英文市場(chǎng)波動(dòng)對(duì)馬來(lái)西亞酒店效率的影響

PublicDisclosureAuthorizedPublicDisclosureAuthorized

PolicyResearchWorkingPaper10953

TheImpactofMarketVolatilityonHotelEfficiencyinMalaysia

DoesHotelSizeMatter?

MohammadAminNesmaAli

WORLDBANKGROUP

DevelopmentEconomicsGlobalIndicatorsGroupOctober2024

PolicyResearchWorkingPaper10953

Abstract

Itisoftenarguedthatsmallfirmsaremoreflexiblethanlargefirms.Asaresult,smallfirmsperformbetterinvolatilemarketscomparedtolargefirms.ThepresentpaperexploresthisideaforarepresentativesampleofprivatehotelsinMalaysia.Specifically,thepaperestimatestheimpactofvolatilityinoccupancyratesonthepuretechnicaleffi-ciencyofsmallversuslargehotels.Aslack-basednon-radialefficiencymeasureobtainedfromthedataenvelopmentanalysismethodologyisused.Theempiricalresultsconfirm

thatsmallerhotelsarebetteratdealingwithvolatilitythanlargehotelsare.Thatis,thereisapositiveandsignificantimpactofhighervolatilityontheefficiencyofrelativelysmallhotels,anegativeandsignificantimpactontheeffi-ciencyoflargerhotels,andnosignificantimpactontheefficiencyoftheaveragehotel.Higherwomen’sownershipalsohelpshotelstodealwithvolatility.Thepaperdiscussesthepolicyimplicationsofthefindings.

ThispaperisaproductoftheGlobalIndicatorsGroup,DevelopmentEconomics.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundtheworld.PolicyResearchWorkingPapersarealsopostedontheWebat

/prwp

.Theauthorsmaybecontactedatmamin@.

ThePolicyResearchWorkingPaperSeriesdisseminatesthefindingsofworkinprogresstoencouragetheexchangeofideasaboutdevelopmentissues.Anobjectiveoftheseriesistogetthefindingsoutquickly,evenifthepresentationsarelessthanfullypolished.Thepaperscarrythenamesoftheauthorsandshouldbecitedaccordingly.Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheInternationalBankforReconstructionandDevelopment/WorldBankanditsaffiliatedorganizations,orthoseoftheExecutiveDirectorsoftheWorldBankorthegovernmentstheyrepresent.

ProducedbytheResearchSupportTeam

TheImpactofMarketVolatilityonHotelEfficiencyinMalaysia:DoesHotel

SizeMatter?

Thisdraft:October2024

By:MohammadAmin*andNesmaAli**

Keywords:Hotels,PureTechnicalEfficiency,HotelSize,DEAJELCodes:

*Correspondingauthor.SeniorEconomist,EnterpriseAnalysisUnit,DECEA,WorldBank,Washington,DC.Email:mamin@.ORCID:

/0000-0002-9451

-3629

**

Economist,EnterpriseAnalysisUnit,DECEA,WorldBank,Washington,DC.Email:

nali4@

Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheInternationalBankforReconstructionandDevelopment/WorldBankanditsaffiliatedorganizations,orthoseoftheExecutiveDirectorsoftheWorldBankorthegovernmentstheyrepresent.

WethanktheEnterpriseAnalysisUnitoftheDevelopmentEconomicsGlobalIndicatorsDepartmentoftheWorldBankGroupformakingthedataavailable.WethankJorgeLuisRodriguezMezaandparticipantsataseminarorganizedbytheWorldBank’sGlobalIndicatorsDepartmentforprovidingveryusefulcomments.Allremainingerrorsareourown.

2

1.Introduction

Volatilityinmarketdemandcharacterizedbyerratic,diurnal,seasonal,andcyclicalfluctuationsinthenumberofvisitorsandoccupancyratesisakeyfeatureofthehotelindustry(seeAlemayehuandTveteraas2020,SaitoandRom?o2018,Parketal.2016,ChenandYeh2012,Jang2004,andHighmanandHitch2002).Whiletherearesomebenefitsfromhighervolatility,moststudiescontendthatthehighcostofadjustinginputsintheshortrun(henceforth,adjustmentcost)outweighsthebenefits.Studiesofthemanufacturingsectorsuggestthattheadjustmentcostvaries,beinglessforsmallcomparedtolargefirms.Asaresult,smallerfirmssufferlessorbenefitmorefromhighervolatility.Thepresentpapermakesafirstattemptatexploringthisideainthecaseofhotels.Weestimatetheimpactofvolatilityinoccupancyratesonaslack-basedmeasureoftheefficiencyofhotelsinMalaysiaobtainedusingtheDataEnvelopmentAnalysis(DEA)methodology.Ourresultsconfirmthatsmallerhotelsarelessnegatively(ormorepositively)impactedbyhighervolatilityinoccupancyrates.Wealsofindthathotelswithmorewomen’sownershipsufferlessfromhighervolatilityinoccupancyrates,andthosethatusetemporaryworkerssuffermore.

Therelationshipbetweenvolatilityindemandandtheprofitabilityorefficiencyofhotelshasbeenexploredinseveralstudies(Section2.1reviewstheliterature).Whiletheevidenceissomewhatmixed,moststudiesfindthathighervolatilityhasanegativeimpactonhotels’performance.Studiesforotherservicesectorsandmanufacturingsectorsreachasimilarconclusion.Thekeyissuehereistheshapeoftheshort-runaverageandmarginalcostcurves.Thesteepertheslopeofthecostcurves,thelargerthedeclineinfirmperformancewhenadjustingoutputtomatchfluctuatingdemand.Efficiencydifferenceshavealsobeenlinkedtohotelsize(seePulinaetal.2010,Pérez-Rodríguezetal.2023,AissaandGoaied2016,SalmanSalehetal.2012,

3

ShyuandHung2012,AssafandAgbola2011),althoughnotalwaysinthesamedirection.Forinstance,Pulinaetal.(2010)findthatmedium-sizedhotelsinSardinia,Italy,aremoreefficientthansmallandlargehotels.However,forhotelsintheCanaryIslands,Spain,Pérez-Rodríguezetal.(2023)findthatefficiencyishigherforlargehotelsthanforsmallandmediumhotels.DeJorgeandSuárez(2014)findaU-shapedrelationshipbetweenefficiencyandthesizeofhotelsinSpain.

Somestudiesofthemanufacturingsectoranalyzehowtheshapeoftheshort-runcostcurves,andthereforetheimpactofvolatilityindemandonfirmperformance,variesbetweensmallandlargefirms(section2.2reviewstheliterature).Theclaimhereisthatproductiontechnologyandtheinternalorganizationofsmallfirmsaremoreflexible,andthereforesmallfirmscanrespondtovolatilityatalowercostthanlargefirms.Incontrast,largefirmshavetheirefficiencynicheinmorestablemarkets,whereeconomiesofscaleallowthemtoachievealowerlong-runaveragecost.Tothebestofourknowledge,theissueofhowhotelsofdifferentsizesrespondtovolatilityhasnotbeenexplored.Thepresentpaperisthefirsttodoso.

Ourresultsshowthathighervolatilityinoccupancyrateshasnosignificantimpactontheefficiencyoftheaveragehotel(seefigure1).However,thereisasharpheterogeneity,withtheimpactbeingpositiveandstatisticallysignificantfortherelativelysmallhotelsandnegativeandsignificantfortherelativelylargehotels(seefigure2).Forourfinalbaselinespecification,aonestandarddeviationincreaseinvolatilityinoccupancyratesisassociatedwithanincreaseinefficiencyby0.073points(about14.2percentofthesamplemeanefficiency)forhotelsatthe25thpercentilevalueofsize,whichissignificantatthe5percentlevel.Thecorrespondingchangeatthe75thpercentilevalueofhotelsizeisadecreaseinefficiencyby0.092points(about18.1percentofthesamplemeanefficiency),significantatthe5percentlevel.Theresultisrobusttoseveralalternativemeasuresofefficiency,hotelsize,andvolatility.Wealsofindthathigherwomen’s

4

ownershipisassociatedwithalessadverse(ormorebeneficial)impactofhighervolatilityinoccupancyratesontheefficiencyofhotels.Tothebestofourknowledge,thisisthefirstpapertofindsuchagenderedeffectinanyindustry.

ThesurveyofhotelsthatweuseisnationallyrepresentativeofregisteredprivatehotelsinMalaysia.Thesurveyprovidesinformationonvarioushotelcharacteristicsandhotels’experienceswithdifferentaspectsofthebusinessenvironment.Weexploitthisrichinformationtoraiseourconfidenceagainsttheomittedvariablebiasproblem.WealsousethemethodologyofOster(2019)toformallytestforomittedvariablebias.Nevertheless,causality-wise,ourresultsshouldbetreatedwithduecautionastheyarebasedonpurecross-sectionaldata.

2.Literaturereviewandconceptualframework

2.1Volatilityandfirmperformance

Fluctuationsinmarketdemandcanaffecttheperformanceofhotelssignificantly.Lowdemandduringoff-peakseasonsresultsinexcessivecapacity(CucciaandRizzo2011,Parketal.2016)iftherearefixedcostsinproductionorahighcostofadjustinginputssuchasthenumberofrooms(seeButters2020)andlabor(seeAlemayehuandTveteraas2020,Parketal.2016).Toohighdemandduringpeakseasonscanalsoputpressureonavailableresources,leadingtopoorerqualityofserviceandlowerperformance(seeParrillaetal.2007).Inanearlytheoreticalcontribution,SheshinskiandDreze(1976)showedthat,comparedtostationarydemand,fluctuatingdemandleadstoahigherexpectedcostperunitofoutput.Atleasttosomeextent,theproblemoffluctuatingdemandcanbesolvedbyusingamoreflexibleproductionmethodortechnology.However,flexibletechnologiesarelimited,andsoistheirefficacyinimprovingefficiency(see,forexample,Bryson2007,Kleinknechtetal.2006).Moreimportantly,flexibletechnologiesmayimpose

5

additionalcosts.AsMills(1984)notes,aplantcertaintooperatexunitsofoutputperweekwillsurelyhavelowercostsatthatoutputthanwillaplantdesignedtobepassablyefficientfromx/2to2x.Othertheoretical(Hagspieletal.2016)andempirical(Meraetal.2017,MerschmannandThonemann2011)studiesmakeasimilarpoint.

Severalpapershaveempiricallyexploredtherelationshipbetweenhotelperformanceandvolatilityindemand.Sáez-Fernándezetal.(2020)findthathigherseasonalityisassociatedwithlowerefficiencyamonghotelsintheBalearicIslands,Spain.ChenandChang(2012)findanegativeimpactofpriceinstabilityontheprofitabilityofhotelsinTaiwan,China.AlemayehuandTveteraas(2019)findthatfor94hotelsinNorway,demandfluctuationsareassociatedwithonlyapartialadjustmentoflabortotheoptimallevelinthelongrun.Thus,theyconcludethatdemandfluctuationscancausehotelstooperateatsuboptimallevels.ChenandYeh(2012)findthatmoredemanduncertaintyisassociatedwithahigherlikelihoodoffailureamonginternationaltouristhotelsinTaiwan,China.SaitoandRom?o(2018)findthatseasonalvariationhasanon-negligibleimpactonthetotalfactorproductivityofhotelsinSpain.Parketal.(2016)findanegativeimpactofmoredemandvolatilityonthelaborproductivityofhotelsin43medium-sizedhotelsintwochainsintheUK.Fernandez-MoralesandMayorga-Toledano(2008)findthatforhotelsinCostadelSolinSpain,underutilizationofcapacityinperiodsoflowdemandcoupledwithfixedcostshasanegativeeffectonproductivity.Similarresultsarealsofoundforotherserviceindustries(seeMorikawa2012,Baker2004)andmanufacturing(seeMerschmannandThonemann2011).

Anegativeimpactofhighervolatilityonhotels’performanceisnotaforgoneconclusion.Therearestudiesthateitherfindnosignificantimpactorapositiveimpactfromhighervolatility.Forexample,JonesandSiag(2009)analyzetheimpactofdemandvariabilityontheproductivityof45chainhotelsintheUK.Theauthorsfindnosignificantimpactofdemandvariability.Ortega

6

andChicon(2013)alsoreportthatseasonalitydoesnotreducelaborproductivityintheSpanishhospitalityindustry.Lado-SestayoandFernández-Castro(2019)findapositiveimpactofseasonalityontheefficiencyofhotelsinSpain,whichtheyattributetocostsavingswhenproductionisconcentratedinafewperiodsoftheyear.Also,higherseasonalitycanimproveproductivitybyallowingbusinessestoundertakemaintenanceorrefurbishmentworkordevelopnewmarkets(seeGrantetal.1997).

2.2Therelevanceoffirmsize

Severalstudieshaveexploredhowthecostofadjustinginputsintheshortrunvariesacrossfirms.Mostofthesestudiesfocusonthemanufacturingsector,andwearenotawareofanysuchstudyforthehotelindustry.Thebroadideahereisthattheadjustmentcost,asreflectedinhowsteeplyshort-runmarginalandaveragecostcurvesrise,dependsinpartonafirm’sinternalorganization,whichvariesbetweensmallandlargefirms.Itisclaimedthatsmallerfirmsaremoreflexible(flattershort-runaverageandmarginalcostcurves)thanlargerfirms,andsosmallerfirmssufferless(orbenefitmore)fromhighervolatility.Largerfirmshavetheirownefficiencyniche,whichischaracterizedbyalowerlongrunaveragecost.Stigler(1939)wasthefirsttoarguethatsmallfirmshavealowercostofadjustinginputsintheshortrunthanlargefirms.Dasetal.(1993)arguetheoreticallyandprovideevidencethat,comparedtolargefirms,smallfirmshavemoreflexibleproductiontechnologies,asreflectedintheflattershort-runaveragecostcurve.Thisallowssmallerfirmstorespondbettertochangingdemandconditions.CavesandPugel(1980)makeasomewhatsimilarpoint.Theyarguethatlargefirmsrelymoreoncapital-intensivemethodsofproductionthathavehighfixedcosts.Greaterrelianceoncapitalandmorespecializedformsofcapitalreduceslargefirms’abilitytoadjusttodemandfluctuations.Incontrast,smallfirmschoosemoreflexible

7

productionmethods,whichentailtheuseofmorenonspecializedinputsandagreaterrelianceonvariablefactorsofproduction.MillsandSchumann(1985)alsoarguethatfirmsizeandflexibilityareinverselyrelatedwithinindustries.Accordingtotheseauthors,smallfirmstendtohavefewerdecision-makersandalesscomplicatedbureaucracythanlargefirms.Theseorganizationalcharacteristicsmeanthatsmallfirmscanrespondmorequicklytochangesinmarketconditions.FiegenbaumandKarnani(1991)alsohighlighttheseandotherorganizationalfactors.UsingdatafromCompustaton3,000companiesfrom83industries,theauthorsfindthatsmallerfirmshavegreateroutputflexibilitythanlargefirmsandthatthisgreateroutputflexibilityimprovestheperformanceofsmallerfirms.Interestingly,theauthorspredictthatthereisnosignificantrelationshipbetweenoutputflexibilityandtheperformanceoftheaveragefirm,apositiverelationshipbetweenthetwoforsmallfirms,andanegativeoneforlargefirms.Wefindsimilarresultsbelowfortheimpactofmarketvolatilityonhotelefficiency.Zimmermann(1995)alsodevelopsatheoreticalmodelandprovidesempiricalevidencefromGermanmanufacturingindustriesthatshowsthatsmallerfirmsaremoreflexible.Hirschetal.(2020)consider2,186firmsintheEUdairyprocessingindustry.Theyalsofindthatlargerfirmshavelowerlong-runaveragecostcurvesorstaticefficiency,whilesmallerfirmsaremoreflexible.Renneretal.(2014)usedataonPolishagriculturalfarmstoestimatetherelationshipbetweenfarmsizeandflexibilityinproduction.Asintheabovestudies,flexibilityiscapturedbytheslopeoftheaveragecostcurve.Thisstudyalsoconfirmsanegativerelationshipbetweenflexibilityandfarmsize.For58manufacturingindustriesinMalaysia,Noretal.(2007)findthatvariationsinsalesarelargerfortherelativelysmallerfirms.Theauthorsarguethatthisisconsistentwithsmallerfirmsrespondingbettertounexpecteddemandfluctuations.

8

3.DEAmethodology

DEAisanon-parametricmethodofestimatingtheefficiencyofdecision-makingunits(DMUs).Beingnon-parametric,DEAdoesnotmakeanyassumptionsabouttheformoftheproductiontechnologyoraboutthedistributionalpropertiesoftheefficiencyestimates.Throughout,wefocusontheinput-orientedmodel,whereefficiencyinvolvesminimizinginputsforagivenlevelofoutputs.Wedosobecausehotelscanonlycontrolinputuseandnotthelevelofdemandoroutput.Amongothers,Perrigotetal.(2009)andHernández-Guedesetal.(2024)useinput-oriented,non-parametricefficiencymeasures.Wealsoassumevariablereturnstoscale(VRS)technologyinsteadofconstantreturnstoscale(CRS)technology.WedosobecauseVRSisgenerallyregardedasamoreaccuraterepresentationofthetruetechnologyforthehotelindustrygivenmarketimperfections,seasonality,scaleeffects,andheterogeneity(seeHernández-Guedesetal.2024formoredetails).Inputminimizationcanhappenwhenallinputsarevariedinthesameproportion(radialmodel)orindifferentproportions(non-radialmodel).TheDEAliteratureonhotelefficiencyhasmainlyreliedonradialmeasures.Thefewstudiesthatuseanon-radialdistancemeasuredosousingthemethodologyofTone(2001).ExamplesincludeAshrafietal.(2013),Dengetal.(2020),andXiaetal.(2018).

Inthefirststep,DEAidentifiesthefeasibleproductionset.ThisincludesalltheDMUs’observedinputandoutputvectors(productionvectors),allproductionvectorswithless(ofoneormore)outputsand/ormore(ofoneormore)inputsthanthatofanyobservedDMU(freedisposalassumption),andalllinearcombinationsoffeasibleproductionvectors(convexityassumption).Inthesecondstep,theefficiencyfrontierisconstructed,whichconsistsofallfullyefficientfeasibleproductionvectors.Aproductionvectorisfullyefficientifthereisnootherfeasiblevectorwiththesame(orhigher)outputthatuseslessofallinputs(radial)orsomeinputs(non-radial).In

9

thelaststep,aDMU’s(in)efficiencyiscalculatedasitsdistancefromthefrontier.Itcapturesthemaximumpercentagereductionininputswhenmovingtothefrontierwithoutreducingtheoutput.

DEAwasfirstproposedasalinearprogrammingproblembyCharnesetal.(1978).TheauthorsemployedaradialdistancemeasureandassumedCRStechnology.Bankeretal.(1984)alsoemployedtheradialdistancemeasurebutreplacedCRSwithVRStechnology.Tone(2001)introducedanon-radialmodelthatallowsfordisproportionatechangesininputsinestimatingaDMU’sdistancetothefrontier(slack-basedmeasureofefficiency).

Ourbaselineormainefficiencymeasureistheinput-orientedslack-basedmeasureofTone(2001)withVRStechnology.Forthismeasure,theefficiencylevelforDMU0usingthex0inputvectorandproducingthey0outputvectorequalsφ0whichisobtainedbysolvingthefollowinglinearprogrammingproblem:

subjectto

yr0=(r=1,…..,s)

∑=1λj=1,λj≥0(?j),s

wherejdenotesthefirm,rdenotestheoutput,idenotestheinput,s(inputsexcesses)andoutputs(outputsshortfalls),respectively.Thelinearprogrammingproblem

10

isrepeatedseparatelyforeachfirmj=1,….,n.Foranygivenfirm,thevalueof。equalsthemaximumpercentagereductionininputs(averagedoverallinputs)thatispossiblebymovingtheconcernedDMUtothefrontierwhilemaintainingitsoutputvector.

Ourmainresultisrobusttoseveralalternativemeasuresofefficiency.Tothisend,weemploythetraditionalradialDEAefficiencymeasureduetoBankeretal.(1984)assumingvariablereturnstoscale(BCCefficiency).Next,werelaxtheconvexityassumption(discussedabove)intheBankeretal.(1984)model.Thisgivesrisetothefreedisposalhullefficiency(FDHefficiency)firstintroducedbyDeprinsetal.(1984).Theslack-basedmeasureofefficiencydefinedaboveischaracterizedbyseveralDMUsthatarefullyefficient.Thisreducesthemodel’sdiscriminatorypowerindistinguishingbetweenDMUs’efficiency.Further,theslack-basedefficiencymeasureissusceptibletooutliersbecauseahandfulofhighlyefficientDMUscanmakeallothersappearhighlyinefficient.Weaddressthesepotentialshortcomingsbyemployingtheslack-based“superefficiency”measurebasedontheworkonTone(2002).

1

Themethodologyassignsdifferentefficiencyscoresgreaterthan1toanotherwisefullyefficientDMUdependingonhowmuchitsexclusionfromthesampleaffectstheefficiencyfrontier.ApotentialoutlierDMUisonewithahighsuperefficiencyscore.AnotherissuewithDEAmodelsisthattheytendtooverestimatetheefficiencyofDMUs.ThishappensbecausetheremaybesomeDMUsintheuniversenotincludedinthesamplethataremoreefficientthanalltheDMUsinthesample.Wecorrectfortheupwardbias(Biascorrectedefficiency)usingthebootstrappingmethodologyofSimarandWilson(2007).WenotethatthebiascorrectedefficiencymeasuresarecurrentlyavailableonlyforradialDEAmodels.Thus,thebiascorrectionisappliedtotheBCCefficiencydescribedabove.

1Dengetal.(2020)alsousethismeasureofefficiencyforhotelsinChina.

11

4.Datasourceandmainvariables

Ourdatasourceisafirm-levelsurveyofprivatefirmsinMalaysiaconductedbytheWorldBank’sEnterpriseSurveys(ES)in2019.TheESarenationallyrepresentativesurveysofprivateregisteredmanufacturingandservicesfirmsthathavefiveormoreemployees.Thesurveysarestratifiedbysize,sector(withinmanufacturingandservice),andregionwithinthecountry.Informalorunregisteredfirmsandthosewithfewerthanfiveemployeesareexcludedfromthesample.Wefocusonthesampleofhotels(ISICRev.3.1industrycode5510)intheES.OurestimationsbelowuseallthehotelssurveyedbytheESforwhichinformationisavailableonthemainvariablesofinterest.Thereare90hotelsinthebaselinesample.SamplingweightsprovidedbytheEStocorrectforoversamplingareusedthroughout.IntheAppendix,aformaldefinitionofallthevariablesusedintheregressionsisprovidedinTableA1.SummarystatisticsareprovidedinTableA2.

Ourbaselineresultsarebasedonestimatingthefollowingequation:

Efficiencyir=α+β1xir*Hotelsizeir+β2xir+β3Hotelsizeir

+Regionfixedeffectsr+Baselinecontrolsir+uir…..(1)

wheresubscriptidenotesthehotel,rdenotestheregion(city),Xisvolatilityinoccupancyrate(definedbelow).Thekeyparameterofinterestisβ1,whichcaptureshowtheimpactofvolatilityinoccupancyrateonahotel’sefficiencyvarieswiththesizeofthehotel.Intherobustnesschecks,severalinteractiontermcontrolsareaddedtoequation(1).TheestimationmethodusedisOrdinaryLeastSquares(OLS)withrobuststandarderrorsclusteredontheregiontimesstarratingofthe

12

hotel(definedbelow)level.Ourmainefficiencymeasureisboundedaboveby1.Hence,wealsouseTobitestimationmethodasarobustnesscheck.

2

4.1Dependentvariable

Ourdependentvariableisameasureoftheefficiencyofhotels.Asmentionedabove,ourmainefficiencymeasureistheinput-orientedslack-basedmeasureofpuretechnicalefficiencyassumingVRStechnology(SBMEfficiency).Thisisanon-radialmeasurebasedontheworkofTone(2001).Weuseoneoutputandthreeinputs.Theoutputisthetotalannualsalesofthehotel,andtheinputsarethetotalannuallaborcost,thenumberofroomsinthehotel,andthetotaloperationalcostproxiedbythetotalannualcostofelectricity.Sales,laborcost,andelectricitycostareforthelastfiscalyear.Thechoiceoftheoutputandinputsisdrivenbytheexistingliteratureanddataavailability.Severalstudieshaveusedannualsalesrevenueasanoutputmeasure(seeAlemayehuandTveteraas2020,Hernández-Guedesetal.2024,Barros2005,DeJorgeandSuárez2014).Likewise,thenumberofrooms(Barros2005,DeJorgeandSuárez2014),laborcost(Barros2005,AssafandAgbola2011,Hernández-Guedesetal.2024),andoperatingexpenses(seeLado-SestayoandFern′andez-Castro2019,Hernández-Guedesetal.2024),whichincludeelectricitycost,havebeenusedasinputsinseveralstudies.ThemeanvalueofSBMEfficiencyis0.51(or51percent),thestandarddeviationis0.21,andtherangeis0.13to1.Thus,atypicalhotelinMalaysiacanreduceallitsinputs(onaverage)byabout49percentwhilemaintainingitsoutput.About16percentofthesurveyedhotels(8.8percentwithsamplingweights)arefullyefficient.Figure3showsthedistributionofSBMEfficiency.

2BothOLSandTobitestimationmethodsareusedintheDEAliterature.WeprefertheOLSmethodbecauseitoffersaneasierestimationandinterpretationfortheinteractionterm.Estimatinginteractiontermsinnon-linearmodelssuchasTobitiscomplicated(seeAiandNorton2003).

13

OurmainresultisrobusttoseveralalternativemeasuresofefficiencythatwerediscussedinSection3.Theseincludeslack-basedsuperefficiency,biascorrectedefficiency,BCCefficiency,andFDHefficiency.Alltheefficiencymeasuresconsideredassumeaninput-orientedmodelwithvariablereturnstoscale.Outputsandinputsareasstatedinthepreviousparagraph.TableA1intheAppendixprovidesmoredetailsontheseefficiencymeasures.

4.2Mainexplanatoryvariables

Ourmainexplanatoryvariablesarehotelsize,volatilityinmarketdemandproxiedbyvolatilityinoccupancyrateexperiencedbyahotelinthelastyear,andtheinteractiontermbetweenthetwo.Forhotelsize,weusethelogofthenumberofpermanentfull-timeworkersemployedatthehotelattheendofthelastfiscalyear.Forvolatility,weusetheinformationprovidedintheESonthehighestoccupancyrate(percentage)lastyear,thelowestoccupancyratelastyear,andtheaverageoccupancyratelastyear.Specifically,ourvolatilitymeasureequalsthedifferencebetweenthehighestoccupancyrateandthelowestoccupancyratedividedbytheaverageoccupancyrate(VolatilityinOccupancy).Suchrange-basedmeasuresofagivenphenomenonaretypicalintherelatedliterature(seeFerranteetal.2018,Lundtorp2001).ThemeanvalueofVolatilityinOccupancyequals0.81,andthestandarddeviationis0.49.Thefocusoftheempiricalexercisebelowisontheinteractiontermbetweenhot