<%BANNER%>

Making Future More Feasible for Independently Owned Hotels

MISSING IMAGE

Material Information

Title:
Making Future More Feasible for Independently Owned Hotels Application of Time Series Models for Hotel Sales Forecasts as Means to Gain a Competitive Advantage
Physical Description:
1 online resource (83 p.)
Language:
english
Creator:
Sorokina, Ekaterina Igorevna
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Recreation, Parks, and Tourism, Tourism, Recreation, and Sport Management
Committee Chair:
Semrad, Kelly J
Committee Members:
Mills, Brian M
Pardalos, Panagote M

Subjects

Subjects / Keywords:
forecasting -- hotels -- modelling -- pricing
Tourism, Recreation, and Sport Management -- Dissertations, Academic -- UF
Genre:
Recreation, Parks, and Tourism thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
The high number of closures within independently owned hotel sector has reached around 80% of all hotel closures in the US market in 2013. The lack of the brand name and recognition in the market brings this sector under higher financial risks. In addition,the complexity and uncertainty of hotel financial performance is determined by diverse internal and external environmental factors that may include: increased industry competition, the instability of economic conditions,technological advances, marketplace developments, and changes in customer preferences. Therefore, there is a need in reliable forecasting tools in order to reduce the adverse effects that may be associated with such environmental factors on hotel financial performance. However, the independently owned hotels may not have sufficient resources to utilize elaborate forecasting methods or expensive software solutions. The available financial resources could be scarce considering that independently owned hotels have to compete with the larger hotel chains. In addition, some of the forecasting tools are demanding in terms of expertise. While the importance of effective forecasting tools for the independently owned hotels seems to be apparent, this sector may hesitate to invest financial resources into forecasting solutions due to cost and time constrains. Thus, the study recommends time series models as an accessible and reliable alternative for independently owned hotels. The study further focuses on testing several time series models in order to assist hotel managers in determining which forecasting method(s) may be the most accurate in terms of formulating future hotel sales expectations.
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 Ekaterina Igorevna Sorokina.
Thesis:
Thesis (M.S.)--University of Florida, 2013.
Local:
Adviser: Semrad, Kelly J.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045737:00001

MISSING IMAGE

Material Information

Title:
Making Future More Feasible for Independently Owned Hotels Application of Time Series Models for Hotel Sales Forecasts as Means to Gain a Competitive Advantage
Physical Description:
1 online resource (83 p.)
Language:
english
Creator:
Sorokina, Ekaterina Igorevna
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Recreation, Parks, and Tourism, Tourism, Recreation, and Sport Management
Committee Chair:
Semrad, Kelly J
Committee Members:
Mills, Brian M
Pardalos, Panagote M

Subjects

Subjects / Keywords:
forecasting -- hotels -- modelling -- pricing
Tourism, Recreation, and Sport Management -- Dissertations, Academic -- UF
Genre:
Recreation, Parks, and Tourism thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
The high number of closures within independently owned hotel sector has reached around 80% of all hotel closures in the US market in 2013. The lack of the brand name and recognition in the market brings this sector under higher financial risks. In addition,the complexity and uncertainty of hotel financial performance is determined by diverse internal and external environmental factors that may include: increased industry competition, the instability of economic conditions,technological advances, marketplace developments, and changes in customer preferences. Therefore, there is a need in reliable forecasting tools in order to reduce the adverse effects that may be associated with such environmental factors on hotel financial performance. However, the independently owned hotels may not have sufficient resources to utilize elaborate forecasting methods or expensive software solutions. The available financial resources could be scarce considering that independently owned hotels have to compete with the larger hotel chains. In addition, some of the forecasting tools are demanding in terms of expertise. While the importance of effective forecasting tools for the independently owned hotels seems to be apparent, this sector may hesitate to invest financial resources into forecasting solutions due to cost and time constrains. Thus, the study recommends time series models as an accessible and reliable alternative for independently owned hotels. The study further focuses on testing several time series models in order to assist hotel managers in determining which forecasting method(s) may be the most accurate in terms of formulating future hotel sales expectations.
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 Ekaterina Igorevna Sorokina.
Thesis:
Thesis (M.S.)--University of Florida, 2013.
Local:
Adviser: Semrad, Kelly J.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045737:00001


This item has the following downloads:


Full Text

PAGE 1

MAKINGFUTUREMOREFEASIBLEFORINDEPENDENTLYOWNEDHOTELS:APPLICATIONOFTIMESERIESMODELSFORHOTELSALESFORECASTSASMEANSTOGAINACOMPETITIVEADVANTAGEByEKATERINASOROKINAATHESISPRESENTEDTOTHEGRADUATESCHOOLOFTHEUNIVERSITYOFFLORIDAINPARTIALFULFILLMENTOFTHEREQUIREMENTSFORTHEDEGREEOFMASTEROFSCIENCEUNIVERSITYOFFLORIDA2013

PAGE 2

c2013EkaterinaSorokina 2

PAGE 3

Toallwhohavebeenthereforme 3

PAGE 4

ACKNOWLEDGMENTS IwouldliketoexpressmysinceregratitudeforallthesupportandunderstandingtomyadvisorDr.KellySemrad.Herprofessionalism,purposefulness,andthoroughknowledgehaveguidedmethroughouttheeducationattheDepartmentofTourism,Recreation,andSportManagement.IamdeeplythankfultoDr.Semradforhermentorship,whichhelpedmetopursuemyacademicgoals.Additionally,IwouldliketothankmythesiscommitteemembersDr.BrianMillsandDr.PanosM.Pardalosfortheirvaluableassistanceinmyresearchwork.TheyhavebeenhighlyattentivetomyinquiresandwerereadilyavailabletoassistattimeswhenIneededtheiracademicguidance.IalsowanttoexpressmyappreciationtoDr.Sagas,thechairoftheDepartmentofTourism,RecreationandSportManagement.Hissupervisionandconsiderationforstudents'academicprogressaregreatlyvalued.Lastbutnotleast,IwouldliketothankmybelovedhusbandAlexey,mymotherandmyuntimelygonefather.TheyhavealwaysencouragedmeandhelpedmetocometothepointwhereIamnow,andIamextremelygratefulfortheirsupport. 4

PAGE 5

TABLEOFCONTENTS page ACKNOWLEDGMENTS .................................. 4 LISTOFTABLES ...................................... 7 LISTOFFIGURES ..................................... 8 LISTOFSYMBOLS .................................... 9 ABSTRACT ......................................... 10 CHAPTER 1INTRODUCTION ................................... 12 1.1ChallengesforIndependentlyOwnedHotels ................ 15 1.2ContemporaryForecastingPractices ..................... 18 1.3ProblemStatement ............................... 23 1.4PurposeoftheStudy .............................. 24 1.5Limitations ................................... 27 2REVIEWOFLITERATURE ............................. 29 2.1BackgroundInformation ............................ 29 2.1.1CharacteristicsoftheYield-ManagementSystems ......... 32 2.1.2CharacteristicsoftheIndependentlyOwnedHotels ......... 34 2.2CommonlyUsedForecastingMethodsandModels ............. 36 2.2.1JudgmentalMethod .......................... 37 2.2.2DescriptiveStatistics .......................... 38 2.2.3RegressionModels ........................... 38 2.2.4EconometricModels .......................... 39 2.2.5ForecastingSoftwareSolutions .................... 40 2.2.6TimeSeriesModels .......................... 41 3METHODOLOGY .................................. 43 3.1TheAdditiveSeasonalDecomposition .................... 44 3.2TheAdditiveExponentialSmoothing ..................... 45 3.3TheAutoregressiveIntegratedMovingAverage(ARIMA) .......... 46 4APPLICATIONOFTIMESERIESMODELS .................... 49 4.1PreparationoftheData ............................ 49 5

PAGE 6

4.2ResultsoftheUnitRootTests ......................... 52 4.3ForecastingSoftware .............................. 53 4.4Findings ..................................... 55 4.4.1PerformanceoftheAdditiveSeasonalDecomposition ....... 55 4.4.2PerformanceoftheAdditiveExponentialSmoothing ........ 57 4.4.3PerformanceoftheAutoregressiveIntegratedMovingAverage(ARIMA)Model ............................. 62 4.5SummaryoftheFindings ........................... 65 5CONCLUSIONSANDIMPLICATIONSOFTHESTUDY ............. 67 5.1Conclusions ................................... 69 5.2LimitationsoftheStudy ............................ 70 5.3ImplicationsfortheIndustryPractitioners .................. 72 5.4FutureResearch ................................ 74 5.5SummaryoftheStudy ............................. 75 REFERENCES ....................................... 77 BIOGRAPHICALSKETCH ................................ 83 6

PAGE 7

LISTOFTABLES Table page 2-1TheCharacteristicsoftheLodgingIndustry .................... 35 4-1TheAdditiveExponentialSmoothing-ForecastedADRsfortheSTRCompetitiveSetofHotels(2010) ................................. 61 4-2TheAdditiveExponentialSmoothing-ForecastedADRsfortheIndividualHotel(2010) ......................................... 61 4-3TheARIMA-ForecastedADRfortheSTRCompetitiveSet(2010) ....... 63 4-4TheARIMA-ForecastedADRfortheIndividualHotel(2010) .......... 63 4-5TheResearchQuestionsandHypothesesInvestigatedintheStudy ...... 66 7

PAGE 8

LISTOFFIGURES Figure page 1-1ExistingRoomSupplyfor2009and2013YearintheUSMarket ........ 12 1-2HotelClosuresintheUS(1979-2013) ....................... 14 4-1ADRoftheSTRCompetitiveSetofOrlandoHotels(2007-2009) ........ 50 4-2ADRoftheIndividualHotel(2007-2009) ...................... 51 4-3TheAdditiveSeasonalDecompositionAppliedtotheSTRCompetitiveDataSet(2007-2010) ................................... 57 4-4TheAdditiveSeasonalDecompositionAppliedtotheIndividualHotelDataSet(2007-2010) ................................... 58 4-5TheAdditiveExponentialSmoothingForecastedADRfortheSTRCompetitiveDataSet(2010) ................................... 60 4-6TheAdditiveExponentialSmoothingForecastedADRfortheIndividualHotelDataSet(2010) ................................... 60 4-7TheARIMAForecastedADRfortheCompetitiveSetofHotels(2010) ..... 64 4-8TheARIMAForecastedADRfortheIndividualHotel(2010) ........... 64 8

PAGE 9

LISTOFSYMBOLS,NOMENCLATURE,ORABBREVIATIONS ADF AugmentedDickey-FullerADR AveragedailyrateARIMA AutoregressiveintegratedmovingaverageMAPE MeanabsolutepercentageerrorRevPAR RevenueperavailableroomRMSPE RootmeansquarepercentageerrorSTR Smithtravelresearch 9

PAGE 10

AbstractofThesisPresentedtotheGraduateSchooloftheUniversityofFloridainPartialFulllmentoftheRequirementsfortheDegreeofMasterofScienceMAKINGFUTUREMOREFEASIBLEFORINDEPENDENTLYOWNEDHOTELS:APPLICATIONOFTIMESERIESMODELSFORHOTELSALESFORECASTSASMEANSTOGAINACOMPETITIVEADVANTAGEByEkaterinaSorokinaAugust2013Chair:KellySemradMajor:Recreation,Parks,andTourism Thehighnumberofclosureswithinindependentlyownedhotelsectorhasreachedaround80%ofallhotelclosuresintheUSmarketin2013.Thelackofthebrandnameandrecognitioninthemarketbringsthissectorunderhighernancialrisks.Inaddition,thecomplexityanduncertaintyofhotelnancialperformanceisdeterminedbydiverseinternalandexternalenvironmentalfactorsthatmayinclude:increasedindustrycompetition,theinstabilityofeconomicconditions,technologicaladvances,marketplacedevelopments,andchangesincustomerpreferences.Therefore,thereisaneedinreliableforecastingtoolsinordertoreducetheadverseeffectsthatmaybeassociatedwithsuchenvironmentalfactorsonhotelnancialperformance. However,theindependentlyownedhotelsmaynothavesufcientresourcestoutilizeelaborateforecastingmethodsorexpensivesoftwaresolutions.Theavailablenancialresourcescouldbescarceconsideringthatindependentlyownedhotelshavetocompetewiththelargerhotelchains.Inaddition,someoftheforecastingtoolsaredemandingintermsofexpertise.Whiletheimportanceofeffectiveforecastingtoolsfortheindependentlyownedhotelsseemstobeapparent,thissectormayhesitatetoinvestnancialresourcesintoforecastingsolutionsduetocostandtimeconstrains.Thus, 10

PAGE 11

thestudyrecommendstimeseriesmodelsasanaccessibleandreliablealternativeforindependentlyownedhotels.Thestudyfurtherfocusesontestingseveraltimeseriesmodelsinordertoassisthotelmanagersindeterminingwhichforecastingmethod(s)maybethemostaccurateintermsofformulatingfuturehotelsalesexpectations. 11

PAGE 12

CHAPTER1INTRODUCTION TheUShotelmarketiscurrentlyexperiencingasteadygrowth;theroomsupplyhasincreasedfrom4,796,704in2009to4,902,217roomsin2013year.Independentlyownedhotelsrepresent31%oftheexistingroomsupplywhichistheequivalentof1,538,048rooms( STR 2013 ).Theindependenthotelsectorhasthelargestroomsupplyfollowedbytheuppermidscale-868,318rooms(Figure 1-1 ).However,thisnumbermaybemisleading,becauseitdoesnotindicatethenumberofhotelclosuresandnewentrantsintheUSAmarket. Figure1-1. ExistingRoomSupplyfor2009and2013YearintheUSMarket Currently,theindependentclassofthehotelshasthehighestnumberofclosures(Figure 1-2 )( Wilson 2013 ).Independentlyownedhotelsarecommonlymanagedbytheownersoftheproperty.Majorityofthehotelswithinthissectorarestillcompletelyindependentofthelargerchains;however,thenumberofthefranchisingagreements 12

PAGE 13

betweenindependentlyownedhotelsandlargerchainsisincreasing( Sturmanetal. 2011 ).Theuniqueoperatingstructureoftheindependentlyownedhotelsandexposuretohighernancialrisks(e.g.,insufcientnancialsupport)( Mayock 2011b )maybethereasonsforthegrowingnumberofthefranchisingagreementsandhotelclosureswithinthissector.Otherrisksthatindependentlyownedhotelsareexposedtoinclude:1)limitednancialresources( Mayock 2011b );2)inabilitytocompetewithlargerchains;3)intensecompetitioninthemarketlandscapes( Enzetal. 2009 );4)lackoftherecognitionamongconsumers( Sturmanetal. 2011 );and,5)loweroccupancyrateswhencomparedtothelargerchains( STR 2010 ; Mayock 2011b ).Giventheuniqueanddiverserisksthatindependentlyownedhotelsareexposedtowhencomparedtootherhotelsectors,itbecomescriticalthatmanagersofindependentlyownedhotelscentrallyfocusonrevenuemaximizationthroughthesuccessfulsalesofthecoreproduct)]TJ /F1 11.955 Tf 12.62 0 Td[(room-night.Thisisespeciallytruewhenconsideringthatindependentlyownedhotelshavelimitednancialresources. Inotherwords,independentlyownedhotelsdonothavethesecurednancialsupportfromcorporateheadquarterssuchasthatofmajorhotelchains.Yet,thesehotels(i.e.independentlyownedandchainhotels)competeinthesamemarketsforthesameconsumersbutwithadiscrepancyregardingavailableresourcesthatassistshotelsinthecapturingofmarketshares.Generally,theresourcesthatindependentlyownedhotelshaveaccesstoaremuchlessthanthoseavailabletochainhotelsmakingitdifcultforindependentlyownedhotelstocompetewithlargechains.Oneofthemajorchallengesthatresultsfromlimitednancialresourcesishavingaccesstoadequateforecastingtools(e.g.softwareanddatasystems)thatmayassistindependentlyownedhotelmanagerswithsettingfutureroomsalesobjectivesandroomratesthatwillsellinmarketconditionsthatareunknown.Thevastmajorityoflargechainhotelshave 13

PAGE 14

Figure1-2. HotelClosuresintheUS(1979-2013) anaccesstoforecastingsoftwareandcorporateguidanceregardingappropriatesalesexpectationsandforecastingfuturesellableroomrates. Therefore,theobjectiveofthisstudyistoinvestigatetheeffectivenessoftimeseriesmodelsforhotelsalesforecastingintermsofthemodelsaccuracyandaccessibilitytoindependentlyownedhotelmanagers.Theaccuracyofthemodelswillbedeterminedbasedontheabilityofthemodelstorepresentactualsalesdata(i.e.,seasonaluctuationsandtrend).Theunitofanalysisisaveragedailyrate(ADR).ADRisacommonperformanceindicatorinthelodgingindustrythatmanagersusetoshowtherevenueacquiredperahotelroomthathasbeensold.Theavailabilityofinformationregardingfuturesalesmayaccompanyindependentlyownedhotelmanagersinrevisingcurrentpricingstrategies.Additionally,giventheinformationaboutthefuturesales,managersmaymanipulatetheresourcesinsuchawaythatwillleadtoanincreasein 14

PAGE 15

theproductivitylevelofthehotel.Theaccessibilityofthemodelswillbedenedbasedonthreecharacteristics:cost,time,andexpertiserequiredtoutilizethemodels. Thefollowingsectionsofthechapterwillinvestigatetheimportanceofhotelsalesforecastingintherelationtofuturehoteloperationaldecisions.Additionally,thechapterwillincludethedescriptionofseveraluniquechallengesthatmayhaveanimpactonindependenthotelmanagersabilitytoaccuratelyforeseefuturesales,especiallywhencomparedtothehotelclassesthatarenotindependentlyowned.Thenwillfollowabriefreviewofcommonlyusedforecastingmethodsandmodelsinthelodgingindustry(e.g,judgmentalmethod,regressionanalysis,applicationofeconometric,timeseriesmodels,etc.).Theresearchproblemandthepurposeofthestudywillbestatedbasedontheprecedingoverviewofthechallengesforindependentlyownedhotelmanagersandimplicationsofpoorqualitysalesforecastingfortheindependentlyownedhotelunit.Thechapterwillthenproceedtothereviewofthemodelsandsoftwareusedforthemodelsapplicationalongwithadescriptionofthedataobtainedinordertotestthemodels.Thedescriptionoftheutilizedtimeseriesdatawillbefollowedbytheresearchquestions.Theconclusionwillconcentrateonthesignicanceofthisstudyaswellasontheresearchlimitations. 1.1ChallengesforIndependentlyOwnedHotels Independentlyownedhotelsmayofferawiderangeofprices;therefore,thehotelscouldvaryfromeconomyclasshoteltoupscaleorevenluxuryhotels.Ontheotherhand,chainhotels(suchasHilton,ResidenceInn,HolidayInn,LaQuintaInnsandSuites,andExtendedStayAmerica)usecompetitiveparitytopriceroomrateswithinthesamepricepointsoratthesamepricethresholds.Thus,inorderforindependentlyownedhotelstostaycompetitiveinthemarketplacetheyhavetocompetewiththelargerchainsthathaveabrandnameaswellasareferencepriceforaroompurchased 15

PAGE 16

underthebrandname.Areferencepricerepresentstheexpectedpriceforaproductorservice;thus,thepricethatfallsbelowthisreferencepricelevelisconsideredaslow,andpricethatisabovethereferencepriceisperceivedashigh( Karande&Magnini 2011 ).Inotherwords,referencepriceisthepricewhichconsumersarewillingtopayforaproductoraservice. Whileindependenthotelsmayuseforecastingtoofferacompetitiveprice,themajorityofindependentlyownedhotelsdonotpossessareferencepriceorabrandnameinthemindsofmostcustomers.Thelackofabrandnameforahotelpropertyisoneofthechallengesthatindependentlyownedhotelmanagersmayencounter( Sturmanetal. 2011 ).Whenconsideringhighlycompetitiveenvironments,independentlyownedhotelsrelyonlyontheirownmarketingstrategiesatthebusinessunitlevel.Whereaschainhotels,havethemarketingeffortsfromthebusinessunitaswellasthecorporateheadquarters.Theimplementationandexecutionofthemarketingstrategiesthatareusedtomaintainacompetitivepositioninthemarketmayrequirehighlyprofessionalmanagementteamsaswellasadditionalnancialinvestmentsthatindependentlyownedhotelsmaynothave. Themanagersofchainhotelsmaypossesshigherlevelofskillsandprofessionalismduetospecicexpectationsandrequirementsoftheirpositions.Onthecontrary,thehotelmanagersofindependentlyownedhotelsmaylacksomeoftheprofessionalskillsforeffectiveoperationofthehotelthatcouldincludemaintaininghighproductivitylevelsandthehotelspositioninthemarket( Burgess 2007 ; Sturmanetal. 2011 ).Anotherchallengefortheindependentlyownedhotelsistheirsmallsize.Whilechainhotelsmaybalancenancialresourcesamongmultiplepropertiestoensurenancialsecurityforallofthem;independentlyownedhotelssolelyrelyontheirownnancialperformanceformanagementoftheresourcesandestablishmentoffuturebudgetingandpricing 16

PAGE 17

strategies.Thus,theabilitytoaccuratelyforecasthotelsalesisacriticalnecessityforsuccessfulnancialoperationofanindependentlyownedhotel. Cassar&Gibson ( 2007 )examinedforecastingpracticesinsmallrmsandexplainedthattheabilitytoaccuratelyforecastsalesisindispensableforthesurvivalofthermgiventhatthefuturegrowthandprotabilityofthecompanydependsonthisability. STR ( 2009c )(SmithTravelResearch)reportedthatoccupancyrateshaddecreasedonaverageby10.7%intheUSlodgingindustryasaresultoftheeconomicdownturnin2008.STRcollectstheinformation(averagedailyrate,revenueperavailableroom,occupancy,revenue,etc.)fromtheUSandinternationalmarketsthatcomprisemorethan5.7millionofrooms.Thecollectedinformationisthenaggregatedintothereportswhichserveasbench-markingtoolsforthehotelswithinthemarketormarketsector( Hood&Mandelbaum 2012 ). TheindustryexpertshaveobservedthatthehoteloccupancylevelintheUSisslowlyrecovering.However,theADRsintheindustryareslowtorecouptothelevelthattheADRswereatpre-recession;thedecreaseofADRby8.8%in2009wasfollowedby1.8%dropin2010( STR 2009a ).ADRwithintheindependentsectorhasincreasedonlyby0.3%in2010whencomparedto2009( Mayock 2011b ).Thismeansthatasoccupancyratesincreasetheamountofrevenuethathotelscaptureisstilllessthenpre-recessionarytimes. Interestingly,theindependentlyownedhotelclassexperiencedmoremarketentrantsandagrowthof4.9%in2010( Mayock 2011b ).ThisgrowthrateisalarmingandneedscautioninthatthesenewmarketentrantsaswellastheformerexistingindependentlyownedhotelsmaybeexposedtodiversenancialrisksduetolowoccupancyratesandlowADRs.STRreported56.2%averageoccupancylevelin2010forindependentlyownedUShotelsascomparedto71.2%ofluxuryand75.3%ofupperupscalechainhotels.Forhotels,accuratesalesforecastingisnecessaryforefcient 17

PAGE 18

planningandallocationofresourceswhenconsideringchallengingconditionsofthelodgingindustrythatcreateuncertaintyaboutfuturermperformance(e.g.volatiledemandandperishablecoreproduct).Perishabilitymeansthatifaroomnightisnotsoldonanygivennightthatthepotentialrevenuefromthatsalecannotberedeemedandislostforever( Hanksetal. 1992 ).Whenconsideringthatthecoreproductisplaguedbyperishableinventorycoupledwiththevolatilenatureoftherelationshipbetweenaxedroomsupplyanduncertainfutureroomdemand;theabilityofmanagerstoaccuratelyforecastroomsalesinordertomaximizermproductivitylevelsiscritical. Themaximizationofproductivitymaybeaparticulardifculttaskforindependentlyownedhotelsgiventheconstraintsthatthesehotelsoperateunder.Therefore,independentlyownedhotelshaveaneedforeffectiveandaccurateforecastinginordertoincreasetheirproductivity.Thisisespeciallytruewhenconsideringthefollowinginformation:a10%increaseinsalesforecastaccuracyintheairlineindustryledtoarevenueincreaseof5%-20%( Aghazadeh 2007 ; Lee 1990 ).Itispossiblethatsuchndingscouldalsoapplywithinthelodgingindustrygiventhatbothofthesesub-industriesaresectorswithinthehospitalityindustry.Also,bothofthesesectorspossessthenecessarycharacteristicsthatarerequiredtoimplementyieldmanagementpractice(i.e.perishabilityofthecoreproduct,theclearsegmentationofmarkets,volatiledemand,relativelyxedcapacity,advancedsales,andtheappropriatecoststructure)( Kimes 1989 ). 1.2ContemporaryForecastingPractices Theforecastingpracticesinthelodgingindustrymayvaryfromsimpleandexperience-basedformulationoffuturesalestotheapplicationofadvancedforecastingsoftwaresolutions. Mentzer&Cox ( 1984 )demonstratedthatsalesforecastsweremostlybasedonmanagersusingsubjectivemethods.Thismeansthatdecisionsweremadeby 18

PAGE 19

agroupofexecutivemanagersregardingfutureexpendituresandprices.Thisapproachseemstoremaininthecurrentmanagementpractices. Croes&Semrad ( 2012 )referredtothismethodasbeingwidelyimplementednowadaysbythehotelmanagerstowardssettingfutureroomrates.Another,commonlyusedapproachtoidentifyingthefuturesalesisapplicationofobjectivemethods. Mentzer&Cox ( 1984 )indicatedthatamongobjectivemethods,regressionanalysiswasmorefrequentlyusedforforecasting.ThissectionwillreviewcurrentstateofforecastingpracticesandwhetherobjectivemethodsreferencedbyMentzer&Coxarestillutilized. Although,forecastingisamultistageprocessthatdependsonmanydiversefactors(i.e.economic,political,climatic,social),mosthotelmangersrelyontheirexperienceformakingoperationaldecisionsforahotel.Forexample,thesedecisionsmayincorporatethesettingofmarketingobjectives,developmentofnewpricingstrategies,marketingstudiesofconsumerbehavior,determininganinuenceoffutureeventsonconsumersdemand,stafngdecisions,roominventoryallocation,andcapitalinvestmentdecisions( Frechtling 2001 ; Limetal. 2009 ; Tony&Poon 2012 ).Managersexperienceisgenerallybasedontheinformationaboutcomingeventswhicharesimilartopasteventsthatinuencedroomssales,thesalesdatafrompastseasons,andmanagersperceptionofthecurrentoperatingenvironmentwhenconsideringpast,current,andanticipatedfuturemacroandmicroenvironmentalconditionsinuencingproductivitylevelsofthehoteloperations( Cranage 2003 ; Limetal. 2009 ).Forinstance,independenthotelmanagersintheOrlando,FLmarketmayusepastconventiondataandtouristinformationtomakeadecisionaboutthenumberofroomsthatshouldbeallocatedtoeachofthetargetsegments.Undoubtedly,hotelmanagersexpertiseisindispensableforhotelsalesforecasting( Aghazadeh 2007 ),becausetheypossessvaluableinformationregardingpasthotelsperformance.However,Croes&Semraddiscussedthatsome 19

PAGE 20

ofthehotelmanagersmaygroundtheirdecisiononlyontheirowninstinct. Croes&Semrad ( 2012 )furthernotedthatthismethodmightnotaccountforacomplexnatureofthelodgingindustry. Anotherapproachthatwasfoundtobefrequentlyusedbythehotelmanagersforsalesforecastingwasdescriptivestatistics(e.g.,averages,standarddeviations,andsomecorrelations)( Kahn 1998 ).However,hotelsalesforecaststhatarederivedfromsuchdescriptiveperformancemeasuresmaybeinaccuratewhenconsideringthenon-stationaryconditionsthatmaypresentinthetimeseriesdatasetproperties.Thismeansthatthedataholdsmemory;thedatapointsarenotfreefrominuenceofoneanother( Frechtling 2001 ).Ifthehistoricdataisnottreatedtoaccommodatesuchnon-stationaryconditionsthesalesforecaststhataregeneratedfromthisdatamaygivemisleadingorerroneousresults. Regressionmodelsarealsobroadlyavailableforecastingmodelsandcouldbestillutilizedbythehotelmanagersinordertoformulatetheirfutureexpectations.However, Song&Witt ( 2000 ),afterapplicationofseveralregressionmodelswithnon-stationarydatafoundthatthemodelspresentedmisleadingandunreliableresults,otherwiseknownasspuriousregression(i.e.,anonexistentrelationshipbetweenvariables).Eventhough,thisndingisnotnewforregressionmodels,SongandWittexplainedthatthisissuehasbeenneglectedbypractitionersandforecastersinthetourismarea.Additionally, Limetal. ( 2009 )pointedoutpossibleineffectivenessofregressionmodelsduetomultipledemandcharacteristicsandoverallcomplexityofthelodgingindustry. Amongaforementionedforecastingmethodsandmodels,econometricmodelsmaybethemostaccurateinmanagingstochasticdataandtreatingthedataforunitroots(i.e.,whenthedatapointsaredependentoneachother)whichisaprevalentconditionoftimeseriesdatainthelodgingindustry( Semrad 2010 ).However,thecomplexityand 20

PAGE 21

time-consumingnatureofthesemodelsshouldbetakenintoaccountrst.Independenthotelmanagersmaylacksufcientstatisticalexpertisethatwouldbenecessarytosettheparametersofthesemodelsandmaylacktheabilitytointerpretthestatisticalresultsfromsuchcomplexoutput.Therefore,modelsthatfallwithinaneconometricdomainarecomplicatedtoexecute. Anothermethodthatmaybeimplementedforsalesforecastingistheapplicationofrevenuemanagementandforecastingsoftwaresolutions.Forecastingsoftwaremayhaveseveraladvantageswhencomparedwiththepreviouslyreviewedmethodsandmodels.Forexample,forecastingsoftwaresolutionsmaybemoresophisticated,becausesomeofthesolutionsallowchoosingfromseveralforecastingtechniques( Moonetal. 2003 ).However,purchasingofforecastingsoftwaresolutionsrequiresadditionalupfrontexpendituresthatisquiteexpensiveandinsomecasessubsequentmonthlyexpenditures(i.e.,forconsultingandmonitoringofhotelnancialperformance).Theindependentlyownedhotelsmaynothavesufcientnancialresourcestouseservicesofthecompaniesthatofferdifferentforecastingsolutions. Thus,theexpectationsarethatthepricingstrategiesoftheindependentlyownedhotelsareformulatedsubjectivelyorbyusingdescriptivestatistics.Bothofwhichareinsufcientinformationpertainingtoforecastingaccuratefuturesalesthatmaybeusedbythehotelmanagersintheirdecision-making(i.e.,futureprices,expenditures)( Yksel 2007 ).Thisisespeciallytrueforindependentlyownedhotelsgiventhatthesectorhaslowerroomdemandthanothersectorsbutisalsocurrentlyaddingtotheroomsupplybythedevelopmentofmoreindependenthotels.Thisplacesthehotelsectorunderhighernancialrisk.Forinstance,aforecastoffutureoccupancyratesandhotelnancialperformancemighthaveaneffectonthesuccessiveinvestmentdecisions(i.e.,investmentsrequiredformarketingdepartment,foodandbeveragedepartment). 21

PAGE 22

Therefore,itbecomesindispensabletodeterminefuturecostbasedonthedemanductuations. Nooteboometal. ( 1988 ),denedtwodifferenttypesofcostwhichareactualandnormalizedcosts,wherenormalizedcostisexpectedcostwithnoeffectofdemandvariation.Thesalesforecastsarenecessaryinordertoaccuratelydeterminefuturelevelsofbothcosts.Thus,otherpossibleconsequencesofpoororinaccurateforecastscouldbeunder-orover-budgetingfordifferentrevenuegeneratinghoteldepartments,identifyingwrongtargetsegments,settingofineffectivepricingstrategiesfortheroomsandservicesprovided,under-orover-estimationofworkforceneeded,etc.( Frechtling 2001 ; Weatherford&Kimes 2003 ; Zakharyetal. 2009 ).Thefuturehotelpricingstrategiesthatareformulatedbasedoninaccuratesalesforecastsmaynegativelyaffectfuturedemandandresultinincreasedoperatingcostsanddecreasedtotalrevenues. Theavailabilityoftheadequateforecastingtoolsiscrucial,especiallyfortheindependentlyownedhotelsduetohighlyvolatilenatureofdemand( Frechtling 2001 ; Jang 2004 ).Inadditiontoconstantseasonalvariationthedemandisdependentuponchangingcustomerpreferences,andincreasingcompetitioninthelodgingindustry( Enzetal. 2009 ).Therefore,alloftheseconditionsmayimpairhotelmanagersabilitytoobjectivelyidentifyfutureroomsales.UShotelclosureanalysispublishedbySTRintheyearof2010and2013providedevidentiarystatisticsthatillustratestheeffectofvolatiletourismdemandandchangingmacroenvironmentalconditions(e.g.softeconomictimes)onthelodgingindustry.Accordingtothereports,1,400hotelswereforcedoutofbusinessnationwidebetweentheyearsof2010-2012ofwhich1,144wereindependentlyownedhotels.Thismeansthatonly256ofcorporatechainhotelswererequiredtoclosetheirdoorsofoperation.Thismaybeduetothecorporatechainhotelshavingaccessto 22

PAGE 23

thenecessaryresourcestoaccuratelyforecastsaleswhileindependentlyownedhotelslacksuchresources. 1.3ProblemStatement Thelackofeffectiveandpracticalforecastingmodelsthatrequireminimaltime,costandexpertisefortheirapplicationdenestheresearchproblem.Independentlyownedhotelmanagersneedaccurateyetpracticalforecastingmodelsthatmayidentifysalesuctuationsovertime.Withoutsuchforecastingmodelsindependentlyownedhotelmanagersmaynotensuretheirnancialsecurityormarketpositionwhencompetingagainstlargerchains.Thegrowingnumberoftheclosuresoftheindependentlyownedhotelscomparativelytolargerchainsdemonstratesthatindependentlyownedhotelsareathighernancialrisks.TheSTRUSanalysisofhotelclosures( Wilson 2013 )indicatedthatforlastthreedecades80.8%ofthehotelclosuresaccountedfortheindependentlyownedhotels. Anotherchallengefortheindependentlyownedhotelisincreasingcompetition.STRUShotelpipelinereportsforJanuary,2012andFebruary,2013showedthatwithinthisperiodtherewasanincreaseintheroomsupplyby35,669rooms.Moreover,currently74,052ofthenewroomsareunderconstruction.Thenumberofnewprojectsincreasedby39.7%comparedto2012,whilein2012theincreasewasequalonlyto3.4%( STR 2012 2013 ).Consideringhighernancialrisksandextremecompetition,thehotelmanagersoftheindependentlyownedhotelsneedtohaveadequateforecastingtools.Theeffectofenvironmentfactorsmaybedetrimentalifthehotelcannotquicklyreacttothechangesinthemarket.Whilesalesforecastingiscrucial,someofthesophisticatedforecastingtoolsaredemandingintermsoftime,cost,andexpertise.Therefore,itisimportanttondforecastingmodelsthatmayassistthemanagersofindependentlyownedhotelstodeterminetheirfuturesales.However,consideringthe 23

PAGE 24

possibilityoflimitednancialresourcesandinsufcientskillsamongindependentlyhotelmanagers( Burgess 2007 ),themodelsalsoneedtobeaccessibleandrelativelysimpleinapplication. Theapplicationoftimeseriesmodelsmaybeapowerfultoolforhotelmanagerstoimprovetheirforecastingabilityinachangingenvironmentthatreectsindustrygrowth,volatileeconomicconditions,andchangingcustomerpreferences( Cranage 2003 ; Limetal. 2009 ).Suchmethodsassubjective,descriptivestatisticsandregressionanalysismaynotaccountforallconditionsthataffecthotelsales.Onthecontrary,timeseriesmodelsareabletodetectchangingpatternsinthedata,forexample,trends,cyclesandseasonalinuence( Andrewetal. 1990 ; Limetal. 2009 ).Whileeconometricmodelsandforecastingsolutionsaresuperiortothesesimpleandaccessiblemethods,theymayrequireadditionalinvestmentsandspecicexpertise.Incontrast,thetimeseriesmodelsthataregoingtobesuggestedrequireminimalexpertiseandcouldbeimplementedusingopensourcesoftware.Additionally,timeseriesmodelsincomparisonwithothercommonlyusedmethodswerefoundtoberelativelysimplisticinapplication( Cranage 2003 ; Cranage&Andrew 1992 ).Regardlessoftheirsimplicity,themodelsarestillbelievedtoprovidehotelmanagerswithrelativelyaccuratesalesforecasts( Aghazadeh 2007 ; Andrewetal. 1990 ; Cranage 2003 ; Lietal. 2006 ; Songetal. 2011 ). 1.4PurposeoftheStudy Thepurposeofthisstudyistotestthreetimeseriesmodelsthatcouldprovidehotelmanagerswithanaccuratesalesforecast.ThestudywillanalyzetheperformanceofthreepracticaltimeseriesmodelsbycomparisonofthefoundsalespatternsandgeneratedforecastswiththeactualhotelsalesdatathatwasacquiredfromadiverserangeofhotelsintheOrlando,Floridahotelmarket.Themodelsareconsideredpracticalbecausetheyarewidelyavailableforecastingtoolsandrequireminimalexpenditures, 24

PAGE 25

time,andexpertisetomakeanactualforecast.Inotherwords,themodelscouldbeappliedwithoutspecialassistanceandtheresultsarefairlyeasyformanagerswithlimitedstatisticalknowledgetointerpretandincorporateintomanagerialstrategies. Themodelstobetestedinthisstudyaretheadditiveseasonaldecomposition,theadditiveexponentialsmoothing,andtheARIMA(autoregressiveintegratedmovingaverage).Themodelswerepreviouslyutilizedforsalesforecastinginthehospitalityindustry( Aghazadeh 2007 ; Cranage 2003 ; Song&Witt 2000 ).Themodelsalsosatisfytheconditionofpracticality,becausetheyareaccessibleandrequireminimumexpertise.Additionally,thisstudyfocusesondeterminingthemodelthatgeneratesthemostaccuratesalesforecastsgiventhecurrentmarketsconditioninthelodgingindustry.Forpurposesofthisstudy,effectivenessisdenedastheabilityofthemodelstogiveanaccurateforecastandtheabilitytoreectpatternsintheavailabledataalongwithbeingpracticalforhotelmanagers.Anapplicationoftimeseriesmodelsforhotelsalesforecastingcouldgivenecessaryinformationforday-to-dayhoteloperations,suchasbudgetingofresources(e.g.,investments,andlaborexpenditures)( Aghazadeh 2007 ). Theadditiveseasonaldecompositionisasimpletimeseriesmodelthatprovidesageneralunderstandingaboutpastsalesdata.Themaindisadvantageofthistechniqueisthatitdoesnotproduceanyforecast.However,themodelisabletogiveaclearrepresentationofthesalesuctuations(i.e.,seasonality,trend,cycle,andrandomuctuations)thatmaybeusedbymanagerstoimprovetheirforecastingabilities.Therefore,thisstudywillinvestigatehowaccuratelythismodelresemblesthepatternsinthehotelsalesdatasuchastrend,cycle,andseasonality. Cranage ( 2003 )suggestedthewiderapplicationoftheexponentialsmoothingmodelswithinlodgingindustry.However,whencomparedtomoreelaboratedforecastingmodels,additiveexponentialsmoothingmaygeneratelessaccurateresults( Aghazadeh 25

PAGE 26

2007 ; Andrewetal. 1990 ).Themodelismoreappropriateforshorttermratherthanlongtermforecasting.Themodelhasalsoafewadvantagesthatneedtobetakenintoaccount.First,itissimpleinapplication,andsecond,therequirementoftimeandcostareminimal.Forexample,ahotelrevenuemanagermayutilizethemodeltodeterminethesalesfortheupcomingweekendusingsalesdatafromthepastseasons.Theforecastgeneratedbythemodelwillbecomparedtotheactualsalesdataforthelastyearwithintheacquireddatasetsinordertoidentifytheaccuracyofproducedforecastandifitdeviatesmoreattheendoftheobservedperiod. TheARIMAmodelisamorecomplexmodelwhencomparedtotheadditiveseasonaldecompositionandtheadditiveexponentialsmoothing.Theapplicationofthemodelinvolvesmorestepsthantheapplicationofothertwomodels.Thecomplexityofthemodelalsorequiresahigherlevelofexpertise.However,theavailabilityoffreesoftwarethatiscapableofrunningthemodelmakesARIMAamorepracticalandaccessiblemethodtothemanagersoftheindependentlyownedhotelsthentherevenuemanagementsolutionssoftwareusedbychainhotelsthatisexpensivetopurchaseandmaintain.Inaddition,theARIMAwasfoundtoproduceresultssuperiortoadditiveexponentialsmoothing( Chu 1998 ).Therefore,theforecastedvaluesproducedbytheARIMAmodelwillbecomparedwiththeforecastoftheadditiveexponentialsmoothingaswellastotheactualsalesdatainordertodeterminethemodelsapplicablevaluetoindependentlyownedhotelmanagers. Inthepreviouspassage,itwasmentionedthatavailablesoftwaremayassistinapplicationoftheproposedmodels.However,thecurrentstudywillgobeyondrecommendingtoutilizeanyavailablestatisticalsoftwareconsideringthefactthatnancialresourcesinindependentlyownedhotelsmaybelimited.Forthatreason,allthemodelswillbeappliedusingopensourcestatisticalsoftwareRProject.The 26

PAGE 27

softwareissuppliedwithmultiplemanualsthathavestep-by-stepinstructionshowtoimplementeachofthemodels.ThedatasetsthatwillbeutilizedinthecurrentstudywereobtainedfromadiverserangeofhotelslocatedinOrlando,FL.TheavailablehotelsalesdatarepresentedmonthlyADRindollarsfrom2007to2010.TheADR( 1 )isdeterminedbydividingtotalrevenuefromtherooms(RT)bythenumberoftheroomsthatweresold(NS). ADR=RT NS(1) Basedontheoverviewofthethreemodelsthefollowingresearchquestionswereformulated: Q1:Willtheadditiveseasonaldecompositionmodelreectaccuratepatternsofthehotelsalesdata? Q2:Doesadditiveexponentialsmoothinggenerateaforecastthatrepresentstheactualhotelsalesdata? Q3:DoestheARIMAperformbetterincomparisontotheadditiveexponentialsmoothingmodel? 1.5Limitations Oneofthemainlimitationsofthisstudyistherequirementofsalesdatatobestationaryandifthisrequirementisnotsatisedtheapplicationofthemodelsmaybecomecumbersomeandwillrequiremoretimeforthedatatransformation.AnotherlimitationisthecomplexityoftheARIMAmodel,eventhoughitmaybeimplementedinmostpopularsoftware,itrequiresmorestepsfortheapplicationandsomeexpertisetoidentifytheparametersofthemodelbeforeaforecastisgenerated.Inaddition,timeseriesmodelsingeneralarenotsensitivetosuddenchangesandinordertoadjusttheforecast,timeseriesmodelsneedtoincorporatemoredatapoints,otherwisetheforecastmaybeinaccurate Frechtling ( 2001 ). 27

PAGE 28

Besidesthelimitationsthatarerelatedtotheapplicationoftheproposedtimeseriesmodels,thereareotheradditionallimitationsregardingtheresultsofthisstudy.First,theapplicationofthemodelsandsubsequentcomparisonoftheresultswillbeconductedonthedatasetsacquiredfromOrlando,FLhotelmarketonly.Therefore,theresultsmaynotbevalidforotherhotelmarketsintheUSAandinternationalmarkets.Thislimitationmayariseduetodifferencesbetweenhotelmarkets;theeffectofspecicconditionsonhoteloperationsinoneparticularmarketmaybemagniedoronthecontrarynegligibleinanothermarket.Forexample,thehotelsalesofthenorthernstatesduringfallseason(e.g,Maine,NewHampshire,Vermont,Massachusetts,RhodeIsland,andConnecticut)maybeaffectedbythenumberoftouristswhocometoseeautumnfoliagewhichstartsfromSeptemberandcontinuesthroughNovember. Howard ( 2011 )reportedthatVermonthad3.6milliontouristswhospent$332millionduringfallof2011.Whileautumnfoliagecontinuestoattracttouriststothenorthernstates,hotelmarketsofsouthernstatesremainrelativelyunaffectedbythisspecicattributeoftheseasonchange. Anotherlimitationofthisstudyisthattimeseriesmodelsarerecommendedforapplicationwithinonesector,namelyindependentlyownedhotels.Thenumberofhotelclosuresandexposureofthissectortohighernancialriskssuggeststhatindependentlyownedhotelsmaylacktheresourcesneededforpurchasingofforecastingsoftwaresolutionswhileotheravailableforecastingtoolsmaybeineffectiveorinaccessible.Incomparison,largerchainhotelsmayhavemoreopportunitiesavailabletothemintermsofconsultingandabilitytopurchasesoftwaresolution;thus,theapplicationoftimeseriesmodelsmaybeirrelevanttothissector.Lastly,currentstudydoesnotutilizeanyconceptualframeworkduetopurelypragmaticpurposes. 28

PAGE 29

CHAPTER2REVIEWOFLITERATURE 2.1BackgroundInformation Detailedandaccurateforecastingiscrucialforallareasofhoteloperationsinordertooperatesuccessfullyinacompetitiveenvironment( Andrewetal. 1990 ; Limetal. 2009 ; Tony&Poon 2012 ).In2009,theindependenthotelsectorhadthesecondlowestpercentofroomsupplygrowth(i.e.,0.5%).While,thetotalsupplyofroomsownedbyindependenthotelsgrewbytheyearof2013,thetotalpercentofsalesgrowthremainedrelativelylow:1%( STR 2009b 2013 ).Giventheinformationregardingfuturesales(e.g.theamountofsalesduringtheweekendswhencomparedtoweekdays),hotelmanagersmayadjusttheirstrategiesaccordingly.Theadjustmentwillhelptoensurethecompetitivenessandnancialhealthofindependentlyownedhotelunitswiththeappropriateaccesstoforecastingtools. Whiteld&Duffy ( 2011 )discussedtheimportanceofsalesforecastingforconstantmonitoringofthehotelnancialperformanceandforformulatingmanagementdecisionsrelatedtofuturehoteloperations.Thesemanagerialdecisionsmayincludefuturepurchases,budgeting,andstafngstrategies( Tony&Poon 2012 ). Thebriefcomparisonofthechainandindependentlyownedhotelsisessential,inordertoexemplifytheuniquechallengesthatindependenthotelsmayencounterandtheurgeinsalesforecasting.Thechainhotelsaremorestablenanciallythanindependentlyownedhotels,becauseoftheabilitytoensurenancialsecuritytoallthepropertiesortodownsizethenumberofpropertieswhilestillremaininginthemarket( Mayock 2011b ).Onthecontrary,independentlyownedhotelsaresolelydependentontheirownnancialperformance.Theindependentlyownedhotelsaremorenanciallyvulnerablethanchainhotels.Thehigherrevenueacquiredfrommultiple 29

PAGE 30

propertiesallowschainhotelstodevelopmoreeffectiveforecastingmethods,forexampletomonitorthenancialperformanceofotherproperties.Inaddition,chainhotelscouldhavetheresourcesneededtoimplementandmaintainforecastingsolutionsgeneratedbyrevenuemanagementforecastingsoftwaresystems.Thisabilityisnotfrequentlyavailablefortheindependentlyownedhotels,becauseanyforecastingsolutionsgeneratedbyrevenuemanagementforecastingsoftwaresystemsrequireadditionalexpendituresthatcanbequiteexpensive.Moreover,withnoopportunitytoutilizemorepreciseforecastingsolutions,thepressurefordeningfuturesalesobjectivesrelyonlyonmanagersandtheirownperceptionoftheoperatingsituationandmarketingenvironment.Theoperatingenvironmentofthehotelindustryishighlycomplexaswellasthewholestructureoftheindustry( Croes&Semrad 2012 ; Okumusetal. 2010 ). Thelodgingindustryhasuniquecharacteristicsthatresultinnon-stationaryconditionsofthemarketslandscape.Thenon-stationaryconditionsthatarerepresentedbystochasticpropertiesinhistorichotelnancialdatainclude:theperishablenatureofthecoreproduct)]TJ /F1 11.955 Tf 12.63 0 Td[(room-nights,volatiledemandovertime,theabilitytomakeadvancedreservationsaswellaslastminutereservationsandwalk-inreservations,cleardistinctionofmarketsegments,relativelyxedcapacityofsupply,andtheappropriatecoststructurerepresentativeofhighxedcostsofoperationandlowassociatedmarginalcosts( Hanksetal. 1992 ; Kimes 1989 ).Suchconditionsrequirethathotelmanagersareabletogenerateaccurateforecastingmethodsinordertotrackthenancialpositionofthehotelandtomaintainortoimproveitsmarketpositionwithinthecompetitiveset.Ifanaccuratesalesforecastcannotbegeneratedbythemanagementteamofindependentlyownedhotelsitmayimpairtheabilityofthehoteltooperateathighlevelsofproductivity,therebyresultingintheincreasednancialrisk. 30

PAGE 31

Highproductivitylevels(e.g.increasedaveragedailyrate,revenueperavailableroom)areimportantforahoteltofunctionbecausecoststhatareassociatedwithlabor(i.e.salaries,wages,benets)comprise33%oftotalrevenueand43%ofalloperatingexpensesandarexed( Hu&Cai 2004 ).Ifthehotelmanagementteamisnotabletofunctionatahighproductivitylevelthermmayfacenancialjeopardy.Accordingto Baker&Riley ( 1994 ),theabilityofthehoteltomaintainahighproductivitylevelisdependentupontwoconditions.Therstconditionistheavailabilityofaccuratesalesforecasting.And,thesecondconditionistheabilitytomatchsupplyanddemand.Thesecondconditionmaybeachievedbyconstantadjustmentofneededresourcestomatchsupplyanddemand( Harnisch 2008 ).Forexample,arestaurantmaystafftheooraccordinglywiththeappropriateworkforcenumbersrequiredtoprovidetheservicetothecustomersneeds.However,inthelodgingindustryitbecomesamoredifcultmanagementtasktosupplytheresourcesrequiredtomeetconsumerneedswhentheconsumerdemandisuncertain,theroomsupplyisxed,andtheproductperishesifnotsoldthatday.Therefore,itbecomesevenmorecriticalformanagersofindependentlyownedhotelstohavetheabilitytoforecastfuturesalesliketheabilityofcorporatechainmanagementteams. Inorderforhotelmanagementtoproperlygaugeadesirableproductivitylevelandtocovercostsandoperatingexpensesthemanagementteamcannotonlypossessanunderstandingoftheconditionsinthemarketplace.Themanagementteammustalsobeabletomaximizerevenuesbothintheshortandlong-termsalesforecasts.Thismeansthattheuseofyieldmanagementbecomesnecessarywherethecentralfocusistocapturemaximumroomrevenueovertimethroughconstantadjustmentofroomratesbasedonuncertaindemand( Emeksizetal. 2006 ).Therefore,theabilitytoforecastanticipatedsalesisnecessaryforthenanciallongevityofalltypesofhotels. 31

PAGE 32

2.1.1CharacteristicsoftheYield-ManagementSystems Thesuccessofthehoteldependsuponitsabilitytoreducetherisksassociatedwiththecurrentandfutureoperationaldecisions,inparticularhowthedecisionswillaffecttheachievementoftheanticipatedresults( Frechtling 2001 ).Thehoteldataisdynamicanditisshapedbynumerouscharacteristicsinherenttothelodgingindustry(i.e.,externalandinternal)astheyield-managementsystem.Additionally,thesecharacteristicsmayrepresentmultiplechallenges.Variousauthorswhoseinterestsspantheareaoftourismdemand,tourismforecasting,salesforecasting,andyieldmanagementattemptedtodescribethecommoncharacteristicsoftheyield-managementsystems(i.e.,lodging,airlineindustries)andthissectionwillcompriseoftheirdifferentapproachestounderstandingoftheindustry'scomplexnature. Kimes ( 1989 )wasoneofthersttodescribethechallengeswhichhotelmanagersmayencounterwhenattemptingtomaximizerevenuefromhotelroomnightsales.Thechallengesthatshedescribedinvolvespeciccharacteristicsthatcouldbeobservedwithinthehotelindustry.Therstcharacteristicisperishabilityofthecoreproductthatisroom-nights.Perishabilityimpliesthatiftheroomwasnotsold,thenpossibleprotassociatedwiththisroomwaslostforever.Secondcharacteristicisabilitytosegmentmarkets,meaningthatmanagersmayadjustpricesbasedondifferentmarketsegments.Anothercharacteristicistheabilitytoselltheproduct(i.e.,room-nights)inadvancewhichcommonlyaccompaniedbyhighuncertainty.HighlyseasonaluctuationsofdemandalsorepresentthecharacteristicofthelodgingindustrysuggestedbyKimes.Roomdemandvariesbetweenweekdaysandweekends,aswellasbetweensummerandwintermonths;moreover,itmayhavedifferentpatternsintwosubsequentyears.ThelasttwocharacteristicsproposedbyKimesarelowmarginalsalescosts,andhigh 32

PAGE 33

marginalproductioncosts.Highmarginalproductioncostsarerelatedtothexedcapacityofthehotelandinabilitytoexpandroomsupplyonceitismet. Unlike Kimes ( 1989 ), Weatherford&Bodily ( 1992 )lookedmostlyatyield-managementsystemingeneralnotatthelodgingindustrywhentheyproposedtheircharacteristics.Authorsnamedrelativelyxedcapacity,perishabilityoftheproduct,andabilitytosegmentmarketsascommoncharacteristics.Weatherford&Bodilythenlookedatconstrainsforperishable-assetrevenuemanagersusingtheexampleoftheairlineindustry.Kimesexplainedthathotelandairlineindustrypossesssimilarcharacteristics,becausebothofthemareyield-managementsystems.Therefore,constrainsproposedbyWeatherford&Bodilywillbealsoreviewedinthecurrentchapter.Weatherford&Bodilysuggestedthatrstgroupofconstrainsisoperationalconstrainswhichincludexedcapacityandotheroperationalconstrains(e.g.,frequencyofightsinthecaseoftheairlineindustry).Anothergroupoftheconstrainsismarketingconstrains,inparticularlowlevelsofcustomerstoleranceinsomecasescausedbytheoverbookingpractices.Thethirdgroupsuggestedbytheauthorsisstrategicconstrains.Theseconstrainsmayincludediverseactionsfromthecompetitorsandadditionalconstrainsassociatedwiththeimplementationofthelong-termstrategies.Weatherford&Bodilyseparatedcharacteristicsoftheyield-managementsystemandconstrainstoitseffectiveoperation;however,itseemsthattheyareinterchangeable,inotherwordsconstrainsrepresentadditionalcharacteristicsofyield-managementsystems. Similarly, Frechtling ( 2001 )usedamacroapproachandreviewedpossiblechallengesofthetourismindustryratherthanindividualindustrieswithinit.Henamedthefollowingchallenges:theinseparabilityofcostumersfromtheproduction-consumptionprocess,thedependenceofoverallcustomerssatisfactionontheirsatisfactionwithcomplementaryservices,highsensibilityoftourismdemandonnaturalaswellason 33

PAGE 34

thehuman-madedisastrousevents,andtherequirementoflong-terminvestments.Thelodgingindustrybeingapartofthetourismindustrymayreectallofthesecomplexcharacteristics. Anotherperspectiveonthespecichotelindustrycharacteristicswasproposedby Hanksetal. ( 1992 ).Thesecharacteristicsorconditionsarelowvariablecosts,highxedcosts,perishableinventory,variabledemandpatterns,abilitytoforecastfuturedemand,andtosegmentcustomersbasedontheirneeds,behavior,andwillingnesstopay.Thecharacteristicsdescribedby Hanksetal. ( 1992 ),slightlydifferfromthecharacteristicsproposedby Weatherford&Bodily ( 1992 )and Kimes ( 1989 ),nonetheless,allofthemrepresentdeterminantsofthehotelindustry.Intotal,sixteenvariouscharacteristics(Table 2-1 )ofthelodgingindustrywerefoundbasedonthereviewofworksby Frechtling ( 2001 ), Kimes ( 1989 ), Hanksetal. ( 1992 ),and Weatherford&Bodily ( 1992 ). 2.1.2CharacteristicsoftheIndependentlyOwnedHotels Additionally,itseemsimportanttoreviewcharacteristicsthatarespecictoindependentlyownedhotelsasopposedtolargerchains.Firstcharacteristicisanabilitytoofferauniqueproductthatisnotstandardizedcomparedtothelargerchain;therefore,theperformanceofthehotelishighlydependentontheabilityoftheindependentlyownedhotelstoproposeanexclusiveservice/productorabetteroffer.Largerchainsarewell-known;therefore,thelackoftheinformationandlessrecognizablenameoftheindependentlyownedhotels( Sturmanetal. 2011 )mayaffecttheirnancialperformance.Thus,thelackofrecognitionsubstitutesthesecondcharacteristicoftheindependentlyownedhotels.Thirdcharacteristics,isthelackofnancialsecurity, Mayock ( 2011b )emphasizedthatindependentlyownedhotelsarelessnanciallysecurewhencomparedtothelargerchains.Healsoreferredtothedifferentqualityofofferedservicesandproductsacrosstheindependentlyownedhotelsthatmayimpact 34

PAGE 35

Table2-1. TheCharacteristicsoftheLodgingIndustry LodgingindustrycharacteristicsProposedbyauthor(s) 1Perishablecoreproduct Kimes ( 1989 ); Weatherford&Bodily ( 1992 ); Hanksetal. ( 1992 )2Abilitytosegmentmarkets Kimes ( 1989 ); Weatherford&Bodily ( 1992 ); Hanksetal. ( 1992 )3Theproductissoldinadvance Kimes ( 1989 )4Volatilenatureofthedemand Kimes ( 1989 ); Hanksetal. ( 1992 )5Lowmarginalsales Kimes ( 1989 )6Highmarginalproductioncosts Kimes ( 1989 )7Fixedcapacity Weatherford&Bodily ( 1992 )8Lowlevelsofcustomers'tolerance Weatherford&Bodily ( 1992 )9Dependenceontheactionsofthecompetitors Weatherford&Bodily ( 1992 )10Dependenceofalloperationsonthelongtermstrategies Weatherford&Bodily ( 1992 )11Theinseparabilityofcostumersfromtheproduction-consumptionprocess Frechtling ( 2001 )12Thedependenceofoverallcostumerssatisfactiononcomplementaryservices Frechtling ( 2001 )13Thedependenceofthedemandonthedisastrousevents(naturalorhuman-made) Frechtling ( 2001 )14Requirelong-terminvestments Frechtling ( 2001 )15Lowvariablecosts Hanksetal. ( 1992 )16Highxedcosts Hanksetal. ( 1992 ) thedemandofeachindividualproperty.Inordertoavoidtheuncertaintyassociatedwiththelackofexperiencewithaparticularindependentlyownedhotelandthelackofinformationaboutthehotel,customersmayprefertostayinachainhotelbecausetheyarefamiliarwiththebrand.Inconsistencyinthequalityreferstothesecondcharacteristicswhichisthelackoftherecognitionofahotel.Lastly,independentlyownedhotelsarecommonlysmallpropertieswhicharedependentontheirownperformance.Thisdependencemayresultinnancialinsecurityifhighproductivitylevelsarenotachieved.Theinsecurityintermsofnancialsupportrepresentsthethirdcharacteristicthatiscommontohotelsintheindependentlyownedsector. 35

PAGE 36

Inaddition,topreviouslyreviewedlodgingindustrycharacteristicsandspeciccharacteristicsoftheindependentlyownedhotels,thereareotherexternalfactorsthatinuencethelodgingindustry.Thesefactorsincludehighlycompetitivemarket( Enzetal. 2009 ),climaticchanges( Jang 2004 ; Songetal. 2011 ),economicuctuations( Susilovic&Sertic 2010 ),developmentofnewtechnologiesintheareaofhoteloperations,politicalchangesinthecountry,thelevelofattractivenessofadestinationtopotentialvisitors( Songetal. 2011 ).Besidesthepersistentinuenceofexternalandinternalcharacteristicsinthelodgingindustry,therearethetrendsinthetourismindustrythatdirectlyaffecthoteloperations.Thetrendsincludeincreasingcustomersophisticationandsubsequentincreaseincustomers'expectationsandglobalization.Thecomplexityofthelodgingindustryisapparentandhotelsrequireadequatemethodsforsalesforecastingtogainacompetitiveadvantageinthemarket.Especiallythisiscrucialforindependentlyownedhotelsconsideringtheirvulnerablenancialposition.Timeseriesmodelsareamongcurrentlyavailablemethodsandtheymaypossessadvantagesthatcouldmakethemsuperiortoothercommonlyusedmodelsandmethods. 2.2CommonlyUsedForecastingMethodsandModels Cranage&Andrew ( 1992 )examineddifferentforecastingmodelsthroughtheapplicationofthemodelstothehotelsalesdata.Thestudyoutlinedthreecommonlyusedmethodsforsalesforecasting,thesemethodsincludedthefollowing:judgmental,econometric,andtime-seriesmodels.Otheravailablemethodsforhotelsalesforecastingincludeapplicationofthedescriptivestatistics,regressionmodels,andapplicationoftheforecastingsoftwaresolution. 36

PAGE 37

2.2.1JudgmentalMethod Theresearchersdenedthejudgmentalmethodinsalesforecastingasamethodthatisbasedontheprice-settingapproachthatonlyuseshospitalitymanagers'experiencetoformulatefutureroomrates. Mentzer&Cox ( 1984 )foundthatjudgmentalmethodwasmorefrequentlyusedforsalesforecasting.Theyreferredtosuchmethodsassubjectivemethods,forexamplethedecisioncouldbemadebasedoncustomerexpectationsortheopinionoftheexecutiveboard.Thejudgmentalmethodhasbeenwidelyusedbythehotelmanagersintheireverydaypracticebecauseitisrelativelyinexpensive( Frechtling 2001 ).Themethodwasrecognizedbysomeresearchersasthemosteffectiveforhotelsalesforecasting( Cranage&Andrew 1992 ). However, Frechtling ( 2001 )suggestedthathotelmanagerssometimesmaynotpossessenoughknowledgeandexpertisetomakeanadequatedecisionregardingfuturehotelsales.Additionally, Burgess ( 2007 )studiedwhethermanagersinbothchainhotelsandindependentlyownedhotelshavesufcientnancialskillsinordertoeffectivelyoperatetheirproperties.Shefoundthatmanagerscouldbeunqualiedintermsofskillsandknowledge;therefore,managersdecisionsmaybeunreliableandmaynotrepresenttheactualoperationalsituation.Burgessemphasizedtheimportanceofaccurateinterpretationoftheavailableinformationbymanagersinordertomakeanaccurateforecast(i.e.,revenueandcosts). Cranage&Andrew ( 1992 )explainedthatmanyhospitalitymanagershavescarceresourcesthatcouldbeinsufcientforeffectiveutilizationofthejudgmentalmethod.Moreover,theyemphasizethattheaccuracyofthismethodisdoubtful,becausetherewaslittleresearchdoneinthisarea.Inaddition, Makridakisetal. ( 1998 ), Schwartz&Cohen ( 2004 ),and Croes&Semrad ( 2012 )echoedthenotionthatmanagersopinionsmaybebiasedandneedtobecautiouslyappliedtothedevelopmentofanyformofsalesforecasting. 37

PAGE 38

Therefore,thehotel'sfutureoperationswillbeadjustedaccordingtoinaccurateinformationthatdoesnotreecttheactualsituationinthemarket.Theavailabilityofmorepreciseandreliableforecastingmethodsormodelsisnecessary( Cassar&Gibson 2007 ).Onereasonwhysomemanagers,especiallythosemanagingindependenthotels,mayrelysolelyontheirexperienceandknowledgetoforecastfuturesalescouldbeduetothelackofresourcesfordevelopmentandimplementationofmoreprecisequantitativemodelsthatcouldgenerateaccuratesalesforecasts( Andrewetal. 1990 ; Steed&Gu 2009 ).Inotherwords,theymaynotpossesseithernancialresourcestopurchasesoftwareorsufcientprofessionalskillstoaccuratelyinterpretforecastsgeneratedbyrevenuemanagementsolutionsoftware. 2.2.2DescriptiveStatistics Thestudyby Kahn ( 1998 )showedthatdescriptivestatisticswaswidelyappliedforsalesforecasts.Suchstatisticalmeasurescouldbeausefultoolforcategorizingthedata.However,thismethodseemstobeinappropriateduetovolatilenatureofthedemand.Theaveragesmayfailtoaccuratelyrepresentseasonaluctuationsofthehotelsalesdata.Thus,ifmanagersutilizesuchmethodintheirpracticeitmayprovidethemwithvagueandmisleadinginformationregardingfuturesales.Consideringthatindependentlyownedhotelsareinseverecompetitionwiththelargerchains,amistakemadeinthepricingstrategiesmaybeterminal.Insummary,theapplicationofdescriptivestatisticsdoesnoteliminatetheuncertaintyofamarketsconditionsandmaynegativelyinuencesuccessivemanagerialdecisions. 2.2.3RegressionModels Regressionmodelswerefoundamongcommonlyappliedobjectivemethodsforsalesforecastsby Mentzer&Cox ( 1984 ).Aswasnotedinthepreviouschapter Song&Witt ( 2000 )foundthatregressionmodelsmaybeinappropriateifappliedtothe 38

PAGE 39

non-stationarydata. Limetal. ( 2009 )supportedthesendings;theysuggestedthattheregressionmodelsfailtogiveaccurateforecastsbecausetheymaynotaccountforthecomplexityofthelodgingindustry.Theauthorofcurrentstudyfoundthatthemultipleregressionmodelwithunemploymentrate,majorevent,economicindex,timeperiod,andseasonofyearvariablesexplainedonly57%ofvariationintheoccupancyratesofhotelslocatedinGainesville,FLmarket.Theother43%ofvariationoftheoccupancyrateswereleftunexplained.Whileforsomestudies57%maybeappropriate;however,consideringthatamistakeinpricingstrategiesmaybefatalforindependentlyownedhotels,thismodelseemstobeunacceptable. Moreover,theRsquaremaynotbeareliablemeasure,sinceitcouldbeinatedduetospuriousregression( Song&Witt 2000 ).Insummary,itseemsthenthattheregressionmodelsmayfailtoaccuratelyidentifyseasonalchangesofthedemand;thus,theforecastscouldsignicantlydeviatefromactualhotelsalesresultinginaspuriousregression.Thepossiblerepercussionsofspuriousregressionsonmanagerialdecisionsareincreasedcosts,inabilitytomeetthetargetedproductivitylevel,andinstabilityofthepositioninthemarket.Consideringthepossibilitythattherelationshipbetweenvariablesmaybespurious,thepriceadjustmentcouldleadtoaprotlossandincreasedexpenditures.Allofwhicharerepercussionsthatindependenthotelscannotaffordtomakewhencompetinginthemarketplacewithcorporatechainhotelsallwhiletryingtomaintainhighlevelsofproductivity. 2.2.4EconometricModels Anothermethodthatwasreferencedasbeingwidelyappliedinthehospitalityindustrywereeconometricmodelswherehotelmanagersmightrationallydeterminetherelationshipsbetweenthevariables,forexample,actualroomoccupancyratesandfuturehotelsales.Thesemodelshaveafewdisadvantages,theyaremorecomplex 39

PAGE 40

andneedtobeconstantlymonitoredandadjusted( Frechtling 2001 ).Although,otherforecastingmodelsalsorequirenewdatainputandadjustment,econometricmodelsneedlargerdatainputthanunivariatemodels( Wittetal. 2003 ).Moreover,econometricmodelscouldsometimesndspuriousrelationshipsbetweenthedependentandindependentvariables( Cranage&Andrew 1992 ).Inaddition,econometricmodelscouldbemorecostlycomparativelytoothercommonlyusedmethodsduetotimerequiredforthedatacollectionandsubsequentanalysis( Cranage&Andrew 1992 ).Lastly,hotelmanagersneedtohavespecicexpertisetoutilizethesemodelsintheireverydaymanagerialpractice. 2.2.5ForecastingSoftwareSolutions Theavailabilityofforecastingtechnologiesmadethemoneofthepopularmethodsforsalesforecasting( Moonetal. 2003 ). Moonetal. ( 2003 )explainedthatpreviouslytheforecastingtechnologieswerelimitedtoapplicationofonlyoneforecastingtechnique;whilenowadays,thetechnologiesofferachoicebetweendifferentforecastingtechniquesinordertoincreasetheaccuracyoftheforecasts.Thefollowingsolutionsarecurrentlyoffered:TheIDeaSForecastingManagementSystem,SaaSRevenueManagementSolutions,e.FlEXande.FLEXPlatinum,H-Enigma,SASForecastServer,ExecuvueHospitalityBusinessIntelligence,TargetvueHospitalityEnterprisePlanning,Nextgenerationbudgeting,forecastingandreporting,AmadeusRevenueManagementsolution,DataManagement&AnalyticsResource,andCRS-RMS.Despitetheavailabilityofmultiplesoftwaresolutionsthatcouldbeusedforsalesforecasting,Moon,Mentzer,andSmithconcludedthatforecastingaccuracyhasnotimproveddramaticallywiththeincreaseinthesophisticationofforecastingsolutions.Thendingssuggestthattherecouldbedifferentoperationalconstrainsthatinuenceimplementationofnew 40

PAGE 41

forecastingpracticesinthecompanies.Thus,moresophisticatedforecastingsolutionsarenotnecessarilymoreeffectivethanotherforecastingmethodsandmodels. 2.2.6TimeSeriesModels Cranage&Andrew ( 1992 ),and Cranage ( 2003 )proposedtimeseriesmodelsforutilizationinhotelsalesforecasting.Hotelsalesdataischaracterizedbyhighseasonality;therefore,thedataincludesseasonalpatternsthatmayreectchangingweatherconditions,theappearanceofsocialholidays,businessevents,andcalendareffects(i.e.,numberofdaysandweekendsinthemonth,season,year)( Frechtling 2001 ; Limetal. 2009 ).Theapplicationoftimeseriesmodelsmayenablehospitalitymanagerstodepictdatauctuations,suchastrends,cycles,and/orseasonalinuences.Themainadvantageoftimeseriesmodelsincomparisonwithothercommonlyusedmethodsandmodelsistheiraccessibility( Cranage 2003 ; Cranage&Andrew 1992 ).Inaddition,themodelscanproduceforecastclosetotheactualdata( Aghazadeh 2007 ; Andrewetal. 1990 ; Cranage 2003 ; Cranage&Andrew 1992 ; Lietal. 2006 ; Songetal. 2011 ). Ofparticularinteresttothisstudy,istheaccuracyandaccessibilitythatthesemodelsmayprovidetoindependentlyownedhotelmanagerswhenconsideringthathotelswithintheindependentlyownedhotellodgingclassmighthavelimitedresourcesavailabletothem(e.g.,insufcientnancialresourcestopurchaseforecastingsoftware,andinsufcientskillsandknowledgetouseandinterpretadvancedstatisticalforecastingprocedures).Moreover,aspreviouslymentioned,theindependentlyownedhotelsarefacingseverecompetitionfromthelargerchainhotelsbecausetheindependentlodgingclassaccountsforlessthan10%(i.e.,22,436hotels)intheUSA( Dahlstrometal. 2009 ; Mayock 2011a ).Thus,inorderforindependentlyownedhotelstoremaincompetitivewithintheUSlodgingindustrymanagersrequireforecastingtoolstoensure 41

PAGE 42

operationallongevityinthemarketplace.Theavailabilityandqualityoftheinformationregardingfuturesaleswillaffectthestufngdecisions,resourceallocationandcapacitymanagement( Songetal. 2011 ).Theapplicationofpragmaticforecastingmodelsmayassistinthedecisionmakingandminimizetherisksassociatedwithfuturehotel'snancialperformance. 42

PAGE 43

CHAPTER3METHODOLOGY Theavailabilityofeffectiveforecastingmethodsisimperativetothehotelswithintheircomplexoperatingenvironment.Specically,thisistrueforindependentlyownedhotels,becausetheyhavetocompetewiththelargerwell-knownbrands.Moreover,amistakeinthepricingstrategymaybefatalfortheindependenthotels.Although,everyhotelhasaneedinaccurateforecasting,independentlyownedhotelsmaynotpossesssufcientresourcesforimplementationofeconometricmodelsorsoftwaresolutionsanditssubsequentmaintenance.Therefore,managersbecomesolelyresponsibleforformulatingfutureroomrates,while Burgess ( 2007 )foundthathotelmanagersmaynothaveenoughprofessionalskilltofulllthisroleeffectively.Inadditiontodeciencyoftheresourcestoimplementforecasts,theseasonalityofhotelsalesmayimpairmanagersabilitytoobjectivelydeterminefuturesalesprospects.Thus,thereisanurgeinndingforecastingmodelsthatareeffectiveenoughtoproduceanaccurateforecastandrequireminimaleffortsintermsofexpenditures,time,andexpertiseasopposedtoforecastingsoftwaresolutionsandeconometricmodels. Therefore,timeseriesmodelscouldbeoneofthepossiblesolutionstoovercomethisobstacle.Thetimeseriesmodelsarecommonlyavailablemethodsforhotelsalesforecastingandtheymaypossessadvantagesthatmightgivethemasuperiorityovermorecomplicatedandresourcedemandingforecastingmethods(i.e.,econometricmodelsandforecastingsoftwaresolutions).Themodelsthataregoingtobetestedinthisstudyinclude:theadditiveseasonaldecomposition,theadditiveexponentialsmoothing,theARIMA(autoregressiveintegratedmovingaverage). 43

PAGE 44

3.1TheAdditiveSeasonalDecomposition Theseasonaldecompositionmodelisconsideredasabasictimeseriesmodelwhichbreaksdownthedataintofourcomponents:trend,seasonal,cyclicalandrandomcomponent.Commonlythetrendandcyclecomponentsarecombinedtogether( Hilasetal. 2006 ).Themodelmaybeappliedtoadatasetusingmostpopularsoftwarewhereitmayrequireafewstepsbeforeaclearrepresentationofthetrendandseasonalcomponentsisproduced.Theseasonaldecompositionisrepresentedintwoforms:additiveandmultiplicative.Themultiplicativemodelispreferredtotheadditivedecompositionifthedatahasanincreasingtrendandhaveatendencytouctuatemoreduringlatertimeperiods( Cowpertwait&Metcalfe 2009 ).Thehotelsalesdoesnotnecessarysatisfythesetwoconditions;thus,theadditiveseasonaldecomposition( 3 )mightbemoreappropriateforanalyzingofhotelsalesdata. Yt=TCt+St+Rt(3) whereYtistimeseries,TCtisthetrendandcyclecomponents,Stsubstitutesseasonalcomponentinthetimeseries,andRtstandsforrandomcomponent. Theaccessibilityandrelativesimplicityofthemodelaremainadvantagesincomparisontootherforecastingmethodsandmodels.Theadditiveseasonaldecompositiongivesanoverallunderstandingaboutpatternsindatathatmightbeusedtoformulatefuturesalesexpectations.However,themaindrawbackoftheseasonaldecompositionisthemodel'sinabilitytoproduceaforecast.Thislimitationcouldrestrictawiderapplicationofthemodelamonghotelmanagers.Consideringboththeadvantagesanddisadvantagesoftheadditiveseasonaldecompositiontheresearchquestionis: 44

PAGE 45

Q1:Willtheadditiveseasonaldecompositionmodelreectaccuratepatternsofthehotelsalesdata? H0:Theadditiveseasonaldecompositionmodelwillnotrepresentaccuratepatternsofthesalesdata. Ha:Theadditiveseasonaldecompositionmodelwillprovideanaccuraterepresentationofsalesdatapatterns. 3.2TheAdditiveExponentialSmoothing Theadditiveexponentialsmoothingismoreelaboratedcomparedtotheadditiveseasonaldecompositionandrepresentsanintermediatelevelforecastingmodel.Theadditiveexponentialsmoothingisabletodetecttwocomponentsofthesalesdata:trendandseasonal.Thereareseveralexponentialsmoothingmodels;however,onlyadditiveandmultiplicativemodelsareeffectiveforhotelssalesforecastingduetohighseasonality.Thisstudywillutilizeonlyadditivemodelconsideringtheresultsofthestudyby Andrewetal. ( 1990 ),and Yksel ( 2007 )wheretheyfoundthattheadditiveseasonalmodelhasperformedbetterthanthemultiplicativemodelintermsofhotelsalesforecasting.Theadditiveexponentialsmoothingmodelmaygenerateresultsthatareveryneartotheactualhotelsalesdata;therefore,maybesuperiortootherforecastingmodelsconsideringtheamountoftime,expertise,andexpendituresrequired( Witt 1992 ).Theadditiveexponentialsmoothingmodelisrepresentedbythefollowingequations( 3 3 3 3 ):Lt=(At)]TJ /F4 11.955 Tf 11.95 0 Td[(St)]TJ /F5 7.97 Tf 6.59 0 Td[(s)+(1)]TJ /F9 11.955 Tf 11.95 0 Td[()(Lt)]TJ /F7 7.97 Tf 6.58 0 Td[(1+bt)]TJ /F7 7.97 Tf 6.59 0 Td[(1) (3)b1=(St)]TJ /F4 11.955 Tf 11.96 0 Td[(St)]TJ /F7 7.97 Tf 6.58 0 Td[(1)+(1)]TJ /F9 11.955 Tf 11.96 0 Td[()bt)]TJ /F7 7.97 Tf 6.59 0 Td[(1 (3)St=(At)]TJ /F4 11.955 Tf 11.95 0 Td[(St)+(1)]TJ /F9 11.955 Tf 11.95 0 Td[()St)]TJ /F5 7.97 Tf 6.59 0 Td[(s (3)Ft=Lt)]TJ /F7 7.97 Tf 6.59 0 Td[(1+bt)]TJ /F7 7.97 Tf 6.59 0 Td[(1+St)]TJ /F5 7.97 Tf 6.59 0 Td[(s (3) 45

PAGE 46

whereequation 3 3 and 3 representleveloftheseries,trendandseasonalityaccordingly,whiletheequation 3 allowstocalculatethefutureforecast.Thevariablesanditscoefcientsaredenedasfollows,Lisleveloftheseries,islevelsmoothingconstantbetween0and1,Aisanactualvalue,sisthenumberofseasonalperiodsinayear(forexample,fourquarters,twelvemonths),bisthetrendoftheseries,seasonalsmoothingconstantbetween0and1,Sistheseasonalcomponent,seasonalsmoothingconstantbetween0and1,andtissometimeperiod; Theadditiveexponentialsmoothingmodelrequireslesstimeandexpertisethanothermoreelaborativeforecastingmodels,suchaseconometricmodels( Cranage 2003 ).However,theadditiveexponentialsmoothingmodelmaygeneratelessaccurateresultscomparativelytomoresophisticatedmodels.Thisnotionwassupportedinthearticlesof Andrewetal. ( 1990 )and Aghazadeh ( 2007 ).Therefore,thesecondresearchquestionwillinvestigate: Q2:Doesadditiveexponentialsmoothinggenerateaforecastthatrepresentstheactualhotelsalesdata? H0:Additiveexponentialsmoothingmodeldoesnotgenerateaforecastthatrepresentstheactualhotelsalesdata. Ha:Additiveexponentialsmoothingmodelgeneratesaforecastthatrepresentstheactualhotelsalesdata. 3.3TheAutoregressiveIntegratedMovingAverage(ARIMA) TheARIMAisanadvancedforecastingmodel;therefore,itisacomplexmodelthatrequiresmorestepsfortheimplementationcomparativelytotheadditiveseasonaldecompositionandtheadditiveexponentialsmoothing.However,accordingto Witt ( 1992 ), Kulendranetal. ( 1996 ),and Kulendran&King ( 1997 ),theARIMAmodelmayperformbetterthanevenmoredemandingeconometricmodels,whiletheexponential 46

PAGE 47

smoothingwasfoundtobeinferiortoeconometricmodelsinsomecases.Likewise,theARIMAwasfoundtoproducemoreaccurateresultsthantheexponentialsmoothing( Chu 1998 ).TheARIMAmaybeusedfortwelveoreighteenmonthsforecastsifitisbasedonthirtytoftyobservations( Frechtling 2001 ; Hilasetal. 2006 ).Themodelisrepresentedbythegeneralformula 3 ( As'ad 2012 ; Hilasetal. 2006 ): p(B)p)]TJ /F4 11.955 Tf 5.48 -9.68 Td[(BS(1)]TJ /F4 11.955 Tf 11.95 0 Td[(B)d)]TJ /F6 11.955 Tf 5.48 -9.68 Td[(1)]TJ /F4 11.955 Tf 11.96 0 Td[(BSDXt=#q(B)Q)]TJ /F4 11.955 Tf 5.48 -9.68 Td[(BSet(3) wheresistheseasonalityoftimeseries,Bisthebackshiftoperator, p(B)istheautoregressiveoperatoroforderp,disthedifferencingorderofanon-seasonaltimeseries,Xtisdataseries,tisatimeperiod,#q(B)isthemovingaverageoperatoroforderq,p(BS)isthenon-seasonalautoregressivecoefcientofniteorderp,Q(BS)isthenon-seasonalmovingaverageofniteorderQ,etisarandomerror. Baseduponthecomparisonofthetwomodels,thefollowingresearchquestionwillbeinvestigatedinthecurrentstudy: Q3:DoestheARIMAperformbetterincomparisontotheadditiveexponentialsmoothingmodel? H0:ThereisnodifferencebetweentheforecastproducedbytheARIMAmodelandtheadditiveexponentialsmoothing. Ha:TheARIMAwillproducemoreaccurateforecastcomparedwiththeadditiveexponentialsmoothingmodel. Theadditiveseasonaldecomposition,theadditiveexponentialsmoothingandtheARIMAmodelsdifferintermsoftheamountofeffortsrequiredtoimplementeachofthem.However,theavailabilityofreliableandpracticalsoftwaremayminimizethesediscrepancies.Theissueofinsufcientresourcesintheindependentlyownedhotels(i.e.,nances)mayberesolvedthroughapplicationoftheopensourcesoftwarethat 47

PAGE 48

satisestherequirementsofbeingreliableandpracticalatthesametime.Therefore,allthreemodelswillbeappliedinRProjectwhichisfreestatisticalsoftware.ThemainlimitationtotheapplicationofRprojectisthatitrequiressometimetogetfamiliarwiththeinterfaceandformatofthecommands;however,themanualsavailableforRgivestepbystepinstructionsontheuploadingofthedatalesandpreparationofthedataforthemodelsimplementation. ThedatathatwillbeutilizedinthecurrentstudywasobtainedfromamidscalehotelthatoperatesontheWaltDisneypropertyinOrlando,FloridawithanotherdatasetwhichwasacquiredfromSTR.ThedatasetobtainedfromSTRisrepresentativeofacompetitivesetofthehotelslocatedinOrlando,FLmarket.AccordingtoSTRglobalglossaryacompetitivesetisagroupofhotelsbywhichapropertycancompareitselftothegroupsaggregateperformance. TheARIMAandtheadditiveexponentialsmoothingmodelswillbeappliedtotheactualsalesdatafromthreeyears(i.e.,2007through2009)representedbymonthlyADRandthegeneratedresultsfor2010willbecomparedontheabilitytoreecttheactualADRs.Theseasonaldecompositionmodelwillbejudgedonoverallrepresentationofthepatternsinthedata.Themaindisadvantageoftimeseriesmodelsforforecastingistheirinabilitytotakeintoaccounttheeventsthatcoulddrasticallychangebehavioroftravelers(e.g.,naturaldisasters,mega-events)( Frechtling 2001 ).Thefailuretoreectsuddenchangesinthesalesdata,forexample,economicshocks,requiresmodelstoincorporatemoredatafromseverconsequentperiodstogenerateaccurateforecasts( Cranage 2003 ).However,beingmoreaccessiblethanotherforecastingmodelsandsolutionsandlessdemandingintermsoftheresourcesrequired,timeseriesmodelsmaygiveacompetitiveadvantagetoindependentlyownedhotels;therefore,shouldbeconsideredforitsconstantutilization. 48

PAGE 49

CHAPTER4APPLICATIONOFTIMESERIESMODELS 4.1PreparationoftheData Thedatathatwasusedinthestudycomprisedfromtwodifferentdatasets.OneofthedatasetswasacomprehensivedatareportthatwasprovidedbySTRforthelodgingmarketinOrlando,FLfortheyears2007to2010(Figure 4-1 ).STRcollectstheperformancedatafromlocalandinternationalhotelmarketswhichincludemorethan5.7millionrooms.STRoffersdiverseperformancereportsforbench-markingpurposes.ThesereportsincludesuchperformanceindicatorsasADR(averagedailyrate),RevPAR(revenueperavailableroom),supply,demand,revenue,etc.( Hood&Mandelbaum 2012 ).ThedatathatwasextractedfromtheSTRdatareportpertainedtotrendperformanceindicatorstatisticsfortheOrlandohotelmarket. TheseconddatasetwasfromanindependentlyownedmidscalehotelthatislocatedonWaltDisneyWorldproperty.Thesetcontainedproprietaryhistoricalnancialperformancedatafortheyearsof2007to2010(Figure 4-2 ).TheindividualunithotelsubscribestoSTRdatareports;therefore,theproprietarydataisrepresentedintheSTRaggregatestatisticalnumbersthatarereportedbySTRforallofthesectorsincludingindependentlyowned.BoththeSTRandtheindividualunithotelsdataincludedthefollowingvariablesfortheyearsof2007to2010:occupancyrates,ADR,andRevPAR.Analysisofbothdatasetsrevealedthattherewerenomissingvalues;therefore,noreplacementprocedureswereapplied. Thefouryearsofthetwodatasetsthenwereenteredintotheadditiveseasonaldecompositiontotesttheperformanceofthemodel.TheARIMAandtheadditiveexponentialsmoothingmodelsweretestedusingonlytherstthreeyearsfromeachdataset.Thefourthyearwasleftasabenchmarkthatwouldallowaperformance 49

PAGE 50

Figure4-1. ADRoftheSTRCompetitiveSetofOrlandoHotels(2007-2009) comparisonfortheARIMAandtheadditiveexponentialsmoothingmodels'results.InordertogenerateaforecasttheARIMAmodelrequiredtestingofthestationarityconditionsofthedata.Thetimeseriescouldbeconsideredasstationaryifitsmeanandvariancewereconstantovertime( Frechtling 2001 ; Limetal. 2009 ; Songetal. 2009 ).Theadditiveexponentialsmoothingmodelachievedstationaritybydifferencingtheactualvaluesbeforegeneratingaforecast.Thisprocedurewasaccomplishedbydifferencingtheprecedingobservationfromasuccessivevalue( Frechtling 2001 ).Thus,ifthetotalnumberofobservationswasequaltonthenafterapplicationofthedifferencingproceduretoachievestationaryconditions,thetotalnumberofobservationswouldben)]TJ /F6 11.955 Tf 13.11 0 Td[(1.Inotherwords,themodeldidnotrequireanypriorpreparationof 50

PAGE 51

Figure4-2. ADRoftheIndividualHotel(2007-2009) thedatasets.Onthecontrary,theARIMAdidnottreatthedataforstationarityastheadditiveexponentialsmoothingandrequiredtheinputdatatobeinthestationaryformwhenitwasenteredintothemodel. Therefore,bothdatasetsacquiredfromtheindividualhotelandtheSTRcompetitiveset,thatrepresentedtheaveragehotelperformance,hadbeenanalyzedonthesubjectofstationaritybeforethesetswereenteredintotheARIMA.ADRacquiredfromtheSTRcompetitivesetofhotelswasthesetofaverages;thus,wastreatedasstationary.Dependencebetweendatapoints(i.e.,non-stationarityofatimeseries)whichwouldreectmemoryintimeseriesdatasetcouldbeavoidediftheobservationswereaveragedacrossseveralcategories.Inotherwords,theSTRcompetitivedatasetwas 51

PAGE 52

enteredintotheARIMAmodelintheuntransformed(i.e.,initial)form.Theprocedurethatwasusedtoidentifyiftheindividualhoteldatawascharacterizedasstationarywastheunitroottest.UnliketheSTRcompetitivedataset,theseconddataset,whichrepresentedthesalesoftheindividualhotel,wasfoundtobenon-stationary. 4.2ResultsoftheUnitRootTests Therearetwounitroottestwhichcouldbeappliedtothetimeseriesdata:theAugmentedDickey-Fuller( Dickey&Fuller 1981 )andthePhillips-Perron( Phillips&Perron 1988 ).However,onlytheAugmentedDickey-Fuller(ADF)testwasused,becausethePhillips-Perronrequiredalargertimehorizonthanthedatathatwasavailableforthisstudy.Thedatawastestedforaunitrootusingrecommendedstatisticalsoftware-RProject.TheADFtestwasbasedonthefollowinghypothesis: H0:Thedatahasaunitroot;therefore,isnon-stationary. Ha:Thedatadoesnothaveaunitroot;thus,isstationary. AftertheADFtestwasconductedtoinvestigatethestationarityconditionsoftheindividualhoteldataset,itrevealedthatthedatawasnon-stationarywithp-valueequalto0.215.Thismeansthat,thedatapointswerenotfreefrominuenceoftheproceedingdatapoint.Inordertotransformthedataintonon-stationaryformitshouldbedifferenceduntilstationaritywasachieved( Song&Witt 2012 ).Firstdifferencingdidnotremoveaunitrootfortheindividualhoteldataset,becausetheacquiredp-valuewas0.336.Afterthedifferencingwascompletedforthesecondtime,thep-valuegeneratedbytheADFtestwasequalto0.174.TheADFtestappliedtothedatathatwasdifferencedthethirdtimeindicatedthatthedatasetwasstationary,becausethep-valuewasequalto0.01.Thus,stationarityofthetimeseriesdatawasachieved.Insummary,thedatawasdifferencedthreetimesinordertoachievethestationarystate. 52

PAGE 53

ThedifferencingprocedurehasbeencompletedinRProject.Oncethedatawasreadyforapplicationbothdatasetswereenteredintothemodels. 4.3ForecastingSoftware Theavailabilityofforecastingsoftwaresolutioninahotelisessentialduetointensivecompetitioninthemarket( Enzetal. 2009 )andthedependanceofahotel'snancialperformanceondiversefactors( Enzetal. 2009 ; Jang 2004 ; Songetal. 2011 ; Susilovic&Sertic 2010 ).However,independentlyownedhotelsmaynothavesufcientnancialresources( Mayock 2011b )tobuyandmaintainsuchsoftware.Therefore,consideringtheimportanceofminimizingtheexpensesassociatedwiththeintroductionofnewsoftwaretoahotel,Rprojectwasrecommendedfortheapplicationofproposedtimeseriesmodels. Rprojectisopensourcesoftwareandcouldbeaccessibleonacomputerwithaninternetconnection.AlthoughRprojectwillrequiresometimeforthefamiliarizationwiththeprogram'sinterfaceandtypeofthecommands,themainbenetofthissoftwareisitseasyaccessandnoexpense.Rprojectworkswithspecictypesoflesthatarerecognizedascommaseparatedvalues(CSV),thetransformationofanExcelletothisformatiseasilyaccomplished.Themanualsthatareprovidedfortheusersofthissoftwarehaveadetaileddescriptionofthestepsneededforthepreparationofthedataandtheuploadingofaletotheprogram.Additionally,themanualsthatgivetheinstructionsfortheapplicationofproposedtimeseriesarealsoavailable.Hotelmanagersofindependentlyownedhotelscouldaccessthesemanualsinordertoacceleratetheirlearningprocessofhowtoruntheproceduresusedinthisstudy.TheapplicationoftheARIMAmodelmaybeamorechallengingtasksincethemanagersneedtohavesomestatisticalexpertiseinordertoapplythemodelforsalesforecasting. 53

PAGE 54

However,RProjectprovidesseveralmanualsregardingapplicationoftimeseriesmodelsincludingtheARIMA,thatmayhelpmanagerstofamiliarizethemselveswiththismodel. TheanalysisoftheadditiveexponentialsmoothingandtheARIMAmodels'performancewasaccomplishedbytwomeans.First,theforecastedvalueswereplottedwiththeactualADRsfor2010yeartoanalyzeiftheforecastedADRsrepresentedtheactualsales.Second,thequalityofeachforecastwasdeterminedusingthemeanabsolutepercentageerror(MAPE)( 4 )andtherootmeansquarepercentageerror(RMSPE)( 4 ). MAPE=Pnt=0jAt)]TJ /F5 7.97 Tf 6.58 0 Td[(Ftj At n100(4) wherenisthenumberoftimeperiods,AtistheactualADR,FtistheforecastedADR,andtindicatesamonth.Accordingto Lewis ( 1982 ),theforecastsfallintofourgroupswhichareaccurate,good,reasonable,andinaccurateforecast.TheforecastcouldbeconsideredasaccurateifthevalueoftheMAPEislowerthan10%,goodforecastswillhavetheMAPEbetween10%and20%,theMAPEforreasonableforecastsmayfallwithin20%to50%range,andforecaststhathavetheMAPEhigherthan50%aredenedasinaccurate( Frechtling 2001 ; Lewis 1982 ).Inotherwords,thelowervalueoftheMAPEindicatesahigherqualityoftheproducedforecast.However,thecategoriesofferedby Lewis ( 1982 )cannotbetakenasanabsolutestandard( Frechtling 2001 );therefore,theMAPEneedstobeaccompaniedbyanothercriterion-theRMSPE,inordertodeterminethequalityofaforecast.TheadvantageofthiserrormeasureisthattheRMSPEtoleratessmallerrorsandcaptureslargeones( Frechtling 2001 ),ndingtheforecastthatmostsignicantlydeviatesfromactualvalues. RMSPE=vuut Pnt=0At)]TJ /F5 7.97 Tf 6.59 0 Td[(Ft At2 n100(4) 54

PAGE 55

ThecalculatedMAPEandtheRMSPEcouldbefoundinthefollowingsectionspertainingtotheperformanceoftheadditiveexponentialsmoothingandtheARIMAmodel. 4.4Findings 4.4.1PerformanceoftheAdditiveSeasonalDecomposition Theadditiveseasonaldecompositionmodelsegregatedtheinputdataintotrend,seasonal,cyclicalandrandomcomponent.Themodeldidnotproduceanactualforecast;therefore,thegraphicaloutputgeneratedbythismodelwasusedforcomparisonofthepatternsfoundbythemodelandthepatternsintheactualADRs.Thegraphicaloutputproducedonlythreedifferentcomponents,becausethismodelcombinedtrendandcyclecomponentstogether( Hilasetal. 2006 ).Additionally,thegraphicaloutputprovidedtheinitialdatathatwasusedfortheassessmentofthemodel'sperformance.Theadditiveseasonaldecompositionmodelwassuggestedinthestudyduetothemodel'spracticalityandaccessibility. Yt=TCt+St+Rt(4) whereYtistimeseries,TCtistrendandcyclicalcomponent,Stisseasonalcomponent,Rtisrandomcomponentinthetimeseries,andtissometimeperiod.Amongtwotypesofseasonaldecompositionwhichareadditiveandmultiplicative,themultiplicativeisrecommendedforapplicationiftheavailabledatahasanincreasingtrendanddeviatesmoreattheendoftheobservedperiod( Cowpertwait&Metcalfe 2009 ).Thehotelsalesdatamighthavediversepatterns,forexampleanincreasingordecreasingtrend;therefore,theadditiveseasonaldecompositionmodelwasfoundtobemoreappropriateforapplicationtotheavailablehotelsalesdatathanmultiplicative.Theinabilityofthemodeltoproducetheforecastmaybecounterbalancedbytheabilitytoclearly 55

PAGE 56

distinguishthepatternsinthedata;thus,theresearchquestionabouttheperformanceofthismodelwas: Q1:Willtheadditiveseasonaldecompositionmodelreectaccuratepatternsofthehotelsalesdata? Hypothesisthatweretestedbasedonthisresearchquestionare: H0:Theadditiveseasonaldecompositionmodelwillnotrepresentaccuratepatternsofthesalesdata. Ha:Theadditiveseasonaldecompositionmodelwillprovideanaccuraterepresentationofsalesdatapatterns. Inordertotestthehypothesespresentedabove,fouryearsfromeachofthedatasetswereenteredintothemodel.Theadditiveseasonaldecompositionproducedcleartrend,cycle,seasonalandrandomcomponentfortherst(Figure 4-3 )andtheseconddataset(Figure 4-4 ). TheadditiveseasonaldecompositionshowedthedecreasingtrendinsalesoftheSTRcompetitivesetofhotels(Figure 4-3 )andtheindividualhotel(Figure 4-4 ).Thetrendofbothdatasetsclearlypresentedtheeffectofeconomicdownturnof2008onthelodgingindustry.Althoughthedecreasingtrendcouldbedistinguishedbyanobservationoftheinitialdata,themodelproducedthedistincttrendofthisnegativechange.Theseasonalcomponentwasalsoclearlysegregatedfromtherestofuctuationsbythemodel.Figure( 4-3 )indicatedthattherewasadecreaseinsalesaroundtheendofeachyear;however,thenADRrecoveredandreachedthehighestpointwithinnexttwomonths.TheincreasewasfollowedbyanotherdecreasewhereADRfelltothelowestlevelduringtheyear(i.e.,thirdquarter)andthenthepatternrepeated. Thesalesoftheindividualhotelhadaslightlydifferentpattern(Figure 4-4 )whencomparedtothesalespatternoftheSTRcompetitivesetofhotels(Figure 4-3 ).The 56

PAGE 57

Figure4-3. TheAdditiveSeasonalDecompositionAppliedtotheSTRCompetitiveDataSet(2007-2010) individualhotelexperiencedaminorincreaseinsalesattheendofthesecondquarterofeachyearandthenADRdecreasedtothelowestpointasintheexampleofthecompetitiveset.Thelastcomponentthatadditiveseasonaldecompositiondetectedinthedatawastherandomcomponent.Incaseofhotelsalestherandomuctuationsmightrefertoamajoreventtakingplaceinthesurroundingareaorchangingweatherconditions,forexampleunusuallyhighorlowtemperatures. 4.4.2PerformanceoftheAdditiveExponentialSmoothing Theadditiveexponentialsmoothingmodelwaschosenforthestudyafterthereviewofworksby Cranage&Andrew ( 1992 ), Andrewetal. ( 1990 ),and Cranage ( 2003 )wheretheyadvocatedforapplicationofthismodelforsalesforecasting.Inaddition, 57

PAGE 58

Figure4-4. TheAdditiveSeasonalDecompositionAppliedtotheIndividualHotelDataSet(2007-2010) Andrewetal. ( 1990 ),and Wittetal. ( 1992 )foundthatwhilethemodelwasrelativelyeasyinapplication,theforecastsgeneratedbytheadditiveexponentialsmoothingcouldaccuratelydepictactualsales.Theforecastproducedbythemodel( 4 )includedinitialdata( 4 ),segregatedseasonal( 4 )andtrend( 4 )components.Lt=(At)]TJ /F4 11.955 Tf 11.95 0 Td[(St)]TJ /F5 7.97 Tf 6.59 0 Td[(s)+(1)]TJ /F9 11.955 Tf 11.95 0 Td[()(Lt)]TJ /F7 7.97 Tf 6.58 0 Td[(1+bt)]TJ /F7 7.97 Tf 6.59 0 Td[(1) (4)b1=(St)]TJ /F4 11.955 Tf 11.96 0 Td[(St)]TJ /F7 7.97 Tf 6.58 0 Td[(1)+(1)]TJ /F9 11.955 Tf 11.96 0 Td[()bt)]TJ /F7 7.97 Tf 6.59 0 Td[(1 (4)St=(At)]TJ /F4 11.955 Tf 11.95 0 Td[(St)+(1)]TJ /F9 11.955 Tf 11.95 0 Td[()St)]TJ /F5 7.97 Tf 6.59 0 Td[(s (4)Ft=Lt)]TJ /F7 7.97 Tf 6.59 0 Td[(1+bt)]TJ /F7 7.97 Tf 6.59 0 Td[(1+St)]TJ /F5 7.97 Tf 6.59 0 Td[(s (4) 58

PAGE 59

whereListhelevelformofseries,islevelsmoothingconstantthatcantakeanyvaluebetween0and1,Aisanactualvalue,sisnumberofseasonalperiodsinayear(forexample,fourquarters,twelvemonths),bisthetrendcomponent,isseasonalsmoothingconstantthatmaytakeavaluebetween0and1,Sistheseasonalcomponent,istheseasonalsmoothingconstantbetween0and1,andtissometimeperiod. Despiteoftheadvantagesofthismodel,theadditiveexponentialsmoothingmayproduceforecastinferiortoothermoreelaboratedforecastingmodels(i.e.,econometricmodels)( Aghazadeh 2007 ; Andrewetal. 1990 ).Therefore,theresearchquestionthatcurrentstudyinvestigatedwas: Q2:Doesadditiveexponentialsmoothinggenerateaforecastthatrepresentstheactualhotelsalesdata? Theresearchquestionwasaccompaniedbythenullandalternativehypotheses: H0:Additiveexponentialsmoothingmodeldoesnotgenerateaforecastthatrepresentstheactualhotelsalesdata. Ha:Additiveexponentialsmoothingmodelgeneratesaforecastthatrepresentstheactualhotelsalesdata. ThemodelwasfoundtoproducetheforecastrepresentativeoftheactualsalesaftertheSTRcompetitivedataset(Figure 4-5 )andtheindividualhoteldataset(Figure 4-6 )wereenteredintotheadditiveexponentialsmoothingmodel.TheblacklineonbothguresrepresentstheADRsfor2007-2009years,thegreenlineistheforecastproducedbythemodel,andtheredlineistheactualADRsfor2010usedforthecomparisonwiththeforecast.ThevaluesofADRforecastedbytheadditiveexponentialsmoothingcouldbefoundintable( 4-1 )andtable( 4-2 ). 59

PAGE 60

Figure4-5. TheAdditiveExponentialSmoothingForecastedADRfortheSTRCompetitiveDataSet(2010) Figure4-6. TheAdditiveExponentialSmoothingForecastedADRfortheIndividualHotelDataSet(2010) 60

PAGE 61

Table4-1. TheAdditiveExponentialSmoothing-ForecastedADRsfortheSTRCompetitiveSetofHotels(2010) JanFebMarAprMayJuneJulyAugSepOctNovDec 100.52107.10103.9598.1390.0684.4178.1972.9572.4488.8286.8584.27 Table4-2. TheAdditiveExponentialSmoothing-ForecastedADRsfortheIndividualHotel(2010) JanFebMarAprMayJuneJulyAugSepOctNovDec 81.4284.6280.8870.3658.9061.2554.6348.4950.1860.9157.5163.47 AlthoughatsomepointsthemodelprovidedtheforecastcloseorequaltotheactualADR,gure( 4-5 )showedthatthemodelmissedsomeoftheuctuations.Figure( 4-6 )alsodepictedthattheadditiveexponentialsmoothingwasnotabletoaccuratelyrepresentseveraluctuations,whileitcouldaccuratelyshowotherpatterns.TheforecastedvaluesoftheSTRcompetitivedataset( 4-5 )deviatedlesswhencomparedtothepredictedvaluesfortheindividualhoteldataset( 4-6 ).Inordertofurtheranalyzethetwoforecasts,theerrormeasurestheMAPEandtheRMSPEwerecalculated.TheMAPEoftherstforecastfortheSTRcompetitivesetwasequalto6.2%;thus,accordingto Lewis ( 1982 ),thisforecastwasaccurate.TheRMSPEmeasurewasalsofoundtobelowfortherstforecast(7%).TheMAPEofthesecondforecastfortheindividualhotelwasslightlyhigher-10.5%;therefore,whileitwaslessaccuratethantherstforecast,itstillcouldbeconsideredasagoodforecast( Lewis 1982 ).TheRMSPEofthesecondforecastproducedbytheadditiveexponentialsmoothingwasalsofoundtobehigherthanfortherstforecast-12.2%.DespiteofthefactthattheadditiveexponentialsmoothingforecastfortheSTRcompetitivedatasethadthelowerMAPEandRMSPE,itcouldnotrepresenttheuctuationsasaccuratelyastheforecastfortheindividualhotel. 61

PAGE 62

4.4.3PerformanceoftheAutoregressiveIntegratedMovingAverage(ARIMA)Model TheapplicationoftheARIMAmodelforsalesforecastmighthaveseveraladvantages.First,theperformanceofthemodelcouldbesuperiortoeconometricmodels( Kulendran&King 1997 ; Kulendranetal. 1996 ; Witt 1992 ).Second,itmightgeneratetheforecastofahigherqualitythanexponentialsmoothing( Chu 1998 ).Third,thismodelmayalsobeappliedonafreestatisticalsoftwareasothertwomodels,whichmakesitaccessibletothehotelmanagersofindependentlyownedhotels.Themodel( 4 )determinedthepatternsinthedataandbasedonthisARIMAproducedaforecast. p(B)p)]TJ /F4 11.955 Tf 5.48 -9.69 Td[(BS(1)]TJ /F4 11.955 Tf 11.95 0 Td[(B)d)]TJ /F6 11.955 Tf 5.48 -9.69 Td[(1)]TJ /F4 11.955 Tf 11.96 0 Td[(BSDXt=#q(B)Q)]TJ /F4 11.955 Tf 5.48 -9.69 Td[(BSet(4) wheresistheseasonalityoftimeseries,Bisthebackshiftoperator, p(B)isautoregressiveoperatoroforderp,disorderofdifferencingnon-seasonaltimeseries,Xtistimeseries,tissometimeperiod,#q(B)isthemovingaverageoperatoroforderq,p(BS)isnon-seasonalautoregressivecoefcientofniteorderp,Q(BS)isnon-seasonalmovingaverageofniteorderQ,etisrandomerror. InastudythatassessedtheperformanceoftheARIMAandtheexponentialsmoothingmodel,theARIMAforecastwasfoundtobesuperiortotheforecastoftheexponentialsmoothingmodel( Chu 1998 ).Thus,thecurrentstudyinvestigatedthefollowingresearchquestionandhypothesiswithregardstotheperformanceoftheARIMAmodel: Q3:DoestheARIMAperformbetterincomparisontotheadditiveexponentialsmoothingmodel? H0:ThereisnodifferencebetweentheforecastproducedbytheARIMAmodelandtheadditiveexponentialsmoothing. 62

PAGE 63

Ha:TheARIMAwillproducemoreaccurateforecastcomparedwiththeadditiveexponentialsmoothingmodel. Theblacklineonthegraphicaloutputforbothforecastsindicatesactualsalesofthecompetitivesetandtheindividualhotelfor2007-2009,redlinerepresentsactualsalesofthehotelsfor2010,andthegreenlineistheforecastgeneratedbytheARIMAforeachofthedatasets.TherstforecastfortheSTRcompetitiveset(Figure 4-7 )producedbythemodeldeviatedfromtheactualsalesatsomedatapointsmorethan$10-$20;therefore,inthiscase,theARIMAgeneratedaforecastinferiortotheforecastproducedbytheadditiveexponentialsmoothing(Figure 4-5 ).ThepredictedvaluesfoundbytheARIMAmodelarerepresentedinTable( 4-3 ). Table4-3. TheARIMA-ForecastedADRfortheSTRCompetitiveSet(2010) JanFebMarAprMayJuneJulyAugSepOctNovDec 96.6099.55101.67107.18106.01104.73106.81103.32101.31102.8899.7898.99 Table4-4. TheARIMA-ForecastedADRfortheIndividualHotel(2010) JanFebMarAprMayJuneJulyAugSepOctNovDec 73.1065.6572.8971.0661.8964.5864.1758.1158.5569.9964.8277.04 SecondforecastoftheARIMAmodelfortheindividualhotel(Figure 4-8 )onthecontraryaccuratelydepictedallofthesalespatternsof2010year;thus,providingtheforecastsuperiortotheforecastoftheadditiveexponentialsmoothing(Figure 4-6 ).However,beforetheactualADRnumbersof2010couldbepredictedbytheARIMAmodel,theforecastedvaluesweretransformedfromthedifferencedformintotheinitialform.Accordingto Frechtling ( 2001 ),totransformtheforecastthatwasproducedbasedondifferencedvalues,theforecastshouldbeaddedtotheactualvaluesforthesameperiodoftime.TheacquiredADRispresentedinTable( 4-4 ). 63

PAGE 64

Figure4-7. TheARIMAForecastedADRfortheCompetitiveSetofHotels(2010) Figure4-8. TheARIMAForecastedADRfortheIndividualHotel(2010) 64

PAGE 65

ThequalityofeachforecastwasadditionallymeasuredusingtheMAPEandtheRMSPE.TheMAPEandtheRMSPEoftherstforecastwereequalto14%and17%accordingly.ThesendingsindicatedthattheforecastoftheARIMAmodelfortheSTRcompetitivesethadapoorerqualitythantheforecastproducedbytheadditiveexponentialsmoothing.Ontheotherhand,thesecondforecastfortheindividualhotelhadtheMAPEandtheRMSPEequalto2.7%and2.8%,thefounderrormeasureswerelowerthantheMAPEandtheRMSPEoftheforecastfortheindividualhotelproducedbytheadditiveexponentialsmoothing.Thus,thesecondforecastfortheindividualhoteloftheARIMAmodelwasmoreaccuratewhencomparedtothesecondforecastoftheadditiveexponentialsmoothing. 4.5SummaryoftheFindings Thechapterinvestigatedthreesetsofresearchquestionsandthehypothesesregardingtheperformanceoftheadditiveseasonaldecomposition,theadditiveexponentialsmoothing,andtheautoregressiveintegratedmovingaverage(ARIMA)models(Table 4-5 ).Theadditiveseasonaldecompositionwastestedontheabilityofthemodeltoresemblethepattersthatcouldbeobservedintheactualsalesdata.Theadditiveexponentialsmoothingwasassessedbasedontwoforecastsproducedforeachofthedatasets,andtheperformanceoftheARIMAmodelwascomparedwiththeperformanceoftheadditiveexponentialsmoothing. Thersttworesearchquestionsandthealternativehypothesesfoundthesupportafterthedatawasimplementedusingtheadditiveseasonaldecompositionandtheadditiveexponentialsmoothingmodels.Theadditiveseasonaldecompositionclearlydisaggregatedthedataintotrend,cyclical,seasonal,andrandomcomponents(Figure 4-3 4-4 ).TheadditiveexponentialsmoothinggeneratedtheforecastedvaluesthatrepresentedtheactualADRs(Figure 4-5 4-6 ).Unlikethersttwomodels,theARIMA 65

PAGE 66

Table4-5. TheResearchQuestionsandHypothesesInvestigatedintheStudy ModelResearchQuestionHypothesesEquation(s) TheadditiveQ1:WilltheadditiveH0:TheadditiveseasonalseasonalseasonaldecompositionmodeldecompositiondecompositionwillnotrepresentYt=TCt+St+Rt( 4 )modelreectaccuratepatternsaccuratepatternsofthesalesdata.ofthehotelsalesHa:Theadditiveseasonaldata?decompositionmodelwillprovideanaccuraterepresentationofsalesdatapatterns. TheadditiveQ2:DoesadditiveH0:AdditiveexponentialLt=(At)]TJ /F5 7.97 Tf 8.47 0 Td[(St)]TJ /F12 5.978 Tf 5.76 0 Td[(s)+(1)]TJ /F16 7.97 Tf 8.47 0 Td[()(Lt)]TJ /F15 5.978 Tf 5.75 0 Td[(1+bt)]TJ /F15 5.978 Tf 5.76 0 Td[(1)( 4 )exponentialexponentialsmoothingmodeldoesnotsmoothingsmoothinggenerateaforecastb1=(St)]TJ /F5 7.97 Tf 8.47 0 Td[(St)]TJ /F15 5.978 Tf 5.76 0 Td[(1)+(1)]TJ /F16 7.97 Tf 8.46 0 Td[()bt)]TJ /F15 5.978 Tf 5.76 0 Td[(1( 4 )generateaforecastthatrepresentstheactualthatrepresentshotelsalesdata.St=(At)]TJ /F5 7.97 Tf 8.47 0 Td[(St)+(1)]TJ /F16 7.97 Tf 8.47 0 Td[()St)]TJ /F12 5.978 Tf 5.76 0 Td[(s( 4 )theactualhotelHa:Additiveexponentialsalesdata?smoothingmodelgeneratesFt=Lt)]TJ /F15 5.978 Tf 5.76 0 Td[(1+bt)]TJ /F15 5.978 Tf 5.76 0 Td[(1+St)]TJ /F12 5.978 Tf 5.75 0 Td[(s( 4 )aforecastthatrepresentstheactualhotelsalesdata. TheARIMAQ3:DoestheARIMAH0:Thereisnodifference p(B)p)]TJ /F5 7.97 Tf 3.88 -6.42 Td[(BS(1)]TJ /F5 7.97 Tf 8.47 0 Td[(B)d)]TJ /F7 7.97 Tf 3.88 -6.42 Td[(1)]TJ /F5 7.97 Tf 8.47 0 Td[(BSDXt=performbetterbetweentheforecastincomparisontoproducedbytheARIMAmodel#q(B)Q)]TJ /F5 7.97 Tf 3.88 -6.41 Td[(BSet( 4 )theadditiveandtheadditiveexponentialexponentialsmoothing.smoothingmodel?Ha:TheARIMAwillproducemoreaccurateforecastcomparedwithadditiveexponentialsmoothing. producedcontroversialresults.TherstforecastofthemodelfortheSTRcompetitivesetwasfoundtobepoorerinquality(Figure 4-7 )thantheforecastoftheadditiveexponentialsmoothing(Figure 4-5 ).However,thesecondforecastfortheindividualhoteloftheARIMAmodel(Figure 4-8 )wassuperiortotheforecastoftheadditiveexponentialsmoothing(Figure 4-6 ).Therefore,therewasnotenoughevidencetorejectorfailtorejectthenullhypothesis,thattherewasnodifferenceintheforecastsproducedbytheARIMAandtheadditiveexponentialsmoothingmodel. 66

PAGE 67

CHAPTER5CONCLUSIONSANDIMPLICATIONSOFTHESTUDY Thelodgingindustryisahighlycomplexenvironment( Croes&Semrad 2012 ; Okumusetal. 2010 )whichrequiresthehotelstodevelopandadaptadequatemanagerialstrategiesinordertoremaincompetitive.Particularly,thepositionofthehotelishighlydependentuponthesensitivityoftheproposedpricingstrategiestothechangesintheoperatingenvironmentandconsumerbehavior.Theprofessionalismandexperienceofthemanagersiscrucialfortheadjustmentofthepricingstrategies( Aghazadeh 2007 );however,thecomplexityoftheoperatingenvironmentandinuenceofunderliningtrendsmayimpairthemanagersabilitytoaccuratelyforeseethefuture. Thedevelopmentofdiverseforecastingsoftwaresolutionshasopenedupanopportunityforthehotelmanagerstosecurethehotel'spositioninthemarket.Yet,theforecastingsoftwaresolutionmaynotbewidelyavailableduetoexpensesassociatedwithit.Independentlyownedhotelsunlikechainhotelsarelessstablenancially( Mayock 2011b );therefore,maynotpossesstheresourcesinordertopurchaseandincorporatetheforecastingsolutionintheireverydaymanagerialpractice.Withoutareliableforecastingmethod,independentlyownedhotelsmaybecomemorevulnerablewhentheycompeteagainstlargerchains.Thestudyhasreviewedpossiblealternativemethodsavailabletoindependentlyownedhotelsforsalesforecasting.Thefoundalternativesincludedjudgementalmethod,applicationofdescriptivestatistics,regressionmodels,econometricmodels,andtimeseriesmodels.However,suchapproachasjudgementalmethod,descriptivestatistics,andregressionmodelsmightnotfullyaccountfortheseasonalityofthehotelsales( Croes&Semrad 2012 ; Limetal. 2009 ; Song&Witt 2000 ).Whileeconometricmodelswerefoundthemostappropriateinmanagingtheseasonaldata( Cranage 2003 ; Cranage&Andrew 1992 ; Semrad 2010 ), 67

PAGE 68

whencomparedtoothertimeseriesmodels,econometricmodelsaremoredemandingintermsoftimeandexpertise. Theadditiveseasonaldecomposition,theadditiveexponentialsmoothing,andtheARIMAmodelswereamongtherecommendedtimeseriesusedinthisstudybecauseoftheirpracticalityandaccessibility.Theadditiveseasonaldecompositionmodeldoesnotproduceaforecast;nonetheless,themodelisableofdistinguishtrend,cycle,seasonal,andrandomcomponentsfromtheinitialdata.Thus,theresearchquestioninvestigatedinthisstudypertainedtothepatternsthatthemodelfoundintheinitialdata.Particularly,ifthesepatternswererepresentativeoftheactualsales. Theadditiveexponentialsmoothingwastestedontheabilityofthemodeltogenerateanaccurateforecastusingtwoavailabledatasets.Lastly,thestudyinvestigatediftheARIMAmodel'sforecastwasmoreaccuratewhencomparedtotheadditiveexponentialsmoothingmodel.Thesuggestedtimeseriesmodelsvaryfromsimple(theadditiveseasonaldecomposition)toanadvancedlevelmodel(theARIMA);however,theutilizationofopensourcesoftwarecouldmakethemmoreaccessibleandlessdemandingintermsoftimeandexpertise.Thus,Rprojectwasutilizedinthestudyinordertotestthemodels. ThemodelswereappliedtotwodatasetsinR,whichincludedthemonthlyADRsobtainedfromSTRcompetitivesetofhotelsandtheindividualmidscalehotelthatoperatesintheOrlando,FLmarket.Sincetheadditiveseasonaldecompositionmodeldidnotgenerateanyforecast,themodel'sperformancewasanalyzedbasedonthepatternssegregatedfromthedata.Theadditiveexponentialsmoothingmodel'sperformancewasevaluatedusingthegraphicalcomparisonoftheproducedforecastwiththeactualvalues.Inaddition,theMAPEandtheRMSPEwerefoundinordertodeterminethequalityoftheforecast.Unliketheadditiveexponentialsmoothing,the 68

PAGE 69

ARIMAmodelrequiredspecicpreparationofthedatabeforegeneratingaforecast.Theindividualhoteldatasetwasfoundnon-stationary,toachievethestationarityformofthedata,thegivenvaluesweredifferenced.TheADFtestshowedthatthedatawasstationaryafterthethirddifferencingwascompleted.Oncetheindividualhoteldatawastransformedintostationaryform,theARIMAmodelwasappliedtoproduceaforecastforeachofthedatasets.Theanalysisofthemodel'sperformancewasconductedalsobygraphicalcomparisonoftheobtainedvaluesandusingtheMAPEandtheRMSPEmeasures. 5.1Conclusions Theadditiveseasonaldecompositionaftertheapplicationtotheavailabledatasets,wasabletodisaggregatethedataintodistincttrendandcycle,andseasonalpatterns(Figure 4-3 4-4 ).Inaddition,themodeldistinguishedtherandomuctuationsthattookplaceduringtheobservedperiodoftime(i.e.,2007-2010).Therefore,thenullhypothesisthatthismodelwouldnotgiveanaccuraterepresentationofsalesdatapatternshasbeenrejectedafterthegeneratedgraphicaloutputwasanalysedandcomparedtotheinitialsalesdata.Theadvantagesoftheapplicationoftheadditiveseasonaldecompositionforsalesforecastingarethesimplicityofthemodelandtheabilityofthemodeltoaccuratelydistinguishthesalespatters(e.g.,adecreasingtrend,theperiodswithlowsales). Theadditiveexponentialsmoothingwasalsofoundtoberelativelyeasyinapplication.AlthoughthemodelmissedafewuctuationswhenitprovidedtheforecastsfortheSTRcompetitivedatasetandtheindividualhoteldataset(Figure 4-5 4-6 ),overalltheadditiveexponentialsmoothingproducedthehighqualityforecasts.TheMAPEandtheRMSPEoftheforecastfortheSTRcompetitivedatasetwere6.2%and7%accordingly.TheforecastfortheindividualhotelhadtheMAPE-10.5%andthe 69

PAGE 70

RMSPE-12.2%.Thus,thenullhypothesishasbeenalsorejected.Thehypothesisstatedthatthemodelwouldnotgeneratetheforecastwhichrepresentedtheactualsalesdata.Similarconclusionsregardingtheperformanceofthemodelwereacquiredby Andrewetal. ( 1990 ),and Wittetal. ( 1992 ). TheapplicationoftheARIMAmodelrequiredingeneralmoretimethanothertwomodels,becauseofthepriorpreparationofthedataandthensubsequenttransformationoftheforecastedvaluestotheinitialform.TherstforecastproducedbytheARIMAmodelfortheSTRcompetitivedatasetdeviatedmorethantheforecastoftheadditiveexponentialsmoothingforthesamedataset(Figure 4-5 4-7 ).TheacquiredMAPE(14%)andtheRMSPE(17%)werenotablyhigherwhencomparedtotheMAPE(6.2%)andtheRMSPE(7%)oftheadditiveexponentialsmoothing.TheperformanceoftheARIMAmodelontheSTRcompetitivesetshowedthattheforecastwasnotsuperiortotheforecastoftheadditiveexponentialsmoothing.Onthecontrary,theforecastoftheARIMAperformedontheindividualhoteldatasetwasfoundtobemoreaccuratethantheforecastpreformedbytheadditiveexponentialsmoothing(Figure 4-6 4-8 ).TheMAPEandtheRMSPEoftheARIMAforecastwereequalto2.7%and2.8%.However,sincethemodelproduceddivergentresults,nonalconclusioncouldbemadeinfavororagainstthealternativehypothesis. 5.2LimitationsoftheStudy Thestudyhasseverallimitationspertainingtoutilizeddatasets,software,suggestedmodels,andtheresultsacquiredafterapplicationofthemodels.ThedatasetswhichrepresentedtheSTRcompetitivesetofhotelsandtheinformationfromtheindividualmidscalehotelwerelimitedtoonlyonehotelmarket-Orlando,FL.Therefore,theresultsofthisstudymaybeinvalidforanotherhotelmarkets.Consideringthecomplexityofthelodgingindustry( Croes&Semrad 2012 ; Okumusetal. 2010 ),it 70

PAGE 71

becomescrucialtounderstandthatwhilehotelmarketsmaysharesomeofthesimilarcharacteristics,othercharacteristicspertainingtothecompetitivelandscapeandthemarketplacethehotelsoperateinmaydiffersignicantly. TheimplementationofthemodelshasbeenconductedinRprojectwhichisopensourcesoftware.Itwassuggestedthattheavailabilityofthefreeforecastingsoftwarecouldhelptoavoidexcessivemoneyexpenditures,especiallythisaspectisimportantifthenancialresourcesarescarce.Thisisdenitelyanadvantageoftheapplicationoffreeforecastingsoftwaresolution;nevertheless,Rprojectwouldrequiretheinvestmentoftimeinordertobeabletoapplytheproposedmodelsonaneverydaybasis.Themanagerswouldneedtofamiliarizethemselveswiththeprocedurestoprepareanduploadthesalesdataintotheprogram,aswellaswiththecommandstorunthesuggestedtimeseriesmodels. Amongproposedtimeseriesmodels,theARIMAwasanadvancedtimeseriesmodelwhichwasmoredemandingintermsoftimeandexpertisethantheadditiveseasonaldecompositionandtheadditiveexponentialsmoothingmodels.Themodelrequiredthestationarityoftheinputdataunlikeothertwomodels;thus,thetransformationoftheinitialdataoftheindividualhoteldatasettothestationaryformwasaccomplished.Inaddition,theARIMArequiredsomestatisticalknowledgeinordertodeterminetheparametersofthemodel.TheforecastedvaluesgeneratedbytheARIMAwereinthedifferencedform;hence,thetransformationoftheforecasttotheactualvalueswasalsoneeded.Lastly,currentstudycouldnotndsufcientsupporttothesuperiorperformanceoftheARIMAwhencomparedtotheadditiveexponentialsmoothing.Asaresultthestudycannotadvocateforawiderutilizationofthemodelforthesalesforecastingbytheindependentlyownedhotels. 71

PAGE 72

WhilsttheapplicationoftheadditiveseasonaldecompositiondidnotrequiremultiplestepsasinthecaseoftheARIMAmodel,themodeldidnotproduceaforecast.Theadditiveseasonaldecompositiondisaggregatedclearlythepatternsfromtheinitialdata;however,themodelonlyshowedcurrenttrendandseasonalpatterns.Thus,theadditiveseasonaldecompositionneedstobeaccompaniedbyanotherforecastingmodelthatisableofdeterminingthefuturesales. Theadditiveexponentialsmoothingonthecontrarycouldgenerateanactualforecastandatthesametimethemodelwasalsosimpleinapplication.Thelimitationtotheapplicationofthismodelisthatmodelassumesthatthepasteventswilltakeplaceinthefuture( Armstrong 1985 ).Nonetheless,thismaynotbethecaseinthehotelindustrywiththevolatiledemandpatterns.Therefore,theforecastoftheadditiveexponentialsmoothingmodelcouldbenegativelyaffectedbytherapidlychangingtrendandseasonalpatterns.Thislimitationisvalidforalltimeseriesmodels.Inordertoproduceanaccurateforecastthemodelsrequiretheinputofmoredatathatincludesthechangingpatterns( Frechtling 2001 ). Timeseriesmodelsmightbeaneffectivemethodforsalesforecastingwithintheindependentlyownedhotelsector.Nevertheless,largerchainscouldaffordmoreelaborateforecastingsoftwaresolutions.Thus,thendingsofthestudymaybeirrelevanttothissector;however,thistopicneedstobeadditionallyexplored.Thendingsalsorelatedonlytothetestedtimeseriesmodels;hence,couldnotbegeneralizedtoothertimeseriesmodels.Lastly,thestudyhadapurelypracticalpurposeanddidnotincorporateatheoreticalframework. 5.3ImplicationsfortheIndustryPractitioners Amongstthereviewedtimeseriesmodels,thestudyfoundthatonlytwomodelsperformedaccuratelywhenappliedtothesalesdata.Thesemodelsaretheadditive 72

PAGE 73

seasonaldecompositionandtheadditiveexponentialsmoothing.Theadditiveseasonaldecompositionmodelcouldassistmanagersindeterminingcurrentsalespatterns,suchastheseasonaluctuationsandtrend.However,itissuggestedthattheadditiveseasonaldecompositionshouldbeaccompaniedbytheadditiveexponentialsmoothingmodel,becausetheadditiveexponentialsmoothingisabletoproduceaseeminglyhighqualityforecast. Theforecastingofsalesisessentialforahotelinordertoprotectthehotel'spositioninahighlycompetitivemarket( Cranage&Andrew 1992 ; Limetal. 2009 ; Tony&Poon 2012 ).Theutilizationofthetimeseriesmodelsmaydiminishsomeofthenancialrisksthatindependentlyownedhotelsexperience.Moreover,timeseriesmodelsmayhelptomanagetheresourcesmoreeffectivelyatlowcostforthehotel.Itbecomescrucialfortheindependentlyownedhotels,becausetheymighthavelimitedaccesstothenancialresourcesunlikelargerchains.Theavailabilityofreliableforecastingmodelsandthelevelofaccuracyoftheforecastedsaleswilldirectlyaffectthedecisionwhichrelatetostafng,planningoftheworkingshifts,purchasingdecisions,allocationoftheavailableresources,andcapacitymanagement( Aghazadeh 2007 ; Frechtling 2001 ; Limetal. 2009 ; Songetal. 2011 ; Tony&Poon 2012 ). Limetal. ( 2009 )suggestedthatforecastingplaysanimportantroleintheyieldmanagementandthemaximizationoftheroomrevenue.Insummary,twotimeseriesmodelswerefoundtobeaneffectivetoolforsalesforecasting,becausetheyarerelativelysimpleinapplicationandcouldprovidemanagerswiththeaccuraterepresentationofthecurrentandfuturetrendsinthesales. AsfortheARIMAmodeltheopposingresultsbetweenthetwodatasets(i.e.STRandtheindependenthotelunit)maydemonstrateacautionsignforindependenthotelmanagers.ThecautionsignwarnshotelmanagersthattheyshouldnotonlyformroomsalesprojectionsfromtheaveragedaggregatehoteldatathatSTRprovides 73

PAGE 74

toallsubscribersbutshouldinsteadtreatthedataasabenchmarkforthehotel'sperformance.Infact,ascouldbeexpected,ifmanagerscreateaforecastfromaveragedaveragestheresultscannotbeaformofaccuratepredictionasseenbytheARIMAforecast.However,theARIMAmodelfortheindividualunithotel(whosedatawasalsopartoftheaveragedaggregatedatasetfromSTR)wasabletoprovideasuperiorlevelforecastforthatindividualhotelwhenonlyusingitsdata.Thisisvaluableinformationforindependenthotelmanagerstolearnandunderstand.TheSTRreportsshouldbetreatedasbenchmarkassessmentsandnotusedtopredictroomsalesandadjusttherates.Theaverageddatacouldincorporateseveralsalespatternsfromarangeofthehotels,whichmightcharacterizeageneralsituationinaparticularhotelmarket.Nonetheless,inordertoformulatefuturesaleshotelmanagersshouldhaveaclearunderstandingaboutthetendenciesandseasonalpatternswhichdeterminethehotel'snancialperformance. 5.4FutureResearch Thecurrentstudyexaminedtheperformanceofonlythreetimeseriesmodels;therefore,futureresearchmayincorporatenewmodelsinadditiontotheadditiveseasonaldecomposition,theadditiveexponentialsmoothing,andtheARIMAmodels.ThesuggestedmodelscouldbefurthertestedonthedatasetsrepresentativeofdifferentUShotelmarketsandinternationalmarketstoanalyzeiftheperformanceofthemodelsisconsistent.ThisisespeciallyneededfortheARIMAmodelduetoinconclusiveperformanceinthisstudy.TheinadequateperformanceoftheARIMAmodelontheSTRcompetitivedatasetmayrelatetoanissuethattheADRswereaveragedacrossdiverserangeofthehotelsthatincorporatedmultipletrendsandpatterns.Theforecastbytheadditiveexponentialsmoothingalsowasnotabletoresemblesomeofthepatternsof 74

PAGE 75

theSTRcompetitivedataset.Thus,theforecastingusingtheARIMAmodelrequiresadditionalexamination. Additionally,theperformanceofthemodelsmaybetestedusingdifferentunitofanalysis,forexample,revenueperavailableroom(RevPAR)oroccupancyrates.Futurestudiescouldalsoinvestigatethecommonlyutilizedforecastingmethodsandmodelsoftheindependentlyownedhotels.Whiletherecouldbeanurgeinndingaccessibleandeffectiveforecastingtools,itseemsthatthereisagapintheresearchregardingthisissue.Likewise,furtherresearchmayconcentrateontheutilizationofthejudgementalmethodandtheeffectivenessofthisapproachforsalesforecastinginthecurrentmarketsituation. 5.5SummaryoftheStudy Thestudydrewtheattentionuponthehighratesofclosuresamongtheindependentlyownedhotels( Wilson 2013 ).Theindependentlyownedhotelsmightnothaveanaccesstotheelaboratedforecastingmethods;therefore,thetimeserieswererecommendedasanaccessibleandeffectivealternative,whichcouldassistinreducingtheuncertaintyassociatedwithfuturesales.Comparisonoftheexistingforecastingmethodsandmodelsshowedthattimeseriesmodelswerelessdemandingittermsofexpenditures,time,andexpertise,butatthesametimewereabletomanagehighlyseasonalhotelsalesdata.Subsequently,threetimeseriesmodelswereproposedandtested,thesemodelsweretheadditiveseasonaldecomposition,theadditiveexponentialsmoothing,andtheARIMA.Amongthesemodels,theadditiveseasonaldecompositionandtheadditiveexponentialsmoothingmettheexpectationsregardingtheirabilitytoresemblethepatternsinthesalesdata.Theadditiveseasonaldecompositionclearlyshowedcurrenttendenciesinthedata,whiletheadditiveexponentialsmoothingproducedactualforecastswiththelowMAPEandRMSPE.TheARIMAmodelalsoshowedpromise 75

PAGE 76

inproducingaccurateforecastsforindividualhoteldatabutwasnotanapproachthatshouldbeusedwithdatathatrepresentsaggregateaverageddatalikethatofSTR. TheresearchquestionregardingtheperformanceoftheARIMAmodelinvestigatediftheforecastofARIMAwassuperiortotheadditiveexponentialsmoothing.However,thestudycouldnotconrmthisassumption;thus,furtherinvestigationisneeded.Basedonthendingstheadditiveseasonaldecompositionandtheadditiveexponentialsmoothingmodelswererecommendedforsalesforecastingbythemanagersoftheindependentlyownedhotels.Thecurrentstudyhasseverallimitationspertainingtotheapplicationofthemodels,theirperformance,andthedatasetswhichwereusedformodels'testing.Thefoundlimitationscouldbeusedforfutureresearchinordertobeabletomakeamoregeneralconclusionabouttheperformanceofthetimeseriesmodelsandtheireffectivenessforsalesforecastingwithintheindependentlyownedhotelsector. 76

PAGE 77

REFERENCES (2009a).PwC:USforecastshows2010demandrecovery. URL http://www.hotelnewsnow.com/Articles.aspx/2173/PwC-US-forecast-shows-2010-demand-recovery (2009b).STRreportsUShotelpipelinefor2009. URL http://www.hotelnewsnow.com/Articles.aspx/2017/STR-reports-US-pipeline-for-September-2009 (2009c).STRreportsU.S.performanceforJanuary. URL http://www.hotelnewsnow.com/Articles.aspx/710/STR-reports-US-performance-for-January (2010).STR:USresultsforweekending19June. URL http://www.hotelnewsnow.com/Articles.aspx/3566/STR-US-results-for-week-ending-19-June (2012).STRreportsUShotelpipelineforJan.'12. URL http://www.hotelnewsnow.com/articles.aspx/7540/STR-reports-US-hotel-pipeline-for-Jan-12 (2013).STRreportsUShotelpipelineforFebruary. URL http://www.hotelnewsnow.com/Articles.aspx/10087/STR-reports-US-hotel-pipeline-for-February Aghazadeh,S.(2007).Revenueforecastingmodelsforhotelmanagement.TheJournalofBusinessForecasting,26(3),33. Andrew,W.P.,Cranage,D.A.,&Lee,C.K.(1990).Forecastinghoteloccupancyrateswithtimeseriesmodels:anempiricalanalysis.JournalofHospitality&TourismResearch,14(2),173. Armstrong,J.S.(1985).Fromcrystalballtocomputer.Wiley,NewYork. As'ad,M.(2012).FindingthebestARIMAmodeltoforecastdailypeakelectricitydemand. URL http://ro.uow.edu.au/cgi/viewcontent.cgi?article=1011\&context=asearc 77

PAGE 78

Baker,M.,&Riley,M.(1994).Newperspectivesonproductivityinhotels:someadvancesandnewdirections.InternationalJournalofHospitalityManagement,13(4),297. Burgess,C.(2007).Dohotelmanagershavesufcientnancialskillstohelpthemmanagetheirareas?InternationalJournalofContemporaryHospitalityManagement,19(3),188. Cassar,G.,&Gibson,B.(2007).Forecastrationalityinsmallrms*.JournalofSmallBusinessManagement,45(3),283. Chu,F.-L.(1998).Forecastingtourismdemandinasian-paciccountries.AnnalsofTourismResearch,25(3),597. Cowpertwait,P.S.,&Metcalfe,A.V.(2009).IntroductorytimeserieswithR.Springer. Cranage,D.(2003).Practicaltimeseriesforecastingforthehospitalitymanager.InternationalJournalofContemporaryHospitalityManagement,15(2),86. Cranage,D.A.,&Andrew,W.P.(1992).Acomparisonoftimeseriesandeconometricmodelsforforecastingrestaurantsales.InternationalJournalofHospitalityManage-ment,11(2),129. Croes,R.,&Semrad,K.J.(2012).Discountingworksinthehotelindustry:astructuralapproachtounderstandingwhy.TourismEconomics,18(4),769. Dahlstrom,R.,Haugland,S.A.,Nygaard,A.,&Rokkan,A.I.(2009).Governancestructuresinthehotelindustry.JournalofBusinessResearch,62(8),841. Dickey,D.A.,&Fuller,W.A.(1981).Likelihoodratiostatisticsforautoregressivetimeserieswithaunitroot.Econometrica:JournaloftheEconometricSociety,(pp.1057). Emeksiz,M.,Gursoy,D.,&Icoz,O.(2006).Ayieldmanagementmodelforve-starhotels:computerizedandnon-computerizedimplementation.InternationalJournalofHospitalityManagement,25(4),536. Enz,C.A.,Canina,L.,&Lomanno,M.(2009).Competitivepricingdecisionsinuncertaintimes.CornellHospitalityQuarterly,50(3),325. Frechtling,D.(2001).Forecastingtourismdemand.Routledge. Hanks,R.D.,Cross,R.G.,&Noland,R.P.(1992).Discountinginthehotelindustry:anewapproach.TheCornellHotelandRestaurantAdministrationQuarterly,33(1),15. 78

PAGE 79

Harnisch,O.(2008).Productivitymanagementinthehospitalityindustry. URL http://www.hospitalitynet.org/news/4035884.html Hilas,C.S.,Goudos,S.K.,&Sahalos,J.N.(2006).Seasonaldecompositionandforecastingoftelecommunicationdata:Acomparativecasestudy.TechnologicalForecastingandSocialChange,73(5),495. Hood,S.,&Mandelbaum,R.(2012).Today'sbenchmarkingtools.TheBottomline,27(4),18. Howard,Z.(2011).U.S.autumnfoliageattractshordesoftourists. URL http://www.reuters.com/article/2011/10/17/us-travel-usa-leaves-idUSTRE79G6NQ20111017 Hu,B.A.,&Cai,L.A.(2004).Hotellaborproductivityassessment:adataenvelopmentanalysis.JournalofTravel&TourismMarketing,16(2-3),27. Jang,S.S.(2004).Mitigatingtourismseasonality:Aquantitativeapproach.AnnalsofTourismResearch,31(4),819. Kahn,K.B.(1998).Benchmarkingsalesforecastingperformancemeasures.TheJournalofBusiness,99. Karande,K.,&Magnini,V.P.(2011).Therelativeuseofcontextualandtemporalreferencepricecomponentsinhotelandairlinepurchases.JournalofHospitality&TourismResearch,35(1),119. Kimes,S.E.(1989).Thebasicsofyieldmanagement.TheCornellHotelandRestaurantAdministrationQuarterly,30(3),14. Kulendran,N.,&King,M.L.(1997).Forecastinginternationalquarterlytouristowsusingerror-correctionandtime-seriesmodels.InternationalJournalofForecasting,13(3),319. Kulendran,N.,etal.(1996).Modellingquarterlytouristowstoaustraliausingcointegrationanalysis.TourismEconomics,2(3),203. Lee,A.O.(1990).Airlinereservationsforecasting:Probabilisticandstatisticalmodelsofthebookingprocess.Tech.rep.,Cambridge,Mass.:FlightTransportationLaboratory,Dept.ofAeronauticsandAstronautics,MassachusettsInstituteofTechnology,[1990]. Lewis,C.D.(1982).Industrialandbusinessforecastingmethods:Apracticalguidetoexponentialsmoothingandcurvetting.ButterworthScienticLondon,UK. 79

PAGE 80

Li,G.,Wong,K.K.,Song,H.,&Witt,S.F.(2006).Tourismdemandforecasting:Atimevaryingparametererrorcorrectionmodel.JournalofTravelResearch,45(2),175. Lim,C.,Chang,C.,&McAleer,M.(2009).Forecastingh(m)otelguestnightsinnewzealand.InternationalJournalofHospitalityManagement,28(2),228. Makridakis,S.,Wheelwright,S.C.,&Hyndman(1998).Forecastingmethodsandapplications. Mayock,P.(2011a).Independenthoteldevelopmentslowmoving. URL http://www.hotelnewsnow.com/Articles.aspx/4948/Independent-hotel-development-slow-moving Mayock,P.(2011b).Independenthotelsegmentshowsresiliency,strength. URL http://www.hotelnewsnow.com/Articles.aspx/4942/Independent-hotel-segment-shows-resiliency-strengths Mentzer,J.T.,&Cox,J.E.(1984).Familiarity,application,andperformanceofsalesforecastingtechniques.JournalofForecasting,3(1),27. Moon,M.A.,Mentzer,J.T.,&Smith,C.D.(2003).Conductingasalesforecastingaudit.InternationalJournalofForecasting,19(1),5. Nooteboom,B.,Kleijweg,A.,&Thurik,R.(1988).Normalcostsanddemandeffectsinpricesetting:Astudyofretailing.EuropeanEconomicReview,32(4),999. Okumus,F.,Altinay,L.,&Chathoth,P.(2010).Strategicmanagementforhospitalityandtourism.http://www.sciencedirect.com/science/book/9780750665223. Phillips,P.C.,&Perron,P.(1988).Testingforaunitrootintimeseriesregression.Biometrika,75(2),335. Schwartz,Z.,&Cohen,E.(2004).Hotelrevenue-managementforecastingevidenceofexpert-judgmentbias.CornellHotelandRestaurantAdministrationQuarterly,45(1),85. Semrad,K.J.(2010).Discounting:AnEmpiricalJusticationforItsValueintheLodgingIndustry.Ph.D.thesis,UniversityofCentralFloridaOrlando,Florida. Song,H.,Li,G.,Witt,S.F.,&Athanasopoulos,G.(2011).Forecastingtouristarrivalsusingtime-varyingparameterstructuraltimeseriesmodels.InternationalJournalofForecasting,27(3),855. 80

PAGE 81

Song,H.,&Witt,S.F.(2000).Tourismdemandmodellingandforecasting:moderneconometricapproaches.Pergamon. Song,H.,Witt,S.F.,&Li,G.(2009).Theadvancedeconometricsoftourismdemand,vol.13.Taylor&Francis. Song,H.H.,&Witt,S.F.(2012).Tourismdemandmodellingandforecasting.Routledge. Steed,E.,&Gu,Z.(2009).Hotelmanagementcompanyforecastingandbudgetingpractices:asurvey-basedanalysis.InternationalJournalofContemporaryHospitalityManagement,21(6),676. Sturman,M.C.,Corgel,J.B.,&Verma,R.(2011).TheCornellSchoolofHotelAdminis-trationonHospitality:CuttingEdgeThinkingandPractice.Wiley. Susilovic,A.,&Sertic,M.(2010).Impactofeconomiccrisisontourismdemand.FacultyofTourismandHospitalityManagementinOpatija.BiennialInternationalCongress.Tourism&HospitalityIndustry,(pp.1659). URL http://search.proquest.com/docview/763422441?accountid=10920 Tony,S.,&Poon,Y.T.(2012).Revenuemanagement:resolvingarevenueoptimizationparadox.InternationalJournalofContemporaryHospitalityManagement,24(4),507. Weatherford,L.R.,&Bodily,S.E.(1992).Ataxonomyandresearchoverviewofperishable-assetrevenuemanagement:yieldmanagement,overbooking,andpricing.OperationsResearch,40(5),831. Weatherford,L.R.,&Kimes,S.E.(2003).Acomparisonofforecastingmethodsforhotelrevenuemanagement.InternationalJournalofForecasting,19(3),401. Whiteld,R.,&Duffy,A.(2011).Extendedrevenueforecastingwithinaserviceindustry.InternationalJournalofProductionEconomics. Wilson,C.(2013).UShotelclosureanalysis. URL http://www.hotelnewsnow.com/articles.aspx/10128/US-hotel-closure-analysis Witt,S.F.(1992).Modelingandforecastingdemandintourism.AcademicPress. Witt,S.F.,Newbould,G.D.,&Watkins,A.J.(1992).Forecastindomestictourismdemand:Applicationtolasvegasarrivalsdata.JournalofTravelresearch,31(1),36. 81

PAGE 82

Witt,S.F.,Song,H.,&Louvieris,P.(2003).Statisticaltestinginforecastingmodelselection.JournalofTravelResearch,42(2),151. Yksel,S.(2007).Anintegratedforecastingapproachtohoteldemand.MathematicalandComputerModelling,46(7),1063. Zakhary,A.,Atiya,A.F.,El-Shishiny,H.,&Gayar,N.E.(2009).Forecastinghotelarrivalsandoccupancyusingmontecarlosimulation.JournalofRevenue&PricingManagement,10(4),344. 82

PAGE 83

BIOGRAPHICALSKETCH EkaterinaSorokinaacquiredherbachelor'sdegreeinmanagementattheDepartmentofEconomics,IndustrialManagementandEngineeringatVoronezhStateTechnicalUniversity,Voronezh,Russiain2008.Ekaterinastartedhermaster'sdegreeattheDepartmentofTourism,RecreationandSportManagement,UniversityofFloridainJanuary2012.Ekaterina'sresearchworkpertainstopracticalforecastinginthelodgingindustryandminimizationofthenancialrisksintheindependentlyownedhotelsector.DuringthecourseofhereducationattheUniversityofFlorida,Ekaterinahasbeengrantedoutstandinginternationalstudentawardandcerticateofoutstandingachievement.Sheadditionallyreceivedtwocerticatesaftercompletionofteachingdevelopmentandtechnologyprograms.Ekaterina'sinterestinhospitalitymanagementledhertobecomeamemberoftheHospitalityFinancial&TechnologyProfessionals(HFTP)organizationinApril2012.ShealsojoinedResort&CommercialRecreationAssociationinSeptember2012;theorganizationthatunitesresearchesandpractitionersinthehospitalityindustry.Ekaterinaappliesherknowledgeandexperiencenotonlyintheacademicwork,sheishighlyinvolvedinvolunteeringfortheeventsthattakeplaceinlocalcommunities.ShehasbeenfortunatetoworkwiththeGainesvilleSportCommissionandtheDepartmentofCulturalAffairsinGainesville. 83