4630.00080.00130.00080.0008-0.85Sell70218503253.5080.00080.00000.00080.00020.09 Information26Total294367582773.3130.00030.00030.00040.0004-1.69TechnologyBuy120810652833.3000.00000.00070.00010.0007-1.53Indet.54702662603.3790.00020.00030.00030.0004-2.46Sell118854272793.2970.00080.00000.00070.0001-1.50 Continuedonnextpage

PAGE 80

IndustryFirmsTradeTradesSizePriceA-LTLT-BA-LTLT-BTradeSectorDir.atTrd.atTrd.afterTrd.afterTrd.Dur. Materials29Total177034753743.7190.00040.00050.00040.00040.29Buy85326453743.7120.00000.00070.00010.00060.34Indet.14845514113.7590.00080.00130.00060.0008-0.32Sell76862793663.7190.00060.00010.00060.00020.36 Telecommunication25Total254072828352.8650.00050.00060.00050.0005-1.26ServicesBuy118964928502.9330.00000.00100.00010.0009-1.14Indet.27917609872.4120.00070.00130.00060.0007-2.34Sell107190307792.9070.00090.00000.00090.0001-1.11 Utilities23Total112968283993.6030.00030.00040.00030.00030.78Buy54934634013.6040.00000.00050.00010.00050.81Indet.7799235033.6300.00080.00150.00050.00060.23Sell50234423793.5970.00050.00000.00050.00010.84

PAGE 81

ThistablesummarizesNBBOquoteswithdurationfromTAQin2006categorizedbyexchangesofnationalbestaskpriceandnationalbestbidprice.Sincebidandaskpricesarenotreasonablesometimesbeforeandaftertradinghourswhentradingisrare,someofthedatamayshowextremelylargespreads. AskBidFirmsNBBOAskAskBidBidNBBOB-AEx.Ex.RevisionPriceSizePriceSizeDur.Spread AA1614695813.3911373.399921.0810.0017C54033603.815203.811050-1.5990.0003D143143362.9315882.9329790.5090.0019M7265532.0225392.0138111.5300.0027P1613112873.368923.36818-0.3270.0016T159679293.707013.701620-1.8380.0005X3313.887873.961000.853-0.0771 BB139101188.85137-4.191415.59613.0373C59148933.8610073.861252-0.5320.0012D29434943.7513623.752691-1.6200.0005M234573.6610413.661966-0.1410.0009N351308683.7313413.733019-0.4770.0005P59492933.8211503.821217-1.0530.0010T6288073.8511383.841141-0.8780.0162X114.193004.18100-0.7650.0064 CA54104203.8111383.81521-1.5500.0003B138183234.0311953.35845-0.2250.6860C252185882623.4116553.391663-0.7890.0121D2433979113.5343513.531306-1.2060.0006 Continuedonnextpage 81

PAGE 82

AskBidFirmsNBBOAskAskBidBidNBBOB-AEx.Ex.RevisionPriceSizePriceSizeDur.Spread M87371343.8410843.841382-0.3810.0008N14271788583.768793.761374-0.5210.0007P258211765113.4012503.401030-1.1050.0030T252313586103.3122243.302268-1.2420.0053X321203.723433.72100-0.8470.0053 DA152312392.9715262.9711260.6210.0021B30418203.7531643.751133-1.5930.0011C2413828623.6111713.613790-1.2240.0006D272100625383.7024083.662270-1.6810.0381M911449023.6525743.651773-1.1490.0007N148305703353.6719453.672176-1.0560.0006P27280944883.6919573.691226-1.5540.0036T1144861773.0015703.007030-1.2300.0009X484883.863913.85107-1.1530.0047 MA5202281.9431541.9325851.5400.0030B265043.6317213.621136-0.2710.0008C89358243.7915313.791150-0.4230.0008D961832323.6322693.632595-1.1330.0007M7039862.9219832.9221150.6980.0090N996665613.5823003.582208-0.0810.0007P1052100693.6318203.621172-0.7090.0010T84219603.3526013.352352-0.3800.0013 Continuedonnextpage 82

PAGE 83

AskBidFirmsNBBOAskAskBidBidNBBOB-AEx.Ex.RevisionPriceSizePriceSizeDur.Spread X12183.689503.67100-1.0390.0108 NB361842063.7248203.721222-0.4510.0005C14281825293.7418543.74900-0.5210.0007D148367325413.6629673.661937-1.0720.0006M1007279083.6031443.604167-0.0770.0007N1481674750763.5625793.561909-0.1890.0007P148407454533.6422553.641014-0.7040.0009T13934200303.6526513.651284-0.3850.0007X8487623.738423.72101-0.5130.0046 PA1612702603.407243.40807-0.4030.0017B63615363.8217503.82978-1.0030.0010C258222139563.3811493.381342-1.0540.0029D27087929063.7013023.691817-1.5270.0128M1092125813.6516613.651599-0.7180.0009N148376129093.6610063.661666-0.6830.0009P273231267983.479533.45940-0.8900.0261T264313684303.3115253.301855-1.0710.0060X8723253.816983.79100-0.6820.0172 TA159125793.7214643.71708-1.9410.0005B90104243.969013.37825-0.5460.5942C252294837873.3220853.312098-1.2610.0051D1135550622.8678472.862052-1.1710.0009M84232793.4716573.471988-0.3640.0013 Continuedonnextpage 83

PAGE 84

AskBidFirmsNBBOAskAskBidBidNBBOB-AEx.Ex.RevisionPriceSizePriceSizeDur.Spread N13929729543.6811233.681795-0.3730.0007P264279182733.3317883.331412-1.1110.0038T264413596433.2541513.244245-1.2820.0102X311014.085734.071000.2290.0095 XA21124.071004.074791.6950.0022B243.721003.713750.3500.0061C281773.721003.72261-0.7600.0054D527503.781023.78379-0.9030.0052M6133.621773.61846-1.1600.0068N8790903.721153.72578-0.5640.0050P8127243.721193.72316-0.6910.0067T31873.841923.83613-0.7570.0058X24853.651003.41100-0.2910.2316 84

PAGE 85

NBBOSummarybyTradeTimeInterval ThistablesummarizesNBBOquoteswithdurationfromTAQin2006categorizedbytradetimeintervals. IntervalFirmsNBBOAskAskBidBidNBBOB-ARevisionPriceSizePriceSizeDur.Spread Weadaptaregressionmethodologyandanotherobjectiveofthisstudyistodeterminewhetherthecoecientsassociatedwiththeabovefactorsdiersignicantlybetweenstableandtransitionregimes.Amarketmakerwhonoticeschangesinthesecoecientswouldadjusthisbid-askspreadscorrespondingly. Atrader'smotivationforherdecisionscanbeassessedthroughvariousvariables,amongwhichtradedurationanddirectionarethemosttelling.Forexample,ifthepriceismorefavorabletobuy-initiatedtraders,theywilltrademoreintensively.Hencethedurationwillbeshortandtheorderimbalancewillincrease.Thespreadbetweentheaskpriceandthelasttradedpriceafteratrade,andthespreadbetweenthelasttradedpriceandthebidpriceaftertradeindicateamarketmaker'sreactiontothebuy-andsell-initiatedtrades,respectively.Tradepriceandsizemayalsorevealatrader'sintentions. Amarketmaker'sdecisiononNBBOrevisionscanbeassociatedwithquoterelatedvariables.Theirinteractionscanalsobeusefultodierentiatethedegreestowhichinformationisincorporatedbetweenrevisionwithtradeandthosewithout.NBBO 85

PAGE 86

NBBOSummarybyIndustrySector ` ThistablesummarizesNBBOquoteswithdurationfromTAQin2006categorizedbyindustrysectorclassiedbyGICS. SectorFirmsNBBOAskAskBidBidNBBOB-A(%)RevisionPriceSizePriceSizeDur.Spread Energy27641409913.6611443.66921-0.8430.0040ConsumerDiscretionary29662014353.689213.68779-0.5580.0041ConsumerStaples29560135503.577543.56731-0.5960.0063Financials29722075773.659223.64800-0.8240.0046HealthCare26597360693.5233473.522541-0.6580.0035Industrials30691075473.4713383.471294-0.8760.0052InformationTechnology29539022623.5911363.581012-0.5550.0070Materials26729455553.2516963.241685-1.5220.0035TelecommunicationServices25572879353.0899843.088714-1.0060.0050Utilities23489188173.5810273.58847-0.2410.0047

PAGE 87

TradeindexofintradaytradingITi= Tradeduration,timeintervalbetween(i1)standithtradesPi= Log-scaletradedpriceonithtradeAi;Bi= Themostrecentlog-scaleaskandbidpricesbeforetrade,followingalgorithmof Ellis,Michaely,andO'Hara ( 2000 )AiPi1= SpreadbetweenaskandlasttradedpricesPi1Bi= SpreadbetweenlasttradedandbidpricesDi= Tradedirection(ordinal)A+i;B+i= RevisedlogscaleaskandbidpricesrightaftertheithtradeSTi= TradesizeonithtradeNNi= NumberofNBBOrevisionsinatradedurationofITiNTi= NumberoftradessincethelasttradewiththesamedirectionasthatoftheithETi= Exchangecodeofithtrade(categorical)j= NBBOrevisionindexofintradaytradingINj= NBBOduration,timeintervalbetween(j1)standjthNBBOquotesAj;Bj= LogscaleaskandbidpricesofjthNBBOquoteNDj= DirectionofNBBOrevision(-1:bidpricechanged,0:indeterminate,1:askpricechanged)SNAj;SNBj= SizeofaskandbidquotesENAj;ENBj= Exchangecodeinaskandbidquotes(categorical) 87

PAGE 88

EvolutionofMeanandStandardDeviationofTradeDurationinEventWindows Figure3-5. EvolutionofMeanandStandardDeviationofRealizedVolatilityinEventWindows Ifweweretoperformastatisticalanalysisforeachrmandtradedatainthesamplewewouldneedtohandleamassivedatasetofintradaytradesandquotes2.Asisconventionalinanevent-study,weneedtorelyonsomeaggregation,withthegoalofndingsignicantresultsonthebasisof61daywindows(30dayspriortotheannouncementand30followingit). Figure 3-4 toFigure 3-9 showtheevolutionofaverageandstandarddeviationofselectedvariablessuchastradeduration,realizedvolatility,bid-askspread,bid-askspreadaftertrade,log-scaleprice,andtradesizeinaggregatedeventwindow.Realizedvolatility 3.2.1 has196,867,860rows,andatablecontainingconsolidatedquotedatahas620,461,738rows.

PAGE 89

EvolutionofMeanandStandardDeviationofBid-AskSpreadinEventWindows Figure3-7. EvolutionofMeanandStandardDeviationofBid-AskSpreadAfterTradeinEventWindows Figure3-8. EvolutionofMeanandStandardDeviationofLog-ScalePriceinEventWindows 89

PAGE 90

EvolutionofMeanandStandardDeviationofTradeSizeinEventWindows wherePiisthelog-priceontheithtrade. Foramodelhavingtradedurationlog(ITi)asresponsevariable,weproposethefollowingregression: log(ITi)=0+1Di+2STi+3Pi+4ETi+5(AiPi1)+6(Pi1Bi)+7log(ITi1)+8Di1+9STi1+10ETi1+11NNi+12NTi+i; whereiiswhitenoise. ForamodelhavingtradedirectionDiasresponsevariableweproposeanorderedlogitmodel.Thelatterisjustiedonthegroundsthatdecisionstotradeondirectionwhetheritisbuy,indeterminate,orsellwouldberankedaccordingtothetrader's 90

PAGE 91

whereDiisalatentobservationsatisfyingDi=8>>>>>><>>>>>>:1ifDi<1;0if1Di<2;1if2Di: Modelshavingotherresponsevariablessuchasspreadbetweenaskandlasttradedpricesaftertrade,A+iPi,spreadbetweenlasttradedpriceandbidpriceaftertrade, 91

PAGE 92

EvolutionofMeanandStandardDeviationofR2inModellog(ITi) FromTable 3-8 ,regressionmodelsPiorSTihaverelativelylowR2onaverage.Thereforetransactionpriceortradesizearenotlikelytobeexplainedbyanyregressor.Modelsforlog(ITi),A+iPi,andPiB+ihavehigherR2.Figure 3-10 to 3-15 illustratetheevolutionofR2inaggregatedeventwindows.Fromallgraphs,wecanseethatthemeanR2valuesareremarkablylowernearthekeyannouncementdatesthan 92

PAGE 93

SummaryofRegressions ThistableisasummaryofOLSandorderedlogitregressionresultsofmodelsspeciedinSection 3.2.1 .Nisthenumberofregressionsforeachmodel./RSSisaconstantstandarddeviationofi,exceptitisaresidualsumofsquaresforthemodelforDi.Forthelatter,R2iscalculatedbyMcFadden'spseudoR2method.DFrefertodegreesoffreedomforeachregression. ResponseVariablesN/RSSR2DF log(ITi)62865Mean1.7640.413108Std.Dev.0.4250.096168Di44084Mean410130.106326Std.Dev.410150.0812961A+iPi62647Mean0.0020.533112Std.Dev.0.0100.226175PiB+i62647Mean0.0020.543112Std.Dev.0.0120.226175Pi62865Mean0.0060.133108Std.Dev.0.0050.146168STi62865Mean13970.123108Std.Dev.31920.166168 Figure3-11. EvolutionofMeanandStandardDeviationofR2inModelDi EvolutionofMeanandStandardDeviationofR2inModelA+iPi

PAGE 94

EvolutionofMeanandStandardDeviationofR2inModelPiB+i EvolutionofMeanandStandardDeviationofR2inModelPi Regressorsineachmodelsmaybesignicantforsomermsandtradedaysbutnotalways.Inparticular,ifthesignicanceofsomeregressorsaredierentbetweenstableregimeandtransitionregime,theresultmightgivesomeinsightsintoaninformedtrader'sstrategy.InOLSregressions,wedenearegressorassignicantwhenthepvalueislessthan0.001andcollectthesignicantregressions.InorderedlogitregressionformodelDi, Figure3-15. EvolutionofMeanandStandardDeviationofR2inModelSTi

PAGE 95

Trade-relatedregressorsmayhavedistincteectsandtheirsignicanceratiosmayvary,dependingonwhethertheyapplytoastableregimeoratransitionregime.Asignicantratioisthefractionofregressionswithsignicantregressorsoverthetotalnumberofregressions.Ineachpanelofthesegures,theuppergraphisfortheaverageestimateofthecoecientandthelowerisfortheproportionofregressionsinwhichthecoecientissignicant(i.e.withp-value<0:001inOLSandwith3
PAGE 96

EvolutionofRegressorsinModellog(ITi)

PAGE 97

EvolutionofRegressorsinModelDi

PAGE 98

EvolutionofRegressorsinModelA+iPi

PAGE 99

EvolutionofRegressorsinModelPiB+i

PAGE 100

EvolutionofRegressorsinModelPi

PAGE 101

EvolutionofRegressorsinModelSTi

PAGE 102

log(INj)=0+1Bj+2Aj+3ENAj+4ENBj+5SNBj+6SNAj+7NDj+8Bj1+9Aj1+10ENBj1+11ENAj1+12SNBj1+13SNBj1+j: TheNBBOdirectionNDjindicatesthesourceofapricechangeinanNBBOrevision.IfNBBOisrevisedduetoanaskpricechange,thenNDj=1,andifitisduetoabidpricechange,thenNDj=1.Ifitisindeterminate,thenNDj=0.AnindeterminatecasemayincluderevisionswiththesamebidandaskpriceasinthelastNBBObutwithadierentsize,orrevisionswherebothbidandaskpriceshavechanged,orrevisionswiththesamebidandaskpricesbutwheretheaskorbidexchangehaschanged.ThevariableNDjhasimportantimplicationsinthatitshowsthedirectionofincorporationofpublicinformation,whethertheinformationispositiveornegative. AmodelthathasNBBOdirectionNDjasresponsevariablecanbespeciedasorderedlogitmodelsincethedecisionofNBBOdirectionwhetherask,indeterminate,orbidwouldberankedbypreference.ENAjandENBjimplynorankorpreferenceandarethusdenedascategoricaldata. whereNDjisalatentvariablesuchasNDj=8>>>>>><>>>>>>:1ifNDi<1;0if1NDi<2;1if2NDi;

PAGE 103

Modelshavingbid-askspreadAjBj,bidpriceBj,askpriceAj,andbidsizeSNBj,andasksizeSNAjasresponsevariablescanbespeciedasOLSregressionsasfollows: FromTable 3-9 ,regressionmodelswithlog(INj)haveverylowR2onaverage.ThedurationofanNBBOrevisionisnotlikelytobeaectedmuchbyanyothermarketinformation.ModelswithAjBj,Bj,orAjhavehigherR2thantheothers.ThehigherR2saremainlyduetothepersistencebetweenresponsevariablesandlaggedresponsevariablesinregressors.TheSNBjandSNAjmodelsalsohavepersistinglaggedresponse 103

PAGE 104

SummaryofRegressions ThetablesummarizesOLSandorderedlogitregressionresultsondatadescribedinSection 3.2.2 .Nisthenumberofregressions./RSSisaconstantstandarddeviationofj,exceptthatitisaresidualsumofsquaresforthemodelNDj.ForthemodelNDj,R2iscalculatedbyMcFadden'spseudoR2method.DFisthedegreesoffreedomofeachregression. ResponseVariablesN/RSSR2DF log(INj)64010Mean1.7820.076410Std.Dev.0.2820.077285NDj41489Mean1:21090.18113182Std.Dev.3:410100.06815001AjBj64010Mean0.0670.736411Std.Dev.0.1590.277285Bj64741Mean0.0280.886422Std.Dev.0.0790.207337Aj64741Mean0.0200.896422Std.Dev.0.0710.197337SNBj64741Mean7710.576422Std.Dev.252790.187337SNAj64741Mean7140.596422Std.Dev.12620.197337 Figure3-22. EvolutionofMeanandStandardDeviationofR2inModellog(INj) variables,whichimpliesthatthemajorityofthebidandasksizesarepassedontothenextNBBOduetothepricerevisionwithouttrade. Figures 3-22 to 3-28 showtheR2evolutiononaggregatedeventwindows.Fromallgraphs,wecanseethatthemeanR2valuesareremarkablylowerduringtransitionregime,implyingthatNBBOrevisionsaremoreaectedbyexogenousinformationduringtransitionregime. 104

PAGE 105

EvolutionofMeanandStandardDeviationofR2inModelNDj EvolutionofMeanandStandardDeviationofR2inModelAjBj EvolutionofMeanandStandardDeviationofR2inModelBj EvolutionofMeanandStandardDeviationofR2inModelAj

PAGE 106

EvolutionofMeanandStandardDeviationofR2inModelSNBj EvolutionofMeanandStandardDeviationofR2inModelSNAj 3.2.3 ,regressorsaresignicantwhentheirpvaluesarelessthan0.001inOLSandabsolutetvaluesarebetween3and100fororderedlogitregression.Weclassifyregressorsintotwocategories.OneconsistsofNBBOrevision-relatedregressorssuchaslog(INj),Bj,Aj,SNBj,SNAjandNDj.Theotherconsistsofexchange-relatedregressors,suchasENAj,andENBj. Figures 3-29 to 3-35 displaytheevolutionofregressorsinmodelsforlog(INj),AjBj,Bj,Aj,SNBj,andSNAj.Ineachpanelforaregressor,uppergraphrepresentstheevolutionoftheaverageestimateofthecoecientovertheeventwindow,andthelowergraphshowsthepercentageofregressionswherethiscoecientissignicant.AsinSection 3.2.3 ,exchange-relatedregressorsareaggregatedacrossallrmsandtradetimes.Table B-2 inAppendix B.2 reportscoecientestimatesofexchangefactors. 106

PAGE 107

EvolutionofRegressorsinModellog(INj)

PAGE 108

EvolutionofRegressorsinModelNDj

PAGE 109

EvolutionofRegressorsinModelAjBj

PAGE 110

EvolutionofRegressorsinModelBj

PAGE 111

EvolutionofRegressorsinModelAj

PAGE 112

EvolutionofRegressorsinModelSNBj

PAGE 113

EvolutionofRegressorsinModelSNAj

PAGE 114

3.1.3 ,amarketmakerhas3typesofbid-askspreadstrategies,whichcanbeevaluatedonthebasisofexplicitperformancemeasures.Thelattermayproducedierentoutcomes,dependingonwhetherthepricedynamicsevolveinastableregimeortransitionregime.Bystrictlyfollowingthesestrategies,characteristicsofpricedynamicscanbeverydierentfromthoseofrealpriceevolution.Thekeypointofthesimulationistoassesswhetherthesestrategiesresultinmoreecientpricedynamics. Sincepricedynamicsevolverandomly,wesimulate30samplepathsforeachrmin30minutestimeintervals.Althoughthelengthofthepricedynamicsisrathershort,thesimulatednumbersofNBBOrevisionsandtradesarelargeenoughtocapturemostofaspectsofpricedynamics.Thetimeintervalmayhavedistinctiveattributesdependingonwhetherwehaveastableregimeoratransitionregimeandthereforesimulatebothcases.Finally,wehavethreetypesofbid-askspreadstrategies,hencethetotalnumberofsimulationscomprisingallthosecasesis180foreachrm. Allestimablecoecientsinpricedynamics,namelycUB0,cUB1,cUI0,cUS0,cUS1,UB,UI,USarearchivedintwosets,oneforstableregimesandtheotherfortransitionregimes.Standarddeviationsofuninformedbuying,UB,indeterminate,UI,andselling,US,oftradedurations,arefromthesummarydatain 3-1 andin 3-3 .ProportionalcoecientscUB1andcUS1arederivedfromtheestimates^5and^6inregressionmodellog(ITi).InndingcUB0,cUI0,andcUS0,wecannotuse0ofmodellog(ITi)becausethisvaluealldefaultfactorsofexchangesETi,ETi1,andtradedirectionDiandDi1.HencewendthembyindirectwaysuchascUB0= log(ITi)B^5 log(ITi)I;cUS0= log(ITi)S^6 114

PAGE 115

3-10 summarizesthesimulationresultsbyregimeandbid-askspreadstrategy. 115

PAGE 116

Thetablereportsasummaryofsimulationresultsfor27rms,bidandaskpricestrategies,andforthetworegimes(\transi-tion"and\stable").InthecolumnPricingStrategy,`1'indicatesconstantbid-askspread,`2'Bayesian-Nashequilibriumstrategywithpartialmoments,and`3'Bayesian-Nashequilibriumstrategywithorderimbalance.Intherevisiontypecolumn,NBBOWOTmeansNBBOrevisionwithouttradebymarketmaker.Uninf.andInf.meanuninformedandinformedtrading.Meansandstandarddeviationsofrevisionduration,ask,bid,andbid-askspreadaremeasuredonlog-scale. RegimePricingRevisionNMeanStdev.MeanStdev.MeanStdev.MeanStdev.MeanStdev.StrategyTypeDur.Dur.AskAskPricePriceBidBidB-AB-A Stable1Inf.Buy105.029.042.121.022.121.022.121.020.00290.0012Inf.Sell165.485.433.521.663.521.663.521.670.00240.0013NBBOWOT.9168132.123.443.901.193.901.193.901.190.00090.0012Uninf.Buy1013701.472.453.841.033.841.033.841.030.00050.0006Uninf.Indet.256590.731.223.870.903.870.903.870.900.00040.0003Uninf.Sell887481.442.443.841.023.841.023.841.020.00050.0006 2Inf.Buy128.426.463.671.383.231.993.661.380.01030.0067Inf.Sell173.203.062.801.662.851.842.691.770.11290.1168NBBOWOT.8184632.253.514.101.293.931.123.771.200.33071.0954Uninf.Buy2322710.621.873.851.033.691.093.761.070.08200.2667Uninf.Indet.176000.941.555.671.184.270.522.860.922.80381.9039 Continuedonnextpage

PAGE 117

RegimePricingRevisionNMeanStdev.MeanStdev.MeanStdev.MeanStdev.MeanStdev.StrategyTypeDur.Dur.AskAskPricePriceBidBidB-AB-A Uninf.Sell2030130.631.903.171.123.331.153.151.140.01780.0415 3Inf.Buy184.805.512.790.792.740.862.770.790.02260.0526Inf.Sell132.952.213.631.663.931.763.601.690.03330.0404NBBOWOT.6965142.523.813.821.123.821.123.811.120.00430.0686Uninf.Buy2390240.541.763.571.373.431.383.551.380.02560.0733Uninf.Indet.50351.552.313.900.773.890.853.821.020.07740.6678Uninf.Sell2250610.491.682.961.313.071.392.931.340.02880.0774 Transition1Inf.Buy816.6530.043.852.353.842.353.842.350.00220.0010Inf.Sell1923.1642.013.791.913.781.913.781.910.00240.0009NBBOWOT.5403322.666.593.781.413.781.413.781.410.00130.0019Uninf.Buy1109993.439.543.821.073.821.073.811.070.00060.0007Uninf.Indet.421091.444.223.670.943.670.943.670.940.00030.0004Uninf.Sell1052523.288.493.771.063.771.063.771.060.00050.0006 2Inf.Buy1014.7918.194.711.343.361.994.711.340.00860.0106Inf.Sell915.8528.263.980.954.451.673.970.960.01090.0174NBBOWOT.3721222.916.884.191.534.101.474.001.500.18750.7345 Continuedonnextpage

PAGE 118

RegimePricingRevisionNMeanStdev.MeanStdev.MeanStdev.MeanStdev.MeanStdev.StrategyTypeDur.Dur.AskAskPricePriceBidBidB-AB-A Uninf.Buy940982.498.284.581.394.391.313.971.300.60561.4170Uninf.Indet.226922.946.875.881.104.330.612.700.833.18011.7121Uninf.Sell850022.427.853.971.233.681.213.231.180.73871.3417 3Inf.Buy343.2223.114.222.093.582.404.182.050.03830.0464Inf.Sell515.4319.101.540.681.510.911.480.710.05600.0457NBBOWOT.1539942.998.043.641.333.601.333.571.360.06740.3820Uninf.Buy312173.3910.413.701.033.671.033.294.690.40964.7738Uninf.Indet.32313.3718.203.790.843.750.863.254.760.54494.7460Uninf.Sell345692.959.483.939.583.600.913.570.940.35319.7082

PAGE 119

FromTable 3-11 ,wenoticedierencesinperformancemeasuresbetweenthetworegimes.Inastableregime,wherenokeydevelopmentisannounced,theBayesian-NashequilibriumstrategywithorderimbalancedominatesintermsofmeanME(T)andPM(T).Theconstantbid-askstrategyhasabetterperformanceinSE(T).Inthetransitionregime,theconstantbid-askspreadstrategyuniformlyprevailsovertheothertwoBayesianNash-equilibriumstrategies.Welchtwo-samplet-testsareperformedtondstatisticallysignicantdierencesbetweenthe3setsofperformancemeasuresinpricingstrategiesatthe95%condencelevel.p-valuesofallpairwisetestsfailtorejectthenullhypothesisthatthedierencebetweenthemeansofthetwosamplesarezero.Thereforeinthissimulation,wecannotconcludethatoneofthethreebid-askpricingstrategiesispreferabletotheother.However,wewillneedtohavemorepreciseparametersetuptorenethoseresults. 119

PAGE 120

SummaryofPerformanceofPricingStrategies Thetablereportsperformancemeasures,ME(T),SE(T),andPM(T)fromsimulationresultsfor27rms,bidandaskpricestrategies,andregimes.Inthecolumnofpricestrategy,`1'indicatesconstantbid-askspread,`2'Bayesian-Nashequilibriumstrategywithpartialmoments,and`3'Bayesian-Nashequilibriumstrategywithorderimbalance.Thelast4columnscontaintheWelchtwo-samplet-testresults.Indexescolumnshowstheindexesoftwosamplesofpricingstrategies. RegimePricingRealizedp-valuesofWelch2samplet-test StrategyReturnVolatilityME(T)SE(T)PM(T)IndexesME(T)SE(T)PM(T) Stable1Mean0.051.011.930.35-0.07(1,2)0.871:181075:191033Std.Dev.1.110.491.970.230.862Mean-0.027.341.920.4264.12(2,3)0.762:011050.61Std.Dev.0.847.91.920.33145.793Mean05.541.890.3659.14(1,3)0.640.483:061012Std.Dev.0.718.631.930.29237.73 Transition1Mean01.141.850.39-0.07(1,2)0.029:7310226:601045Std.Dev.1.430.761.530.30.822Mean-0.049.522.020.6244.76(2,3)0.0046:021083:171042Std.Dev.1.2811.541.440.685.163Mean-0.032.751.810.481.5(1,3)0.562:641061:301034Std.Dev.0.332.331.450.443.37

PAGE 121

BallandTorous ( 1988 ).However,knowingthecontentofnewscangiveuscausalitytonewdirectionandvolatility.Itcanhelpclassifytransitionregimesmoreclearlythanuncertainevent-datemodelduetothefactthatthesetupofuncertaineventdatealsomakesusassumeuncertaintransitionregimesimilarto PastorandStambaugh ( 2001 ).Thelatterrestrictsourabilitytodeterminemultipletransitionregimesineventperiodsorperformadetailedanalysisoftheregimecontainingtheeventdate.Finallysincethesamenewscanbeinterpretedinmanyways,whichmaybereectedinchangesofdirectionandvolatilityleadingtonewregimes. Asfurtherstudywecanextendouranalysistoliquiditychanges.Inrelationtothekeydevelopmentannouncement,onecananalyzeintradaydataineventperiods,inamannersimilartoobservingtheadverseselectioncomponentinbid-askspreadsasin KrinskyandLee ( 1996 ),orobservingregimechangesinrealizedvolatilityasin LiuandMaheu ( 2008 ). InChapter3,themethodologysimulatesmarketquoteandtraderevisions,closelyimitatingrealmarketpricedynamics.Sincethepricedynamicsispath-dependentaswellastime-dependent,wedonotdevelopaclosedformorparametricmodelofpricedynamics.ClosedformpricedynamicsmaybedevelopedbyinvestigatingconvergencetogeometricBrownianmotionasinchapter3of Merton ( 1990 )undersomeconditionsofconvergence. 121

PAGE 122

Themainreasonwedonotapplyspreaddecompositionmodelsisthatthemodelsfocusondecomposingbid-askspreadintoseveralcomponentstondevolutionofadverseselectioncomponentovertime,butdonotfocusontherelationshipbetweentradedurationandbid-askspread.Combinationofspreaddecompositionmodelsespecially Madhavan,Richardson,andRoomans ( 1997b )and Easley ( 1996 )wouldgivemoreinsight,howeverbothmodelsdoesnotconsidertradeandquotedurationsboth. AlgorithmforextractingNBBOfromconsolidatedquotedatashouldbemoreprecise.MostofempiricalpapersanalyzingquotedatausesPerlscriptprovidedbyWRDStoextractNBBOusesonlyforNYSElistedquote,notforNASDAQ,AMEXandothers.Werunthescriptforalldatawearchived,hencetheremightbesomeissuesasmentionedbyauthorsofthepapers. Insimulation,theperformanceofsimulationisanissue.In30minutesofsimulation,thesimulationtakesconsiderabletimetogenerateallquoteandtradedata.WemayneedtoreprogramwithC++toimproveperformance,ordevelopingapproximationmethod. Anotherissueisthatvolatilityinthisintervalishugefor30minutesoftrading.Inarealdata,wecouldidentifythattransactionpricerevolutiondoesnotfollowthelatestbidoraskpricesnecessarily,norfollowingbidoraskpricesdiscountedbyoperatingcost 122

PAGE 123

Insimulationresults,wecanidentifyexplodingnumberofinformedtrading,becauseofabiggapbetweenwI(t)andw(0).Wecalculatedw(0)fromhistoricalaverageofw(t)inentiretimehorizon,henceitisquitedierentfromevolutionofwI(t).Couplingw(0)equalswI(0)insteadofaveragepricewouldgivemoreprecisecomparisonamongbidaskpricestrategies. InChapter3,weachievedtheobjectives. First,webuildastructuredmethodologyofbidandaskpricedynamicsofmarketmakerbasedonsequentialtradingmodelsundercontinuoustimeframework,directlyreplicatingintradaytradingandquotedata.Themethodologyconsistsoffundamentalvalueprocess,durationprocess,andNBBOrevisionprocess.InNBBOrevisionprocess,wesuggested3possiblebid-askspreadpricingstrategies. Second,weperformempiricalstudyinvestigatingthesignicantdeterminantsaectingdecisionsoftradesandNBBOrevisions.Theempiricalstudyconsistsofvarious 123

PAGE 124

Finally,wesimulatedatawiththemethodologybasedonempiricaldata.Theresultsgivesdierentpreferencesonbidandaskpricingstrategiesbetweeninstableregimeandtransitionregime,howeverthedistinctionisnotstatisticallysignicant. 124

PAGE 125

3{18 ) wherep(D=1jw)=8>><>>:p(D=1jw>0)=DBdt+UBdt;p(D=1jw0)=UBdt: A{1 )andorganizeit,then Foranyoftworealnumberst,s,weknowlimt!0lims!0t s=1; 3{6 ),E(wI(t))=wI(0)+E0@I(t)Xi=1SiTIi1A+E0@Si0@tI(t)Xi=1TIi1A1A:

PAGE 126

Ross ( 1996 ),wecanndE0@I(t)Xi=1SiTIi1A=E0@I(t)+1Xi=1SiTIi1AESI(t)+1TII(t)+1=(I+1)S PluggingEq.( 3{20 )inDBandEq.( A{3 )inE(w),wehave 3{19 ) A{5 )andorganizeit,then PluggingEq.( 3{20 )inDSthenwehave

PAGE 127

3{21 ),( 3{22 ) 3{21 )isidenticaltoderivingEq.( 3{18 ),exceptplugginginDB(t):=cDBI+(t)insteadofDB(t):=cDB(wI(t)w(t))+.AlsoderivationofEq.( 3{22 )isthesameasthatofEq.( 3{19 ),exceptplugginginDS(t):=cDS(wI(t)w(t))insteadofDS(t):=cDSI(t). 127

PAGE 128

Thetablerecordsallcoecientestimatesofexchangefactors.Exchangetypehavefourpossiblevalues.`E':exchangewherethetradeoccurredfromETi,`L':exchangecodewherethelasttradeoccurredfromETi1,`A':exchangecodewheretheaskquotehasbeenpostedfromENAi,and`B':exchangecodewherethebidquotehasbeenpostedfromENBi.Signicantcountrepresentsthenumberofregressionsthathastheexchangefactorwithpvalueislessthan0.01atOLSorabsolutevalueoftvalueisbetween3and100atorderedlogit.Signicantratioistheratiobetweensignicantcountsandallcounts. ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. log(ITi) EA13%-4.12411.5609B89%-2.34001.98800.71820.6211C142519%-1.02611.70740.29940.3288D324020%-0.25163.44160.25930.2648I43%-4.50042.42711.41060.8212M14858%0.30523.73170.50780.8908N694923%-3.6731360.96850.15620.2387P1298825%-3.2350273.52430.25380.3370Q413130%16.62061157.01670.29271.4207T397316%0.096316.11970.22130.3363X554%-5.748311.55832.11311.7017 LA13%-5.56211.4111B1113%2.094610.67071.11270.8293C180624%-0.97051.23930.27860.2711 Continuedonnextpage

PAGE 129

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. D317720%-0.12578.88820.25880.2919I43%-2.05902.75510.85540.7157M180010%0.72882.34930.44350.5068N611320%-0.131719.49270.16990.1952P1505828%-1.085116.88340.27602.7425Q494236%-1.18754.53310.27920.3107T390416%-0.492118.25850.31155.3704X544%-0.692310.51892.20362.0857 LB33.5%0.6261.2350.3280.150C2773.7%0.1631.5200.3220.250D5413.4%0.0031.2230.2680.177I54.2%-2.8862.6320.8940.313M1831.0%0.0341.8860.4340.298N13924.6%-0.1480.7740.1670.122 Continuedonnextpage

PAGE 130

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. P15743.0%-0.0401.9780.2470.201Q6254.6%-0.0420.9640.2040.152T7483.0%0.4053.0430.2180.220 LA25%-0.00130.00080.00030.0001B67%-0.00090.00370.00040.0001C4977%0.00390.08920.00070.0105D7965%0.00050.01610.00050.0040I43%-0.00030.00170.00030.0001M16569%-0.00170.05810.00050.0072N15965%-0.00030.00370.00020.0006P30116%0.00050.02920.00050.0037Q9457%0.00020.04700.00110.0058T12355%0.02290.93610.00050.0045 Continuedonnextpage

PAGE 131

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. X12810%-0.00040.00570.00050.0012 AC189054%-0.330718.02600.06081.2508D243724%-0.12260.85230.00090.0122M201020%-0.00800.18070.00090.0145N939032%-0.00030.00500.00010.0004P1639427%-0.04846.13730.00750.4252T914325%-0.08918.20750.01340.5694X119184%0.00410.00430.00050.0005 BC108729%0.253412.20780.07431.1547D151015%-0.07780.58570.00080.0073M118913%0.00360.11860.00060.0089N338512%0.00110.05270.00010.0019P955815%0.02494.14600.00870.3900T591116%0.03925.27140.01390.4958X25019%-0.00010.00350.00030.0003 BC166445%1.8106101.72480.08632.5736D249525%-0.07171.41670.00250.0825 Continuedonnextpage

PAGE 132

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. M199121%-0.01320.30340.00040.0070N905931%-0.00110.07450.00010.0013P1606626%0.182232.73450.00920.8285T863624%0.340144.64820.01701.1300X115386%0.00420.01080.00060.0035 EA1541%-0.00020.00170.00030.0001B1416%0.00220.00410.00050.0006C77710%0.00010.00410.00020.0007D14019%0.03071.12640.00060.0068I87%-0.00060.00350.00020.0001M272815%-0.00030.00720.00030.0013N371412%-0.00020.01440.00020.0028P598011%0.00000.02950.00050.0048Q172513%-0.00630.26610.00080.0051T267511%0.00130.05710.00060.0123X28422%-0.00030.00510.00040.0005 LA38%0.00120.00090.00030.0001B56%0.00060.00380.00040.0001C5227%-0.00210.05100.00070.0112D8545%0.00320.04770.00110.0108I33%0.00000.00240.00040.0000M179010%-0.00240.09730.00050.0057N16706%-0.00030.01020.00030.0031P30806%-0.00020.02280.00050.0044Q9267%0.00010.03270.00100.0064 Continuedonnextpage

PAGE 133

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. T12835%0.00360.12240.00060.0049X13410%0.00090.01490.00070.0023 LA25%-0.01300.01010.00440.0035B1011%0.00250.01010.00250.0015C108615%-0.00320.11150.00190.0211D209713%0.00060.04840.00150.0091I98%0.00210.01690.00340.0023M178110%-0.00550.27000.00340.0663N466115%-0.00010.00740.00070.0011P737414%-0.00010.01400.00100.0029Q182613%0.00070.02060.00120.0052T369615%-0.00020.01270.00090.0015 Continuedonnextpage

PAGE 134

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. X464%0.02690.08190.00990.0158 LA13%408171525B45%5458710732847356285C2483%218151146771262D4963%-13010940445881I22%4462612715822173M8835%109208188991677N13765%-6666585266672P22224%-2842126066267575Q8376%-30749885234135311T7243%-255739565352672X14111%-105132868427323551

PAGE 135

Thetabledisplaysallcoecientestimatesofexchangefactors.Exchangetypetakesfourpossiblevalues.`A':exchangecodewheretheaskNBBOhasbeenpostedfromENAj,`B':exchangecodewherethebidNBBOhasbeenpostedfromENBj,`LA':exchangecodewherethelastbidNBBOoccurredfromENAj1,and`LB':exchangecodewherethelastaskNBBOoccurredfromENBj1. ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. log(INj) AC167225%6.592.22.923.3D158514%-1.559.11.717.7M4474%1.22.50.60.6N982633%0.41.40.10.4P1617725%2.6920.33.8262.3T906226%-670.065406.111.8698.3X1406%-1.66.71.20.8 BB4879%-1691.92650.5228.9152.6C218031%181.8728.580.2114.7D202218%58.4433.015.660.7M4765%10.7203.16.642.2N977233%0.030.40.25.5P1803628%1113.0146871.312.7250.4T1055527%-327.239585.123.0109.5X1397%2.247.12.516.3 LAC69618%1.769.11.415.2D137512%0.849.11.114.8M3544%3.040.81.415.3 Continuedonnextpage

PAGE 136

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. N1140939%0.57.30.12.7P1526524%-5.4830.42.8265.5T897524%-23.02150.99.2690.9X1526%-4.520.91.44.9 LBB4767%-1582.82591.7219.4165.1C76719%211.4697.051.8116.1D157814%13.7161.05.038.1M3494%15.1142.14.133.6N1179840%0.43.90.10.5P1663126%9.1163.22.829.6T854822%20.4352.47.374.5X1517%-7.258.12.115.1 BC3148.5%-43.894459.49926.97498.443D150214.4%-11.575203.9434.50742.316M5185.8%-1.2150.6850.3210.190N550619.4%-0.3010.6990.0860.162P1141719.6%-0.34991.6681.17320.930 Continuedonnextpage

PAGE 137

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. T575416.6%25.620560.06810.171133.778X50.3%4.5723.6621.1360.773 LAC2487.7%4.37160.3750.8217.453D140813.6%-1.16843.5310.74210.194M6486.8%7.085144.0691.78536.099N626322.0%1.16148.3250.28711.932P1078518.5%0.19859.4770.43412.282T605917.6%6.865286.8493.33685.143X30.1%-2.9630.8680.7690.393 LBC2848.6%10.238112.3543.17626.763D174316.7%3.09889.4481.13221.279M6507.3%-1.3130.6250.3320.184N695524.5%-0.4130.4800.0940.106P1258821.6%-0.61274.5970.37514.586T576216.6%-27.164524.0167.374125.853X30.1%3.8153.4381.0650.974 BB61100%13.4130.2930.0030.010 Continuedonnextpage

PAGE 138

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. C451864%-9.8965.4450.0800.309D352931%-1.8034.8010.0430.231M8749%-0.2922.0070.0030.024N683423%0.0100.6170.0110.064P2245635%-2.51854.3650.0360.777T1520539%-3.0696.8370.0410.322X78841%0.0041.0530.0090.095 LAC70318%0.1942.2930.0210.159D284526%0.3171.6520.0370.180M94310%-0.0040.1900.0040.021N784227%-0.0810.4650.0150.068P1617426%-0.53664.0920.0260.891T958325%-1.522166.7580.0542.353X73531%-0.1340.8070.0260.159 LBB70100%-9.5975.4250.0810.312C121730%5.6169.2750.1740.624D310228%0.5752.3440.0410.176M8189%0.0901.0690.0040.032N773826%-0.0910.6090.0160.103P1706027%0.3753.0480.0310.214T1068628%0.6506.6360.0450.455X72333%-0.1541.4440.0390.275 Continuedonnextpage

PAGE 139

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. M6056%0.0230.3550.0030.015N433715%-0.0500.8540.0110.124P1331521%-0.0472.3340.0200.446T871324%0.1064.6000.0220.450X46121%-0.0390.6510.0150.085 BB101100%-5.6016.1110.0390.036C494367%2.45711.5150.1670.605D408736%0.7943.3420.0400.161M177419%0.0122.0610.0020.035N946732%-0.0150.5660.0060.074P2539340%0.5125.3600.0430.303T1828046%0.6906.3770.0550.348X97549%-0.0410.8890.0120.111 LAC84521%0.1836.4020.0140.236D182516%-0.1331.6630.0320.281M5015%-0.0190.3390.0030.014N515517%0.0460.3990.0080.047P1179018%0.0410.7390.0110.073T827722%-0.0851.9970.0110.136X41517%0.1130.6590.0200.096 LBB8196%3.5146.7820.0450.046C178542%-1.2366.7100.0790.338D375334%-0.1651.7430.0220.153M169518%-0.0220.4050.0010.012N899930%0.0300.4570.0060.055 Continuedonnextpage

PAGE 140

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. P2049832%-0.0952.1360.0150.130T1502638%-0.1144.4610.0230.604X93042%0.0591.1210.0190.156 BB10099%4.1644.0970.0410.039C447561%-1.73211.2710.2011.021D233621%-1.0704.7960.0580.362M5135%-0.2201.2040.0040.021N327411%-0.0220.5350.0070.077P1414122%0.187175.7480.0901.123T1240331%0.987176.9600.0940.791X42021%-0.2474.5230.0220.231 LAC144236%0.0150.6460.0050.036D345131%-0.1086.5830.0220.630M189519%0.0030.1170.0010.018N880630%0.0020.1580.0020.024P2007031%-0.38356.8990.0100.708T1516639%-1.048130.7890.0191.629 Continuedonnextpage

PAGE 141

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. X103442%0.0040.1330.0020.019 LBB8298%-2.8684.5650.0390.037C135232%1.25310.5290.1370.756D182816%0.3513.7730.0350.331M4595%0.0620.5720.0020.010N396713%0.0000.3240.0040.036P1006016%-0.19630.1350.1146.897T836421%0.1327.7050.0350.374X36417%0.2504.6110.0270.246 BB8685%-868623411443480172176109857C286539%557415566079034980199143D513145%201876416804410238161149M313433%303811381114197572517N2109071%222465477825P3620956%440841600378309958204T2313858%640782007869552376725 Continuedonnextpage

PAGE 142

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. X43822%-4586552516514204 LAC3218%444523696363432D6786%387488552132511216M4535%5699476591407N16366%-26059781701258P36306%229710270889313605T26227%-47360521593587351262X964%1835160936392482 LBB6881%828060010892217162934114048C137233%-28332528532151564871708D419938%-410031552101255682373M250727%-17166760643215082253N1593654%-145341175738P2971446%-1268261927788017364T1814546%-15240804912229437805X30314%5938668418204963 BB9594%-64658479827280180304141196 Continuedonnextpage

PAGE 143

ExchangeExchangeSignicantSignicantMeanStd.Dev.MeanStd.Dev.TypeCodeCountRatioEstimateEstimateStd.Err.Std.Err. C194826%869770725739252295262160D10259%1068600985854953659356650M4355%30121307110299915227N307110%574783100262P53268%300376447721721977203420T511113%322536455050123627208490X985%81884856813807050 LAC129232%-1380139673962338D419938%1238273084082717M289329%-7786584429700N1547152%-195290088282P2910246%9708062426914497T1792347%218689944199045X34514%6197922574902689 LBB7286%662594510431994164696137985C63815%-629818513812638897261777D8087%-239285435630012572229038M4345%-49806511113438702N15575%-1712738166384P38216%-10476721112636833108096T29147%-143574242258810828130081X1226%-61224216311036121

PAGE 144

2 144

PAGE 145

Firsttwocolumnsindicatetickersymbolofrmsandchange-pointmodelhavingrregimesthatwewanttocompare.Nextfourcolumnsshows lnBrs>2withcomparedmodelhavingsregimes.Thenextfourcolumnsshows95%uppercondencelimitduetotheone-sidedttest.Thelastfourcolumnsshowsthepvaluesdeterminingnullhypothesisisbeingrejected. ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 AEP10.0017.4336.8620.25NA37.4474.7544.420.000.900.940.90AEP2-17.430.0019.432.832.59NA63.9930.760.050.000.740.52AEP3-36.86-19.430.00-16.611.0325.12NA31.000.050.210.000.26AEP4-20.25-2.8316.610.003.9225.1164.22NA0.060.390.700.00 BIIB10.009.18-1.837.64NA51.1013.6531.710.000.610.340.65BIIB2-9.180.00-11.01-1.5432.73NA31.6648.990.330.000.300.45BIIB31.8311.010.009.4717.3153.69NA34.600.490.640.000.69 Continuedonnextpage

PAGE 146

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 BSX10.0011.59-3.5232.32NA37.677.2477.400.000.730.200.87BSX2-11.590.00-15.1120.7314.49NA12.5271.680.190.000.150.73BSX33.5215.110.0035.8514.2842.75NA81.820.590.790.000.89BSX4-32.32-20.73-35.850.0012.7630.2210.13NA0.100.230.090.00 CAT10.0022.49-13.4839.02NA53.7422.5683.810.000.860.240.91CAT2-22.490.00-35.9716.538.77NA-9.6567.120.100.000.010.69CAT313.4835.970.0052.5049.5262.28NA107.680.700.980.000.93CAT4-39.02-16.53-52.500.005.7734.062.68NA0.070.270.050.00 CELG10.00-1.82-1.79-23.25NA-0.5120.82-6.500.000.000.390.01CELG21.820.000.02-21.433.12NA22.37-4.620.410.000.440.01CELG31.79-0.020.00-21.4524.4122.32NA3.900.490.440.000.06CELG423.2521.4321.450.0040.0038.2446.80NA0.980.970.900.00 EGN10.0026.3055.1483.04NA59.2697.81148.280.000.890.980.98EGN2-26.300.0028.8456.746.66NA58.47115.760.080.000.930.94 Continuedonnextpage

PAGE 147

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 FDX10.002.20-2.8461.01NA17.8313.86106.920.000.510.310.98FDX2-2.200.00-5.0458.8113.43NA6.32103.650.330.000.150.98FDX32.845.040.0063.8519.5416.39NA112.300.530.670.000.98FDX4-61.01-58.81-63.850.00-15.10-13.97-15.39NA0.010.010.010.00 FL10.0016.0416.96111.55NA41.4845.29172.440.000.820.811.00FL2-16.040.000.9295.519.41NA25.59148.680.120.000.471.00FL3-16.96-0.920.0094.5911.3723.74NA153.400.130.420.000.99FL4-111.55-95.51-94.590.00-50.66-42.34-35.77NA0.000.000.000.00 GCO10.0020.8932.469.58NA47.0772.6938.450.000.880.900.67GCO2-20.890.0011.57-11.325.29NA48.8011.100.070.000.670.16GCO3-32.46-11.570.00-22.897.7625.65NA15.430.080.270.000.14GCO4-9.5811.3222.890.0019.3033.7361.21NA0.250.760.820.00 JCP10.0030.6328.7112.44NA67.5957.2230.920.000.900.940.83 Continuedonnextpage

PAGE 148

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 KFT10.00-2.8714.8714.06NA-0.1942.6944.900.000.000.780.74KFT22.870.0017.7316.935.54NA45.0947.150.710.000.830.80KFT3-14.87-17.730.00-0.8012.969.62NA41.820.160.110.000.46KFT4-14.06-16.930.800.0016.7813.2943.42NA0.190.150.480.00 MGM10.0015.8413.8146.57NA44.5230.9098.120.000.790.880.92MGM2-15.840.00-2.0330.7212.83NA32.1191.700.150.000.420.78MGM3-13.812.030.0032.763.2836.18NA87.100.060.500.000.83MGM4-46.57-30.72-32.760.004.9930.2521.58NA0.060.180.140.00 MON10.0054.188.7966.35NA99.8429.10122.650.000.970.710.97MON2-54.180.00-45.3912.16-8.52NA5.8181.680.020.000.060.60MON3-8.7945.390.0057.5611.5296.59NA120.410.190.920.000.93MON4-66.35-12.16-57.560.00-10.0457.355.30NA0.020.370.060.00 Continuedonnextpage

PAGE 149

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 PG10.0011.8137.7239.99NA36.9972.7273.980.000.740.950.97PG2-11.810.0025.9128.1813.37NA52.1972.050.180.000.930.84PG3-37.72-25.910.002.27-2.710.38NA44.450.030.040.000.50PG4-39.99-28.18-2.270.00-6.0015.7039.91NA0.020.130.430.00 PLCE10.002.1526.1274.97NA10.5858.41121.670.000.510.890.99PLCE2-2.150.0023.9772.826.28NA58.04119.440.210.000.860.99PLCE3-26.12-23.970.0048.856.1710.10NA106.870.070.100.000.91PLCE4-74.97-72.82-48.850.00-28.27-26.209.17NA0.000.010.070.00 PTRY10.0011.4314.69-0.68NA30.7137.948.370.000.790.820.31PTRY2-11.430.003.26-12.127.85NA19.0711.800.120.000.550.16PTRY3-14.69-3.260.00-15.388.5512.55NA12.150.120.290.000.15 Continuedonnextpage

PAGE 150

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 QCOM10.003.7927.0028.14NA18.0857.4665.980.000.580.910.87QCOM2-3.790.0023.2224.3610.51NA49.8065.030.250.000.910.82QCOM3-27.00-23.220.001.143.453.36NA37.850.060.060.000.48QCOM4-28.14-24.36-1.140.009.7016.3135.58NA0.090.140.440.00 S10.0013.291.3134.61NA36.6017.7784.510.000.790.470.86S2-13.290.00-11.9821.3110.02NA16.4277.300.140.000.200.72S3-1.3111.980.0033.2915.1440.38NA82.400.370.720.000.86S4-34.61-21.31-33.290.0015.3034.6715.82NA0.110.240.120.00 T10.0029.0431.9930.31NA73.2075.5561.140.000.850.870.94T2-29.040.002.961.2715.12NA8.5254.580.120.000.610.49T3-31.99-2.960.00-1.6811.562.61NA51.520.100.070.000.45T4-30.31-1.271.680.000.5252.0454.89NA0.040.460.500.00 TMO10.000.40-2.94-19.88NA8.9113.5912.510.000.380.310.13TMO2-0.400.00-3.33-20.288.12NA14.9212.470.320.000.310.13 Continuedonnextpage

PAGE 151

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4 TSN10.009.7414.8027.11NA45.3945.1459.940.000.640.760.90TSN2-9.740.005.0617.3725.91NA23.8166.650.290.000.610.70TSN3-14.80-5.060.0012.3115.5413.69NA55.630.180.260.000.66TSN4-27.11-17.37-12.310.005.7131.9031.00NA0.070.250.290.00 UNH10.00-1.881.5134.26NA0.438.6969.080.000.000.450.94UNH21.880.003.3936.154.20NA10.4771.230.470.000.630.95UNH3-1.51-3.390.0032.755.673.68NA66.570.210.100.000.93UNH4-34.26-36.15-32.750.000.55-1.061.07NA0.040.040.050.00 VZ10.0010.62-14.6713.69NA24.221.2034.510.000.850.040.83VZ2-10.620.00-25.293.072.97NA-5.0515.130.060.000.010.56VZ314.6725.290.0028.3630.5445.54NA53.340.910.970.000.96VZ4-13.69-3.07-28.360.007.139.00-3.38NA0.110.240.020.00 WYE10.00-2.117.4614.52NA0.2223.5034.050.000.000.720.86 Continuedonnextpage

PAGE 152

ComparedModelUpperCondenceLimitpvalue TickerModelM1M2M3M4M1M2M3M4M1M2M3M4

PAGE 153

Firsttwocolumnsindicatetickersymbolsandchange-pointmodelhavingrregimes.Thenextfourcolumnsoftherstrowshowstheaveragelengthofregimes,^mssinaneventperiodof61days.Thesecondrowshows95%condencelimitoftwo-sidedttest. TickerMr^m11^m22^m33^m44 AEP223.1637.84(18.82,27.49)(33.51,42.18)AEP318.2513.1829.57(14.5,22)(11.25,15.1)(25.97,33.17)AEP414.3211.6811.5223.48(12.35,16.29)(10.57,12.78)(10.62,12.42)(20.8,26.17) BIIB218.9842.02(15.39,22.58)(38.42,45.61)BIIB316.0711.7333.2(12.47,19.68)(9.32,14.14)(29.47,36.93)BIIB414.299.3810.7226.61(10.47,18.11)(7.18,11.57)(8.92,12.52)(22.7,30.52) BSX216.8344.17(14.42,19.24)(41.76,46.58) Continuedonnextpage 153

PAGE 154

TickerMr^m11^m22^m33^m44 CAT221.4539.55(18.91,24)(37,42.09)CAT319.7611.5529.7(17,22.52)(9.09,14)(26.68,32.71)CAT414.1711.9510.6224.27(11.98,16.35)(10.42,13.48)(8.4,12.83)(21.24,27.29) CELG228.2732.73(23.93,32.61)(28.39,37.07)CELG322.7513.724.55(19.18,26.32)(10.44,16.96)(20.35,28.75)CELG417.5811.2111.7820.43(14.37,20.79)(9.43,12.99)(9.09,14.47)(16.37,24.49) EGN216.9644.04(14.92,19)(42,46.08)EGN314.2113.3833.4(12.54,15.88)(12.38,14.38)(31.5,35.31)EGN412.8111.7911.0625.34(11.27,14.36)(10.7,12.87)(10.37,11.76)(23.45,27.22) FDX219.1741.83(16.08,22.27)(38.73,44.92) Continuedonnextpage 154

PAGE 155

TickerMr^m11^m22^m33^m44 FL219.3541.65(16.6,22.09)(38.91,44.4)FL316.6811.9332.39(13.8,19.56)(10.23,13.63)(29.8,34.99)FL413.9311.410.8124.86(11.93,15.93)(10.25,12.55)(9.71,11.91)(22.88,26.84) GCO220.2940.71(16.14,24.43)(36.57,44.86)GCO321.229.7929.99(17.43,25.01)(8.05,11.53)(26.08,33.9)GCO415.4612.169.4823.91(13.25,17.66)(10.79,13.52)(7.62,11.34)(20.74,27.07) JCP217.8543.15(15.01,20.69)(40.31,45.99)JCP315.1212.733.18(12.99,17.24)(11.63,13.77)(30.37,36)JCP412.6811.910.6325.79(10.79,14.57)(10.7,13.1)(9.82,11.43)(23,28.59) KFT220.6240.38(16.75,24.48)(36.52,44.25) Continuedonnextpage 155

PAGE 156

TickerMr^m11^m22^m33^m44 MGM216.9244.08(14.39,19.44)(41.56,46.61)MGM315.6112.1533.24(12.97,18.25)(10.87,13.43)(30.87,35.6)MGM412.9611.7410.6425.66(11.26,14.66)(10.6,12.88)(9.58,11.71)(23.71,27.61) MON21942(16.37,21.64)(39.36,44.63)MON317.5811.8631.56(14.61,20.55)(10.37,13.36)(28.33,34.79)MON414.8312.169.6424.36(12.59,17.07)(10.84,13.49)(8.73,10.56)(21.45,27.28) PDX223.4937.51(19.78,27.21)(33.79,41.22)PDX320.610.3730.03(16.97,24.23)(8.72,12.01)(26.51,33.55)PDX414.9411.699.3825(12.7,17.18)(9.87,13.5)(8.38,10.37)(21.83,28.17) PG214.5946.41(12.66,16.52)(44.48,48.34) Continuedonnextpage 156

PAGE 157

TickerMr^m11^m22^m33^m44 PLCE218.8242.18(16.19,21.45)(39.55,44.81)PLCE317.611.3732.03(14.6,20.6)(9.4,13.35)(29.08,34.98)PLCE413.712.510.6224.18(11.82,15.57)(10.58,14.41)(9.13,12.11)(21.68,26.69) PTRY220.8640.14(17.99,23.72)(37.28,43.01)PTRY317.7212.6130.67(14.8,20.63)(10.82,14.41)(28.09,33.25)PTRY414.0712.719.8724.36(12.41,15.72)(11.03,14.38)(8.77,10.97)(21.72,26.99) QCOM222.2238.78(17.9,26.54)(34.46,43.1)QCOM319.210.1331.67(14.89,23.51)(8.49,11.76)(27.73,35.61)QCOM413.5511.159.9826.32(11.37,15.72)(9.42,12.88)(8.69,11.28)(22.99,29.66) S221.8939.11(18.02,25.77)(35.23,42.98) Continuedonnextpage 157

PAGE 158

TickerMr^m11^m22^m33^m44 T218.6442.36(16.76,20.52)(40.48,44.24)T316.9212.7531.33(14.44,19.4)(11.23,14.27)(28.3,34.36)T413.5512.1410.9524.35(11.84,15.27)(10.86,13.43)(9.94,11.97)(21.62,27.08) TMO222.2538.75(19.84,24.66)(36.34,41.16)TMO318.5512.4629.99(15.68,21.42)(11.06,13.85)(27.25,32.74)TMO415.0711.8910.0524(12.76,17.38)(10.6,13.17)(8.94,11.15)(21.79,26.2) TSN220.1140.89(17.3,22.92)(38.08,43.7)TSN317.9912.5930.41(15.21,20.78)(10.88,14.31)(27.39,33.43)TSN414.3111.1210.7724.79(12.18,16.45)(9.97,12.26)(9.81,11.74)(21.99,27.6) UNH217.4843.52(15.07,19.89)(41.11,45.93) Continuedonnextpage 158

PAGE 159

TickerMr^m11^m22^m33^m44 VZ220.1640.84(17.62,22.69)(38.31,43.38)VZ317.5312.6130.85(14.42,20.65)(11.68,13.55)(27.92,33.79)VZ414.7210.7710.4225.09(12.08,17.35)(10,11.53)(9.55,11.3)(22.73,27.46) WYE222.5138.49(18.88,26.14)(34.86,42.12)WYE318.3314.9527.73(15.64,21.01)(13.32,16.58)(24.36,31.1)WYE414.413.2111.9821.41(12.8,16)(11.97,14.44)(10.81,13.14)(18.86,23.97) 159

PAGE 160

Tablereportsmeandirectionson30samplesbyeachregime,and95%condencelimitandpvaluefromWelchtwosamplet-test.Nullhypothesisonthetestisthatthedierencebetweenmeansofthedirectionsonconsecutiveregimesarezero. WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 161

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 162

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 163

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 164

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 165

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr

PAGE 166

Tableshowsmeanvolatilitieson30samplesbyeachregime,and95%condencelimitandpvaluefromWelchtwosamplet-test.Nullhypothesisonthetestisthatthedierencebetweenmeansofthevolatilitiesonconsecutiveregimesarezero. WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 167

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 168

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 169

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr Continuedonnextpage

PAGE 170

WelchTwoSamplettest(CondenceLimit)(pvalue) TickerMr

PAGE 171

WereportsOLSestimatesofsimpleautoregressivepaneldatabetweendirectionandvolatilityofcurrentregimeandofthelastregime.Explanatoryvariableisspeciedatcolumn^=^,andcomparingpairofregimesis(i;j),jiscurrentregime. ^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 Continuedonnextpage

PAGE 172

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 AEP2(1;2)^-0.0019-0.0050.17320.3240.28260.96910.31150.0370.1686^0.00830.0904-0.26760.48400.5680.19790.0370.19913(1;2)^0.0038-0.32310.5824-0.78040.38240.46310.22890.16120.1175^0.0055-0.11680.2081-0.094900.44670.21730.16120.1217(2;3)^-0.00380.00570.05240.59330.06850.91380.76940.0010.4073^0.0093-0.0165-0.41540.585300.7530.01310.0010.53364(1;2)^0.0062-0.00160.2019-0.91910.22340.99730.75290.07310.1233^0.0065-0.0194-0.0267-0.12880.00010.9090.91140.07310.1208(2;3)^0.00760.14410.3417-1.85990.28190.64480.68710.12710.0952^0.0030.00170.1263-0.04690.00420.97340.34460.12710.12(3;4)^-0.004-0.02030.03860.57460.08650.71390.91060.03450.1646^0.00660.01320.01930.279600.73190.93580.03450.1624 BIIB2(1;2)^0.0119-0.0939-0.0645-0.40250.00590.64150.49420.01680.2094 Continuedonnextpage

PAGE 173

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 BSX2(1;2)^0.0007-0.0162-0.0098-0.05320.79980.74250.89750.68060.0106^0.0174-0.0117-0.1715-0.124500.87670.13260.68060.0923(1;2)^-0.00390.09210.1874-0.05350.80420.83470.78360.93320.0052 Continuedonnextpage

PAGE 174

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 CAT2(1;2)^0.0043-0.0280.0594-0.45950.0060.71590.34340.00020.5118^0.0107-0.1512-0.067-0.912700.15530.44950.00020.53833(1;2)^-0.0157-0.03710.20231.02980.11140.92960.71450.15050.0823^0.0106-0.0988-0.22980.075600.38280.11670.15050.1682(2;3)^0.00130.0005-0.08810.0280.73190.99390.71040.92180.006 Continuedonnextpage

PAGE 175

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 CELG2(1;2)^-0.0294-0.48590.40930.91280.00070.4660.10630.00020.4283^0.02830.1042-0.30970.448100.82430.07970.00020.43663(1;2)^-0.0188-0.16230.18520.90490.01090.73130.49060.00060.3758^0.01970.0309-0.12840.409500.92260.47740.00060.3746(2;3)^-0.008-0.01180.10590.34430.46620.96460.79050.35620.0335^0.0206-0.016-0.1570.095300.90870.45190.35620.06954(1;2)^-0.01570.13412.8711-0.39830.2710.80220.010.37710.2377 Continuedonnextpage

PAGE 176

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 EGN2(1;2)^-0.00440.12680.12060.34590.20820.39910.54090.14470.087^0.0127-0.2245-0.37810.231300.06060.01320.14470.27693(1;2)^-0.00330.17990.18140.00560.50550.47290.57260.98680.0234^0.0118-0.0929-0.27440.001900.52320.13390.98680.0853(2;3)^-0.0040.04530.05340.41580.04650.53170.64820.01570.2141^0.0092-0.11180.01160.491500.14910.92780.01570.25924(1;2)^-0.01060.1147-0.40690.99590.15240.74530.56760.11790.0991^0.0068-0.02760.15680.09180.0010.79680.46720.11790.1048(2;3)^-0.0109-0.0897-0.42581.08350.02480.41420.22320.01460.2163 Continuedonnextpage

PAGE 177

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 FDX2(1;2)^-0.00160.0007-0.01460.25490.38740.99250.90720.0730.1215^0.01-0.0303-0.05730.464300.77560.73370.0730.13023(1;2)^0.0031-0.050.0669-0.31410.25370.71020.67540.18140.0788^0.0089-0.0773-0.0523-0.215300.48610.69270.18140.0958(2;3)^0.00210.10280.1192-0.43430.57060.6920.69530.05470.155^0.0091-0.17480.02-0.31010.00120.42320.93820.05470.1714(1;2)^0.0205-0.66180.9649-2.52640.13160.2870.47820.03410.26^0.0088-0.1065-0.3941-0.063900.28140.06070.03410.3595(2;3)^-0.00180.0406-0.07950.20050.8080.7390.91030.72630.0111^0.0081-0.0302-0.0240.02390.00050.4710.92160.72630.026(3;4)^-0.0018-0.0295-0.04380.28450.35980.52690.73220.15960.0787 Continuedonnextpage

PAGE 178

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 FL2(1;2)^-0.0020.0028-0.18190.20550.56730.97730.16170.31550.1142^0.01440.07330.05590.188500.4380.6590.31550.07783(1;2)^-0.0007-0.05540.10410.02120.93520.89230.83450.9670.0018^0.01410.2531-0.050.003200.09960.79450.9670.1236(2;3)^-0.0012-0.01070.04580.06290.73530.88990.81480.72710.0101^0.0109-0.01050.29340.07570.00210.90160.16320.72710.08064(1;2)^0.0255-0.49320.3778-2.28790.00510.34190.51550.00070.3667^0.0078-0.22770.2173-0.15930.00080.09010.14980.00070.4175(2;3)^0.0098-0.1077-0.8424-0.09970.39540.64050.32450.87190.0423^0.0085-0.00750.2107-0.01020.01480.91940.44290.87190.0446(3;4)^-0.0007-0.00790.05070.02910.81580.85610.71120.85680.007^0.01460.0486-0.15980.043800.36170.33740.85680.0698 GCO2(1;2)^-0.00750.07410.05450.47480.0050.56570.695500.598 Continuedonnextpage

PAGE 179

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 JCP2(1;2)^-0.0009-0.0661-0.02160.07380.73370.36220.75170.7070.0362^0.01340.05810.0070.074900.42770.91880.7070.02663(1;2)^-0.0099-0.5708-0.17310.7530.04060.04380.58970.04110.2812 Continuedonnextpage

PAGE 180

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 KFT2(1;2)^-0.00450.01950.0140.52520.00010.37160.833300.5771^0.0091-0.0364-0.0611.093100.24510.522300.59033(1;2)^0.00350.0084-0.0371-0.35190.70340.97680.96650.71740.0052^0.00660.0519-0.058-0.014600.37110.74680.71740.0423(2;3)^-0.001-0.00650.04210.220.58290.85160.80830.14170.0819 Continuedonnextpage

PAGE 181

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 MGM2(1;2)^0.0205-0.0041-0.0863-1.0620.00190.98510.63030.00270.3027^0.0162-0.0289-0.0174-0.279200.79870.85010.00270.29913(1;2)^0.0115-0.34760.022-0.5690.20130.51080.95890.25170.0633^0.0105-0.18540.197-0.08820.00130.37140.23470.25170.1105(2;3)^-0.00310.0125-0.06950.32320.28730.75660.49770.03730.2055^0.0150.002-0.09590.483800.96760.44380.03730.20424(1;2)^-0.0005-0.3174-0.0290.27440.92570.15920.89410.39830.0994 Continuedonnextpage

PAGE 182

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 MON2(1;2)^-0.0089-0.056-0.04790.57080.22590.78940.86430.13930.1115^0.0171-0.01-0.29860.143900.92430.02610.13930.273(1;2)^0.0189-0.2162-0.4077-1.21270.01010.50180.28740.01530.2415^0.0116-0.0172-0.0552-0.169800.88690.70280.01530.207(2;3)^0.0051-0.0794-0.39060.01350.34620.44590.19720.95660.0738^0.0168-0.0615-0.40740.008600.45980.08790.95660.11744(1;2)^-0.00320.30830.44080.16160.60340.29870.23050.71690.0752^0.0107-0.0589-0.09330.031800.65720.57170.71690.0171(2;3)^0.0166-0.3753-1.2178-0.33010.07080.16950.06370.37060.1789 Continuedonnextpage

PAGE 183

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 PDX2(1;2)^0.0064-0.1949-0.2585-0.39940.0170.68260.08860.00030.4307^0.0154-0.5942-0.3881-1.02060.00010.43340.11090.00030.42373(1;2)^0.0067-0.03160.3975-0.64630.47660.96170.51030.02420.2132^0.0177-0.1938-0.2329-0.27920.00190.65330.55760.02420.2152(2;3)^0.004-0.04860.0004-0.5370.00440.16130.99300.7142^0.0085-0.0842-0.0109-1.326800.1210.892700.71784(1;2)^0.0031-0.69520.5735-0.52340.74940.52490.39750.07710.158^0.0116-0.07670.2954-0.22020.05740.91420.50280.07710.1212(2;3)^0.0155-0.0882-0.1525-0.91040.03430.58520.5470.0030.2927^0.0157-0.0875-0.1145-0.320.00010.35810.44480.0030.3088(3;4)^0.0042-0.06360.059-0.56570.00430.08550.323500.7091 Continuedonnextpage

PAGE 184

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 PG2(1;2)^-0.0022-0.01470.03030.21790.01330.73220.59720.04960.1598^0.0074-0.0160.00830.643700.82830.93320.04960.14823(1;2)^0.0007-0.2185-0.0459-0.08160.85810.18770.89720.84370.0671^0.0072-0.1074-0.1231-0.018700.17560.46690.84370.0792(2;3)^0.0001-0.0654-0.19840.24280.96420.62580.47770.47480.0503^0.0065-0.0047-0.01820.081600.95240.9110.47480.02234(1;2)^0.0018-0.2968-0.45780.42810.58610.1080.18560.42490.1213^0.00440.11830.19830.057600.07940.11550.42490.1525(2;3)^-0.0019-0.0572-0.36440.85310.79410.87440.71340.21720.0738^0.006-0.0068-0.26770.06790.00120.94720.33520.21720.1015(3;4)^-0.0018-0.0305-0.06850.1880.24850.47870.6470.33540.0677^0.00660.029-0.05330.190100.50360.72350.33540.0508 PLCE2(1;2)^-0.0140.08240.18370.49930.17150.6790.58250.12350.0939 Continuedonnextpage

PAGE 185

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 PTRY2(1;2)^-0.00140.09570.115-0.02570.75780.47390.41030.86680.0558^0.0253-0.1231-0.3887-0.042900.47560.02470.86680.1993(1;2)^0.01850.3221-0.1872-0.71630.02310.40710.67080.03480.1847 Continuedonnextpage

PAGE 186

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 QCOM2(1;2)^0.0101-0.0328-0.0943-0.65050.00610.870.52030.00090.3538^0.0147-0.0964-0.1318-0.542500.59730.32250.00090.38043(1;2)^0.01970.43070.0404-0.8220.15530.7110.96830.10260.1101^0.0106-0.21740.2599-0.12070.04130.62520.50320.10260.1232(2;3)^-0.00520.0567-0.09210.38320.2170.38820.58710.10740.1415 Continuedonnextpage

PAGE 187

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 S2(1;2)^0.0066-0.0255-0.2231-0.37060.00150.7970.057300.5318^0.0181-0.1449-0.4743-1.403700.44960.036500.54953(1;2)^-0.0056-0.2033-0.27720.59910.00860.01590.04150.00030.4852^0.00940.1640.29520.66400.07140.0390.00030.4301(2;3)^0.0039-0.2794-0.3433-0.23970.30490.29860.20730.11120.256^0.0109-0.28940.1415-0.39490.01880.40330.68960.11120.12084(1;2)^0.0041-0.0546-0.2525-0.17110.40210.79690.51760.61010.036 Continuedonnextpage

PAGE 188

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 T2(1;2)^0.0009-0.0068-0.1291-0.01580.71020.9430.44850.93870.0271^0.00980.0218-0.4177-0.014700.81030.00660.93870.26423(1;2)^0.00570.20990.3598-0.80840.28940.5180.48560.13510.1015^0.00770.0880.0411-0.103700.44850.8250.13510.0979(2;3)^-0.00240.04710.02540.25370.18530.40770.87210.13930.0982^0.0074-0.0427-0.01140.323700.50730.94880.13930.08984(1;2)^0.0036-0.5964-0.95650.34390.48830.28210.14540.5960.0872^0.00420.12390.27050.03190.00410.46570.17740.5960.07(2;3)^-0.0050.0390.63590.21140.31950.81680.260.71930.0535 Continuedonnextpage

PAGE 189

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 TMO2(1;2)^0.0013-0.0669-0.0534-0.04680.34730.52090.4690.62310.0344^0.01240.023-0.2257-0.201500.91550.13380.62310.09633(1;2)^0.0091-0.37030.4115-0.85830.35460.62770.67670.14030.0912^0.0071-0.13580.2059-0.09520.02310.59320.53020.14030.1049(2;3)^-0.0001-0.0099-0.10860.09360.93270.5840.06330.44710.1846^0.0108-0.0126-0.11590.239500.66410.22410.44710.11924(1;2)^-0.0084-0.195-1.29891.17270.42010.78380.2180.11490.1537^0.00880.0338-0.02350.07920.00030.85480.93260.11490.1021(2;3)^0.0059-0.15770.1665-0.48380.35010.12810.67420.3460.1542^0.00940.0469-0.175-0.070700.24060.24240.3460.1436(3;4)^0.0027-0.0777-0.0441-0.1580.24130.24160.79820.40910.0838 Continuedonnextpage

PAGE 190

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 TSN2(1;2)^-0.0042-0.03420.03070.20440.14340.72930.76810.19720.0661^0.01580.0433-0.14160.308800.72150.26370.19720.10613(1;2)^0.0081-0.17410.5603-0.47240.44540.78140.4270.33440.0659^0.01360.2723-0.1208-0.07590.00030.27390.67060.33440.0864(2;3)^0.00060.0484-0.1111-0.10240.77590.20830.23010.46940.1324^0.01130.0683-0.0671-0.198400.20220.60670.46940.08154(1;2)^-0.0017-0.0876-0.0239-0.1770.87130.89850.97380.76570.0057^0.01090.19280.1076-0.01960.00070.39580.65620.76570.0637(2;3)^-0.0070.07240.72330.16930.4820.68680.18850.6690.0697^0.01680.0179-0.38720.04220.00010.84180.15740.6690.0793(3;4)^-0.0039-0.106-0.03820.45760.14580.12120.77990.06950.1582^0.00680.10780.20470.26470.00020.03450.04060.06950.3315 UNH2(1;2)^-0.00640.05710.16330.2490.03190.61550.19260.13020.1264 Continuedonnextpage

PAGE 191

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 VZ2(1;2)^0.00490.0238-0.0711-0.42940.06770.92630.72840.09060.1078^0.0089-0.0662-0.1841-0.247300.7340.22960.09060.15153(1;2)^0.0069-0.0979-0.2916-0.44440.03130.76480.43260.1960.1065 Continuedonnextpage

PAGE 192

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2 WYE2(1;2)^-0.0041-0.04680.07120.30160.01430.67240.53350.01140.2363^0.0117-0.0248-0.06620.735700.88610.71140.01140.22733(1;2)^0.0099-0.2579-0.4412-1.02070.02380.51960.32490.00020.4582^0.0082-0.0221-0.0877-0.40220.0020.93020.75730.00020.4215(2;3)^-0.0032-0.1116-0.07560.36270.05290.11090.5030.04150.1975 Continuedonnextpage

PAGE 193

^j=^0+^1^i+^2^i+^3^j^j=^0+^1^i+^2^i+^3^jpvalues TickerMr(i;j)^=^^0^1^2^3^0^1^2^3R2

PAGE 194

Back,Kerry,1993,Asymmetricinformationandoptions,ReviewofFinancialStudies6,435{72. Bagehot,Walter,1971,Theonlygameintown,FinancialAnalystsJournal27,12{14. Ball,CliordA.,andWalterN.Torous,1988,Investigatingsecurity-priceperformanceinthepresenceofevent-dateuncertainty,JournalofFinancialEconomics22,123{153. Biais,Bruno,andPierreHillion,1994,Insiderandliquiditytradinginstockandoptionsmarkets,ReviewofFinancialStudies7,743{80. Black,Fischer,1971,Towardafullyautomatedexchange,FinancialAnalystsJournal27,28{35,44. ,1975,Factandfantasyintheuseofoptions,FinancialAnalystsJournal31,36{41,61{72. Blasco,Natividad,PilarCorredor,andRafaelSantamaria,2006,Doesinformedtradingoccurintheoptionsmarket?somerevealingclues,SSRNeLibrary. Bollen,Nicolas,HansR.Stoll,andRobertE.Whaley,2003,Bid/askspreads:Acomparativeanalysis,. Brennan,MichaelJ.,andAvanidharSubrahmanyam,1996,Marketmicrostructureandassetpricing:Onthecompensationforilliquidityinstockreturns,JournalofFinancialEconomics41,441{464. Campbell,JohnY.,AndrewW.Lo,andA.CraigMacKinlay,1997,Theeconometricsofnancialmarkets(PrincetonUniversityPress). Chakravarty,Sugato,HuseyinGulen,andStewartMayhew,2004,Informedtradinginstockandoptionmarkets,JournalofFinance59,1235{1258. Chen,CarlR.,PeterP.Lung,andNicholasS.P.Tay,2005,Informationowbetweenthestockandoptionmarkets:Wheredoinformedtraderstrade?,ReviewofFinancialEconomics14,1{23. CherianJ.A.,JarrowR.A.,1998,Optionsmarkets,self-fulllingprophecies,andimpliedvolatilities,ReviewofDerivativesResearch2,5{37(33). Chib,Siddhartha,1996,Calculatingposteriordistributionsandmodalestimatesinmarkovmixturemodels,JournalofEconometrics75,79{97. ,1998,Estimationandcomparisonofmultiplechange-pointmodels,JournalofEconometrics86,221{241. 194

PAGE 195

Cowan,ArnoldR.,2007,Eventususer'sguide,. Dumas,Bernard,JeFleming,andRobertE.Whaley,1996,Impliedvolatilityfunctions:Empiricaltests,NBERWorkingPapers5500NationalBureauofEconomicResearch,Inc. Easley,David,MaureenO'Hara,andP.S.Srinivas,1998,Optionvolumeandstockprices:Evidenceonwhereinformedtraderstrade,JournalofFinance53,431{465. Easley,David,etal,1996,Liquidity,information,andinfrequentlytradedstocks,JournalofFinance51,1405{36. Ellis,Katrina,RoniMichaely,andMaureenO'Hara,2000,Theaccuracyoftradeclassicationrules:Evidencefromnasdaq,TheJournalofFinancialandQuantita-tiveAnalysis35,529{551. Fama,EugeneF.,andKennethR.French,1993,Commonriskfactorsinthereturnsonstocksandbonds,JournalofFinancialEconomics33,3{56. Fontnouvelle,PatrickDe,RaymondP.H.Fishe,andJereyH.Harris,2003,Thebehaviorofbid-askspreadsandvolumeinoptionsmarketsduringthecompetitionforlistingsin1999,JournalofFinance58,2437{2464. George,ThomasJ.,GautamKaul,andMahendrarajahNimalendran,1994,TradingVolumeandTransactionCostsinSpecialistMarkets,JournalofFinance,Vol.49,No.4,pp.1489-1505,September1994. Glosten,LawrenceR.,andPaulR.Milgrom,1985,Bid,askandtransactionpricesinaspecialistmarketwithheterogeneouslyinformedtraders,JournalofFinancialEconomics14,71{100. Hull,J.C.,2000,Options,futures,andotherderivatives(Harcourt:Orlando,FL)fourthedn. Jereys,Harold,1961,Theoryofprobability/byHaroldJereys(ClarendonPress,Oxford:)3rded.edn. Krinsky,Itzhak,andJasonLee,1996,Earningsannouncementsandthecomponentsofthebid-askspread,TheJournalofFinance51,1523{1535. Kwiatkowski,Denis,PeterC.B.Phillips,PeterSchmidt,andYongcheolShin,1992,Testingthenullhypothesisofstationarityagainstthealternativeofaunitroot:Howsurearewethateconomictimeserieshaveaunitroot?,JournalofEconometrics54,159{178. 195

PAGE 196

Landsiedl,Felix,2005,Themarketmicrostructureofilliquidoptionmarketsandinterrelationswiththeunderlyingmarket,. Liu,Chun,andJohnM.Maheu,2008,AreThereStructuralBreaksinRealizedVolatility?,JournalofFinancialEconometrics,Vol.6,Issue3,pp.326-360,2008. Llorente,Guillermo,RoniMichaely,GideonSaar,andJiangWang,2001,Dynamicvolume-returnrelationofindividualstocks,WorkingPaper8312NationalBureauofEconomicResearch. Madhavan,Ananth,MatthewRichardson,andMarkRoomans,1997a,Whydosecuritypriceschange?atransaction-levelanalysisofnysestocks,RodneyL.WhiteCenterforFinancialResearchWorkingPapers20-94WhartonSchoolRodneyL.WhiteCenterforFinancialResearch. Madhavan,A.,M.Richardson,andM.Roomans,1997b,Whydosecuritypriceschange?atransaction-levelanalysisofnysestocks,REVIEWOFFINANCIALSTUDIES10,1035{1064. Martellini,Lionel,2000,Ecientoptionreplicationinthepresenceoftransactionscosts,ReviewofDerivativesResearch4,107{131. Merton,R.C.,1990,Continuous-TimeFinance(Cambridge:Blackwell). Pastor,L(ubo,andRobertF.Stambaugh,2001,Theequitypremiumandstructuralbreaks,TheJournalofFinance56,1207{1239. ReutersKnowledge,2006,www.knowledge.reuters.com,. Robert,ChristianP.,andGeorgeCasella,2005,MonteCarloStatisticalMethods(SpringerTextsinStatistics)(Springer). Roll,Richard,1984,Asimpleimplicitmeasureoftheeectivebid-askspreadinanecientmarket,JournalofFinance39,1127{39. Rosenthal,DaleW.,2008,ModelingTradeDirection,SSRNeLibrary. Ross,StephenA,1976,Optionsandeciency,TheQuarterlyJournalofEconomics90,75{89. Ross,SheldonM.,1996,StochasticProcesses,2ndEdition(Wiley). RudigerFahlenbrach,PatrikSandas,2006,Doesinformationdrivetradinginoptionstrategies?,. Vijh,AnandM,1990,Liquidityofthecboeequityoptions,JournalofFinance45,1157{79. 196

PAGE 197

Joon-HuiYoonisadoctorofphilosophyofquantitativenanceconcentrationinindustrialandsystemsengineeringattheUniversityofFlorida.HereceivedhisPh.D.degreeinquantitativenanceforhisdissertationon\InformationAsymmetryinDirectionandVolatility:PriceProcessandTransactionLevelAnalysis."HealsoholdsmasterofsciencedegreefromtheTexasA&MUniversityandisaholderofCertiedinProductionandInventoryManagement.Hehadworkedforfouryearsininformationtechnologyindustrybeforepursuingacademia.Heisinterestedincombininginformationtechnologyandnancialtheoriestomakenancialmarketmoreecient. 197