• No results found

Mediation and moderation in food-choice models: a study on the effects of consumer trust in logo on choice

N/A
N/A
Protected

Academic year: 2021

Share "Mediation and moderation in food-choice models: a study on the effects of consumer trust in logo on choice"

Copied!
8
0
0

Loading.... (view fulltext now)

Full text

(1)

ContentslistsavailableatScienceDirect

NJAS

-

Wageningen

Journal

of

Life

Sciences

j ou rn a l h o m e p a g e :w w w . e l s e v i e r . c om / l o c a t e / n j a s

Mediation

and

moderation

in

food-choice

models:

a

study

on

the

effects

of

consumer

trust

in

logo

on

choice

R.

Zanoli

a,∗

,

S.

Naspetti

a

,

M.

Janssen

b

,

U.

Hamm

b

aUniversitàPolitecnicadelleMarche bUniversityofKassel

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received16February2014

Receivedinrevisedform1October2014 Accepted16January2015

Availableonline12March2015

Keywords: willingness-to-pay WTP-spacemodels Mediationanalysis Moderationanalysis Consumertrust Labelling Organicfood

a

b

s

t

r

a

c

t

Thepaperintroducesawaytoanalysetheinfluenceofmediatingandmoderatingvariablesonwillingness topayinasimpleway.Usingdataon427Italianconsumersregardingdifferentorganiclogos,mediation andmoderationanalysisisappliedforthefirsttimeinadiscretechoicesetting.Wetestedthehypothesis thattrustinlogomediatestherelationshipbetweenthelogoandconsumerchoicefororganiclabelled foodproducts.Resultsdonotallowrejectingthehypothesisthattrustinlogototallymediatestheeffect ofthelogo.Therefore,thewillingnessfororganicproductscouldbeinterpretedas“costfortrust”:the higherthetrustthehighertheperceivedvalue-for-money.Themediationeffectoftrustinlogodoes notvaryacrosspoints-of-purchaseorregions.Ournovelapproachissusceptibleofvariousapplications whenanalysingchoicedataandcanbeextendedfurther.

©2015RoyalNetherlandsSocietyforAgriculturalSciences.PublishedbyElsevierB.V.Allrights reserved.

1. Introduction

Mediationanalysisisameansoftestinghypotheticalprocesses andmechanismsthroughwhichanindependentvariable,Z,might have an indirect effect over a dependentoutcome variable, Y, throughoneormoremediators,M,sometimescalledintervening orprocessvariables[1–4].

Usuallymediationanalysesare appliedin theframework of simple OLSmultipleregression orother methodsof estimation e.g.logit,multilevelmodelling,andstructuralequationmodelling (SEM)(amongothers:MacKinnon[5],Hayes[6],Hayes[7],Valeri &VanderWele[8]).

Moderationreferstotheeffectsofaqualitativeorquantitative variable“thataffectsthedirectionand/orstrengthoftherelation betweenanindependentorpredictorvariableandadependentor criterionvariable”[2].Itisaverydifferentconceptfrom media-tion.Forexample,genderdifferencesmayexistintheimpactof, say,brandawarenessonactualpurchaseintention.Inthiscase,the variablegenderinteractswiththevariablebrandawarenessin pre-dictingtheoutcomeofpurchaseintention.Wesaythatgenderisa moderatoroftheimpactofbrandawarenessonpurchaseintention.

∗Correspondingauthor.

E-mailaddress:[email protected](R.Zanoli).

Inthispaper,weapplymediationandmoderationanalysisto discretechoicedatatoanalysetheroleoftrustinlogoinmediating willingness-to-pay(WTP)fororganiclabelledproducts,by adapt-ingtheoriginalBaron&Kenny’s[2]approachtodiscretechoice models1.Wealsoinvestigatethemoderatingroleofthe point-of-purchaseonthismediation,whileaccountingforregionalvariation. Our resultsshowthat overalltrust intheorganiclogo com-pletelymediatesconsumers’choiceoforganicapplesandeggs.In otherwords,sinceorganiclogosarecuestotrustworthyorganic quality,trustintheselogoscompletelyexplains(causes)consumer choiceandrelatedWTP.Thiscausalityismoderatedbythe point-of-purchase,sinceexposureto(andawarenessof)differentlogos variesindifferenttypesofshops.Thattrustinorganicqualityisa keyfactorinpurchasedecisionsfororganicproductsisnotanew finding(seee.g.Naspetti&Zanoli[9]).Thenoveltyofouranalysis istodirectlyrelatetrustinlogowithWTP,hencerenderingthe

1Recently,mediationresearchershavecriticizedtheoriginalseminalworkof

Baron&Kenny[2]andfocusedmoreontheestimationoftheindirecteffectofMon Y[6].Theseworkscriticizethe“jointsignificance”approachofBaron&Kennyonthe followinggrounds:a)itssupposedlowpowerindetectingtheeffectofthe media-torvariables,b)thefactthattheindirecteffectisnottesteddirectlybut“inferred logicallybytheoutcomeofasetofhypothesistests”[6].Someofthesecritiques havebeenretracted[12]andarenotveryrelevantinourcase,sinceinthispaper wefocusonthemediationeffect(M→Y)andnotontheindirecteffect(X→M→Y). http://dx.doi.org/10.1016/j.njas.2015.01.001

(2)

perceivedproductpriceasanindicatorof“costfortrust”.Dually, themorethetrustembeddedinaspecific(organic)logo,thehigher theperceivedvalueformoney.Givenperceivedcustomervalueis afunctionofperceivedbenefitsandperceivedsacrifices[10,11], webelievethat ourfindingsarerelevantsinceallowtodirectly relatea higherlevelbenefit(trust)toanattribute (price),being logos(extrinsicattributes)justavisiblecueofhigherquality,which inturnincreasestheperceivedvalue.

Finally,thispaperintroducesawaytoanalysetheinfluenceof mediatingandmoderatingvariablesonWTPinasimpleway,using randomparameterestimationor mixedlogitmodel. Coefficient estimatesforattributesthatarefullymediatedbyothervariables should,inprinciple,bezero.Moderationishandledby decompo-singanyheterogeneityobservedwithintherandomparameters.

The rest of the article is organised as follows. Section two describesthechoicedata.Insectionthreeweillustratethemethod ofinvestigationandthedifferenttypesofmodelsusedfor test-ingformediationandmoderationeffects.Sectionfourreportsthe resultsandtheirdiscussion,whilethelastsectionconcludes. 2. Data

WeusepartialdatafromtheJanssen&Hamm’s[13] survey thatanalysed theimpactof differentorganiccertification logos onconsumer’sWTP.Thisstudyfocussedon6EU countriesand twoproducts(applesandeggs)andanalysedhowthelogochoice isinfluencedbyconsumerawarenessandperceptionofdifferent organiclogosandbyconsumercharacteristicsregardingthebuying behaviourfororganicfood.

ChoiceexperimentswereconductedtoelicittheWTPfor differ-entorganiccertificationlogos.Insubsequentstructuredinterviews, datawascollectedonfactorsthatmightinfluenceconsumers’WTP. 2.1.1. Choiceexperiment

Thedataweusehererefertothechoiceexperimentinapples. Recruitedparticipantsneededtobuyorganicapplesatleastonce permonth. The choice of theproduct obviously influencesthe WTP,thoughthescopeofourstudywasnottoobservedifferences betweentheWTPinrelationtotheorganiccategory,asinKrystallis andChryssochoidis[14].

Thehypotheticalchoiceexperimentwasdesignedtoresemble arealbuyingsituationandtobeincentive-compatible.Unlikein otherstudieswithchoiceexperiments(e.g.Loureiro&Umberger [15], Lockshin et al.[16], Lusk &Schroeder[17]), we usedreal organicapplesinsteadofpicturesordescriptionsofproducts. Typ-icalproductinformationwasshownonthepricetag,whichwas identicalacrossthe alternatives(variety, domestic origin). Fur-thermore,theparticipantswerepaida cashincentiveandwere instructedthattheywouldhavetopayforthechosenapplesjust likeinarealshoptoreducethehypotheticalbias[17].Fortax rea-sons,attheendofthechoiceexperiment,theparticipantswere informedthattheycouldnotbegiventheapplestheyhadchosen andtheycouldkeepthemoney.

Eachparticipantcompletedtwochoicesetsfororganicapples. Eachchoicesetcontainedfourproducts,whichlookedidentically butweremarkedwithdifferentorganiclogosandprices(Figure1). Fourdifferentorganiclogosweretested:threeorganiclogosplusan alternativewithoutalogo.Inaddition,theparticipantscouldalso choosenottobuyanyoftheofferedalternatives(“no-buyoption”). The no-buy option was included to make the buying decision morerealistic.Furthermore,previousstudiesshowedthatforced choicemightleadtobiasedresults[18].Thelogoswherechosen torepresentthemostfrequentlyavailablelogosinalldistribution channels(supermarketsandspecialisedorganicshops);theywhere

Fig.1. Choiceexperimentsetting.

chosenonthebasisofpreviousfocusgroupsandasurveyinretail shops.Theold voluntaryEUlogowasusedintheexperiments, sincethesurveywasconductedpriortotheintroductionofthe newmandatoryEUlogo.TheotherlogosusedwereCCPB (Certi-ficazioneecontrolloprodottibiologici),themostcommonprivate organiccertificationlogoinItaliansupermarkets,andDEMETER,a worldwideprivateorganiccertificationlogoforbiodynamic prod-ucts.Thelogoawarenesswasspecificallytestedforbya means ofLikert-typescale(1=completelyunknown7=well-known,with awarenessthresholdratingabove4):theEUlogohadanawareness of83.4%,theothertwologoslower(50.5%and46.5%respectively, inthetotalsample).IndeedtheCCPBlogoismainlyfoundinbrands soldinsupermarkets(awareness;51.9%),whiletheDEMETERone inspecialisedshops(awareness:63.8%).Figure2.

Fourdifferentpricelevelsweretested.Onemonthbeforethe experimentswereconducted,theaveragemarketpriceoforganic apples/eggswasdetermined,basedonwhichthetestedpriceswere chosen(Table1).

Anorthogonalfractionalfactorialexperimentaldesignwith16 differentchoicesetswasusedforthesystematicvariationofthe price levels acrossthe organic labels. The sample was divided intoeight blocks.Althoughit wasnotdesigned tominimize d-efficiency, the design was evaluated post-hoc using Ngene 1.1 (d-efficiency=1.94).

(3)

Table1

PricelevelsforapplesinEuro/kginthechoiceexperiments.

Price IT

level Ancona Bari Milano

-20% 2.33 2.14 2.58

Avg. 2.91 2.68 3.22

+20% 3.49 3.22 3.86

+40% 4.07 3.75 4.51

2.1.2. Structuredinterviews

Thestructuredinterviewsconductedafterthechoice experi-mentscontainedthefollowingquestions:

1.Socio-demographiccharacteristics:gender,age,householdsize, levelofeducationandnethouseholdincome.

2.Consumerperceptionsofthelabels:Theparticipantswereasked to rate each of the three tested labels containing the logos regardingtheirleveloftrustandcredibility.Theanswer cate-goriesareshowninTable2.

Thetwoitemswheresummeduptobuildascale,reflecting con-sumer‘overalltrustinlogo’(forsimplicitywewilllabeltheoverall scaleTRUST).Werecodedconsumersashavingacertaindegree ofTRUSTallrespondentswhoexhibitedaratinghigherthan4in eitherofthetwoitems.Therefore,theTRUSTscalemayhavethree values:0(thelogoisconsideredneithertrustworthynorcredible), 1(thelogoisconsideredeithertrustworthyorcredible),2(thelogo wasratedbothtrustworthyandcredible).ThenewTRUSTscaleis highlyreliable(Cronbach’salpha:0.88).

Dataon427respondentswerecollectedface-to-faceinItalyin FebruaryandMarch2010afterapre-testwith15participantsone monthearlier.Thechoiceexperimentswereconductedinthree dif-ferenttowns(MilanintheNorth,AnconaintheCentreandBariin theSouth)attwokindsofshops/locations:conventional supermar-ketsand/orshoppingcentresandspecialisedorganicfoodshops. Halfofthechoiceexperimentswereconductedat eachkindof shop,approximatelyreflectingthemarketshareofeach kindof shopinItaly,whereorganicmarketsalesaccountfor1,550Millions EurosandaremainlyconcentratedintheNorthandCentre[19]. InItaly,logosandlabelsnowadaysplayanincreasinglyimportant rolealongsidesmall(oftenspecialised)retailinginachievingmore accountable producer-consumer relations; until recently, small retailersandfarmers’marketshaveoccupiedadominantposition while“supermarketscommandacomparativelysmallfractionof therelativelyhighamountoftrustplacedinfood”[20].Thisisnot onlyacharacteristicoftheorganicfoodmarket:inItalysmalland medium-sizedfarmsarethevastmajorityandthereisa strong presenceofsmallspecializedshops(groceries,bakeries,butcheries, etc.),openmarketsandfarmers’markets,especiallyinthe Cen-treandtheSouth.Thisregional/point-of-saledifferencesobviously affecttheconsumersaswell[21].

Table2

Labelratingsintheinterviews.

Item Interviewquestion Answercategories Trust Pleaserateeachofthe

labelsonthefollowing scale.

Scalefrom1to7with: 1=Icompletelytrustthislabel. 7=Idonottrustthislabelatall Credibility Pleaserateeachofthe

labelsonthefollowing scale.

Scalefrom1to7with: 1=Thislabeldoesnotstandfor realorganicproducts.

7=Thislabelstandsforrealorganic products.

Additionalcategory“Idon’tknow”

Table3

Descriptionofthesample:Socio-demographiccharacteristics.

IT n 427 Gender n 427 Female 70.3% Male 29.7% Age n 427 18-44years 42.4% 45-75years 57.6%

Meanageinyears 46.2

Education

n 427

Noformalqualification 0.0% Middleschool(min.8yearsofschool) 1.6% Secondaryschool(min.12yearsofschool) 11.0% Collegeoruniversitydegree 87.4%

Householdsize

n 426

mean 2.8

Householdnetincome

n 426 <600D 4.9% 600Dto<1,200D 14.3% 1,200Dto<1,800D 21.4% 1,800Dto<2,400D 16.7% 2,400Dto<3,000D 13.1% 3,000Dto<3,600D 10.1% 3,600Dto<4,200D 5.4% 4,200Dto<4,800D 4.9% 4,800Dto<5,400D 2.3% 5,400Dormore 6.8%

Thesocio-demographiccharacteristicsofthesampleare pre-sented in Table 3. The level of education was generally high in the sample. However, this result is in accordance with previous studies suggesting that the share of people with a college or university degree is, on average, much higher among consumers of organic food compared to the rest of the population [22–24]. 70% of the respondents were female, reflecting the relationship of foodshoppers in Italy, where mainly women are responsible for food purchases. The mean household size was slightly above average compared to the population.

3. Modelspecification

InthemediationmodelofBaronandKenny[2]thedirecteffect ofZonYisnotedas“Z→Y”andtheindirecteffectofXonZthrough themediatorMas“Z→M→Y.”

Inourcase,wecanpostulatethatorganiclogosmayenhance consumertrusttowardorganicfood,whichinturnmaypositively impacttheWTPfororganicfoodcarryingthoselogos.

FollowingBecker[25]andHuffmanetal.[26],considerthe fol-lowingutilityfunction:

U=U(Q1,Q2,Q3,&,Qi,&,QI−1,QNL;T1,T2,&,Ti,&,TI−1,TNL)

=U(Q,T) (1)

UtilityisbasedonpreferenceofIchoicevariables:organicfood labelledwithvariousorganiclogos(Q1,Q2,Q3,...,QI-1),andwithno

logo(QNL).Theutilityofeachchoiceishypothesizedtobeaffected

bytheleveloftrustassociatedtoeachoftheselogos(T1,T2,...,TI-1, TNL).

Thepricesoforganicfoodwithlogo(ornologo)arep1,...,pI-1, pNL.

(4)

Atanytime,consumersmaximizetheirutilitysubjecttothe conventionallinearbudgetconstraintX:

MaximizeU(Q,T)s.t.

k

pkQk=X (2)

Assuming strictly non-concave utility, we can form the Lagrangeanequationfor(2),thatis

=U(Q,T)+(X−

k

pkQk) (3)

Thefirstorderconditionsare:

U

Qk +

U

Tk =pk

(foreachk=1,2,...,i,NL) (4a)

k

pkQk−X=0 (4b)

Equation(4a)representsthebasisforourmediationanalysis, i.e.theassumptionthatprice–opportunelyscaled–shouldequate thesumofthemarginalutilitiesofchoiceandofthetrustoneach logo.

Undertheassumptionofutility-maximizationbynconsumers facingIalternatives,considertheusualdecompositionofthe over-allutilityofeachalternativei:

Uni=Vni+εni (5)

where εni captures the factors that affect utility that are not

includedinVni.

Inalabelledchoiceexperiment,thelinear-in-parameters rep-resentativeutilityisusuallyspecifiedas:

Vni=˛i+xni ˇ (6)

where,˛irepresentsaconstantthatisspecificforalternativei,xiis

avectorofvariablesthatrelatetoalternativeiasfacedbyconsumer n,andˇarethecoefficientsofthesevariables.

Ifequation(1)holds,andweassumethatthelogosenhance consumertrust,wecanthereforepartitionthevectorxniand

re-writetherepresentativeutilitytotestthemediationeffect:

Vni=˛i+zni+ıTni (6)

where˛irepresentsaconstantthatisspecificforeachalternative

logoi,Tnithemeasuredlevelsoftrustinthelogoiforeachconsumer n,andznithevectorofremainingattributesofeachalternative.

Givenagainthelinearcausalrelationshipinwhich the vari-ableZis presumedtocausethe variableY(Z→Y),a moderator variable (W)–usually a covariate–is a variable that alters the strengthofthecausalrelationship(Baron&Kenny,1986).When thereismediation,ifWmoderatestheindirecteffectofZonY throughM,thismeansthatmediationoftheeffectof ZonYis moderated,somethingreferredtoasmoderatedmediation. Medi-atedmoderation,instead,referstothephenomenoninwhichthe product of Z and a moderator of Z’s effect (W) on Y carry its effectonYthroughM[7].In mediatedmoderation,the moder-ationdisappearswhenthemediatorisintroduced.Inmoderated mediation, thepattern ofmediationvaries asa function ofthe moderator[4].

3.1. BasicmodelandWTP

Sinceonly differencesin utilitymatters[27],whendifferent logosareinvolvedthebasicmodelin(6)maybespecifiedas:

UnEU=ASCnEU+ˇnPRICE+εnEU

UnCCPB=ASCnCCPB+ˇnPRICE+εnCCPB

UnDEM=ASCnDEM+ˇnPRICE+εnDEM

UnNL=ˇnPRICE+εnNL

UnNB=ASCnNB+εnNB

(7)

wherethesubscriptsNLstandsforNologo,andNB“Nobuyoption”, ASCniarethealternativespecificconstant(theNLisnormalizedto zeroforidentification),PRICEisthepriceineurosshowedineach label(whichisobviouslyzerointheNboption).

The random parameter mixed logit model (MXL) for panel observationspopularisedbyTrain[28]obviatesthelimitationsof preferencehomogeneity(allrespondentsaspreferenceclones)and allowsforthemorerealistichypothesisoftastevariationacross respondents.Inamixedlogitmodeltheutilityofindividualifrom alternativejisspecifiedas:

Uij=ˇiXij+eij (8)

WhereXijareobservedvariablesthatrelatetoeachalternative

anddecisionmaker,ˇiisavectorofcoefficientsforthesevariables

forindividualirepresentingthatperson’stastes,andeijisarandom

term.

InMXLspecificationsmarginalWTPforeachattributeofchoice iscalculatedastheratioofcoefficients:

WTP= ˇ(attributelevel)

ˇ(price) (9)

Thiswayofproceedingemploysmodelsspecifiedonthe “pref-erencespace”,whereoftenthepricecoefficientisheldconstant. Thisassumptionmakestheestimationeasierbutrequires undesir-ablerestrictionsonthemodel,suchaconstantmarginalutilityof money.

Besidesmodellingtasteheterogeneityinpreferencespace–i.e. byspecifying thedistributionof coefficients intheutility func-tionandthenusingthemtoderivethedistributionofmarginal WTP–wealsoestimatedthemodel usingthegeneralisedmixed logit (GMXL) model by Fiebig et al. [29]. This model extends therandomparameters“mixed”multinomiallogitmodel(MXL) developedbyReveltandTrain[30]andencompassesalltypesof multinomial logitmodels.We specified themodel allowingfor tasteheterogeneityinallvariablesincludingtheASCs2.TheGMXL model allows testing for scale heterogeneity, which was how-evernotpresentinourdata(␶=0,␴=1asfromFiebigetal.[29]). TheMXLmodelis,therefore,anadequaterepresentation ofour data.

However,theGMXLmodelcanbesimplyre-parameterizedsuch thattheparametersaretheWTPforeachattributeratherthanthe utilitycoefficientofeach attribute.Thatis,insteadof theusual approachofparameterizingthemodelin‘preferencespace’(i.e., coefficientsintheutility),themodelisparameterizein‘WTPspace’ [31,32].

Sinceeachrespondentwaspresentedtwochoicesforeach prod-uct,themodelwasestimatedaspaneldata.

(5)

Table4

ResultsoftheMXL&WTP-spacemodelsfororganicapplechoice(mediationanalysis).

Unmediated Mediated

MXL WTP-sp MXL WTP-sp

Parameters RPa Coefficient Estimate |z-values| Estimate |z-values| Estimate |z-values| Estimate |z-values|

ASCEUlogo RP ␮ 2.03 8.81 2.30 7.59 -0.02 0.06 0.21 0.62 ␴ 2.33 6.20 1.79 4.28 1.89 5.77 1.56 5.55 ASCCCPB RP ␮ 1.17 5.40 1.29 4.36 -0.26 0.77 0.00 0.02 ␴ 1.51 4.31 1.58 3.58 0.77 1.27 0.19 0.22 ASCDemeter RP ␮ 1.10 4.00 0.99 2.59 0.42 1.53 0.61 1.75 ␴ 2.48 6.05 3.02 5.57 1.55 3.86 1.74 4.56 Price RP ␮ -0.89 6.77 (-1.15) (5.82) -0.94 6.56 (-1.39) (2.24) ␴ 1.13 4.59 (0.50) (2.14) 1.27 5.10 (0.58) (1.60) ASCNo-buy NR ␮ -9.84 6.11 7.13 5.43 -11.68 5.95 11.10 1.55 TRUST RP ␮ – – 1.75 8.15 1.87 7.43 ␴ – – 1.25 5.51 1.43 6.95 Nr.parameters 9 9 10 10

FinalLogLikelihood -981.45 -986.06 -892.91 -898.38

Adj.PseudoR2 0.29 0.28 0.35 0.35

AIC 1980.9 1992.1 1807.8 1820.8

Nr.observations 854 854 854 854427

Nr.individuals 427 427 427

–Termwasnotestimatedinthemodel.Thetermsinbracketsarethecoefficientsandstd.deviationsofpriceinpreferencespaceform. Sinceeachrespondentperformedtwochoices,thenumberofobservationsistwicethenumberofrespondents.

a‘RP’standsforrandomparameters,‘NR’standsfornon-random(fixed)parameters.

3.2. InferenceonMediation

InordertotestthattheTRUSTvariablehasmediatingeffectson thechoice,werefertoBaronandKenny[2]stepwiseprocedure:

Step1:Showthatlogosaffectthechoice.Thisstepestablishes thatthereisaneffectthatmaybemediated.Thismaybeestablished byestimatingthemodelspecifiedin(7).

Step2:Showthatthecausalvariables(thelogosandtheprice) arecorrelatedwiththemediatorTRUST.Thisstepisusually per-formedbyregressingthemediatorovertheindependentvariables, X.Inourdiscretechoicecontext,thistestwasperformedby substi-tutingactualchoicewithTRUSTintheequations(7)3(Train[33], p.6)4.

TRUSTEU=ASCEU+ˇPRICEPRICE+εEU

TRUSTCCPB=ˇPRICEPRICE+εCCPB

TRUSTDEM=ASCDEM+ˇPRICEPRICE+εDEM

(10)

Notethat,sinceTRUSTisonlyratedforthethreelogos,neither theNLnortheNButilitieswereestimated.Therefore,oneoftheASC couldnotbeidentifiedandwedroppedASCCCPBfromtherespective

equation.

Step3:Showthatthemediatoraffectsthechoice,while con-trollingfortheeffectsofthelogos.Totestthis,weneedtoestimate theaugmentedmodel(6),whichmaybespecifiedas:

UEU=ASCEU+ˇPRICEPRICE+ıTRUSTTRUST+εEU

UCCPB=ASCCCPB+ˇPRICEPRICE+ıTRUSTTRUST+εCCPB

UDEM=ASCDEM+ˇPRICEPRICE+ıTRUSTTRUST+εDEM

UNL=ˇPRICEPRICE+εNL

UNB=ASCNB+εNB

(7)

Again,notethat,sinceTRUSTisonlyratedforthethreelogos, thevariabledoesappearinneithertheNLnortheNButilities.

3Forsimplicity,inthisandthefollowingequationswehaveremovedthe

sub-scriptn

4TRUSTbeingacountvariableismodelledasfrequencydata.

Step4:Toestablishthat‘overalltrustonlogo’completely medi-ates the logo-choice relationship, the effect of logos on choice controllingforTRUSTshouldbezero.Inotherwords,ifcomplete mediationtakesplace,theASCcoefficientsshouldbezero,whileif partialmediationtakesplace,thecoefficientsoftheASCshouldbe reducedcomparedtotheonesestimatedinStep2.

3.3. Moderationanalysis

Totestthehypothesisthatthemediationeffectsarenot moder-atedbyothermeasuredcovariates,wehaveinteractedthepotential moderatorvariables(covariates)withtherandomparameters(ASC coefficients,PRICE,andTRUST).Doingthisineffectdecomposes anyheterogeneityobservedwithintherandomparameter, offer-inganexplanationastowhythatheterogeneitymayexist[34]. Besides,weconductedthesameanalysisintheunmediatedcase too,tocheckforsimplemoderatingeffectsofthesecovariatesover choice.Ifmoderationexistsintheunmediatedmodelandthen dis-appearinthemediatedone,wecanspeakofmediatedmoderation. Instead,weareobservingamoderatedmediationmodelif media-tionismoderatedbycovariates,i.e.mediationisstrongerforone group(e.g.males)thanforanother(e.g.females).

Table5

ResultsoftheMXLmodelforTRUST.

MXL

Parameters RPa Coefficient Estimate |z-values|

ASCEUlogo NR ␮ 0.32 40.97

ASCDemeter RP ␮ -0.91 69.20

␴ 1.37 84.19

Price NR ␮ 0.01 1.41

Nr.parameters 4

FinalLogLikelihood -3422.77

Adj.PseudoR2 0.06

AIC 6853.5

Nr.observations 838

Nr.individuals 427

Sinceeachrespondentperformedtwochoices,thenumberofobservationsistwice thenumberofrespondentslessmissingobservations.

a‘RP’standsforrandomparameters,‘NR’standsfornon-random(fixed)

(6)

Table6

ResultsoftheMXLmodelsfororganicapplechoice(moderationanalysis).

Unmediated Mediated

Parameters RPa Coefficient Estimate |z-values| Estimate |z-values|

ASCEUlogo RP ␮ 2.19 3.61 0.38 0.89 ␴ 2.21 6.12 1.85 5.66 ASCCCPB RP ␮ .17 0.30 -0.08 -0.2 ␴ 1.59 4.08 0.50 0.68 ASCDemeter RP ␮ 1.41 2.20 0.78 1.95 ␴ 2.37 6.19 1.40 3.84 Price RP ␮ -1.02 3.18 -0.93 6.67 ␴ 1.09 4.44 1.22 5.08 ASCNo-buy NR ␮ -9.76 6.10 -10.40 6.45 TRUST RP ␮ – – 1.67 6.31 ␴ – – 1.27 5.43 SHOPXASCEU HM ␮ -0.34 0.82 -0.54 0.96 SHOPXASCCCPB HM ␮ 0.47 1.44 -0.02 0.04

SHOPXASCDemeter HM ␮ -1.39 2.95 -0.46 0.95

TOWNXASCEU HM ␮ 0.02 0.09 – –

TOWNXASCCCPB HM ␮ 0.36 1.44 – –

TOWNXASCDemeter HM ␮ 0.20 0.70 – –

SHOPXPrice HM ␮ 0.05 0.24 – –

TOWNXPrice HM ␮ 0.04 0.32 – –

SHOPXTRUST HM ␮ – – 0.03 0.12

Nr.parameters 17 15

FinalLogLikelihood -970.32 -892.76

Adj.PseudoR2 0.29 0.35

AIC 1974.6 1815.5

Nr.observations 854 854

Nr.individuals 427 427

–Termwasnotestimatedinthemodel.

Sinceeachrespondentperformedtwochoices,thenumberofobservationsistwicethenumberofrespondents.

a‘RP’standsforrandomparameters,‘NR’standsfornon-random(fixed)parameters,‘HM’standsforheterogeneityinmeanparameter.

4. Results

4.1. TestingforMediation

Table4reportsMXL(preference-space)andGMXL(WTP-space) modelestimationresultsforapples,bothestimatedusing200 Hal-tondraws.Allmodelsallowforaconstantfixedeffectfortheno-buy alternative,whileallrandomparametersareassumedtobe nor-mal.TheWTPspacemodelparametersarealreadyWTPestimates. InthefirsttwocolumnsfromtheleftinTable4theunmediated modelresultsspecifiedby(7)arereported.TheserefertoBaron andKenny’s[2]Step1.Intheremainingtwocolumnsontheright theresultsofBaronandKenny’sSteps3and4arereported,as specifiedin(7).

AllASCparametersreportedinTable4forbothMXL(preference space)andWTP-spaceunmediatedmodelsarehighlysignificant, showingtheroleoflogosinexplainingconsumerpreferencesof organicapples.Allstandarddeviationparametersarealsohighly significant,whichisconsistentwiththehypothesisoftaste hetero-geneity.TheWTP-spaceestimatesaresubstantiallysimilartothose thatmaybecalculatedbytheunconditionalparameterestimatein theMXLspecificationbytakingtheratioofeachASCparameter estimateandtheabsolutevalueofthepriceparameterestimate5. TheyrepresenttheadditionalWTPwithrespecttothe‘NoLogo’ label(referencealternative).Asfoundinpreviousstudies[13,35], theEUlogowashighlyregardedbyItalianconsumersasameansto guaranteeorganicfoodproducts.Theotherlogosarelesserknown, andingeneralscorealoweradditionalWTP.

InTable5wereporttheMXLresultsofregressingTRUSTover logosASCandprice(BaronandKenny’sStep2),estimatedusing200 Haltondraws.Allrandomparametersareassumedtobenormal. SinceestimatedstandarddeviationsofEUASCandpricewerefound

5 These,fortheunmediatedMXLmodelare:EU2.27D,CCPB1.30D,Demeter1.23D.

tobeinsignificantinrestrictiontests,theywererestrictedtozero, implyingfixedcoefficientsandanabsenceofheterogeneity.Not allrespondentsratedthelogosfortrustworthinessandcredibility, resultingin14missingobservations.

TheresultsreportedinTable5showthatindeedtheoverall TRUSTinthelogoissignificantlycorrelatedwiththelogoitself.The EUlogoistheonethatexhibitshighertrust,withnoapparent het-erogeneityamongItalianrespondents,whiletheDemeterlogo–on average–isnegativelycorrelatedwithtrust(itiswell-knownonly byaminorityofrespondents:seeJanssenandHamm[13]).Overall modelfitislow,sincethisisnotafullyspecifiedmodelofTRUSTand isonlyusedtoestablishtherelationshipbetweenTRUSTandthe explanatoryvariables(logoASCsandprice)inthelabelledchoice model.

Theseresultsarefurtherconfirmedbytheresultsofthe medi-ationanalysisreportedinthelasttwocolumnsofTable4.Once thevariableTRUSTisaccountedforinthemodels,theASC param-eter estimates become non-significant (i.e. zero)6, both in the preference-spaceandtheWTP-spacemodels.Notablythestandard deviationsof parametersremain significant(with theexclusion of theCCPB ASC in thepreference space model),showing that consumersexhibittasteheterogeneityandtheparametersare dis-tributednormallyaroundzero.

Themodeldoesnotrejectthehypothesisofcompletemediation ofthelogoeffectsonchoicethroughtheTRUSTvariable.Thiseffect

6Kenny,inhismediationwebsite[4],warnsthatthecoefficientsshouldbezeroin

value,andnotsimplystatisticallynon-significant,astheywereinBaronandKenny [2],“becausetriviallysmallcoefficientscanbestatisticallysignificantwithlarge samplesizesandverylargecoefficientscanbenon-significantwithsmallsample sizes”.However,inourchoicemodeltheTRUSTvariablemaysufferofendogeneity andthiswouldresultininconsistent(butnotinefficient)estimatesofbandc’: sincethesamplesizeislarge,andwehaveestimatesofboththeparameters(which arenonsignificant)andofthestandarddeviationsoftheparameterdistributions (whichinsteadarehighlysignificant),wecanbefairlyconfidentinourconclusions regardingcompletemediation.

(7)

issharedbyalllogos,thoughitisclearthatthehighesteffectis ontheEUlogo,whichexhibitedthehighestWTP.Inotherwords, itappearsfromTable4thatconsumers’trustinlogoiscompletely entangledwiththelogoitself.Indeedtheorganiclogoisanextrinsic cueforthe(organic)qualityoftheproduct(inourcase,applesor eggs).

4.2. TestingforModeration

Wehavetestedthemoderationeffectoftwocovariates:SHOP& TOWN.Thefirstisadichotomousvariable(0=specialisedorganic shop,1=conventionalsupermarket.Thelatterisacategorical vari-able referring to where interviews where collected (Northern, Central,andSouthernItaliantowns).

SinceTOWNwasnotsignificantasmoderatorintheunmediated model,wehaverunthemediatedmoderationanalysisonlywith thevariableSHOPasmoderator.Table6reporttheresultsofour estimationsinpreferencespace.

Intheunmediatedmodel,SHOPappearstobeasignificant mod-eratorfortheDemeterlogo.Indeed,aswehavealreadymentioned, thislogoismorefrequentlyseeninspecialisedorganicshop(and consumersbuyinginthoseshopsaremoreawareofit).The nega-tivesignintheinteractiontermmeansthatthemarginalutilityof Demeterlogoishigherforthosebuyinginthespecialisedshop.

However,oncelogos aremediatedby TRUST,SHOPstopsto explainpreferenceheterogeneityinDemeterlogoASCs:the mod-erationeffectofSHOPismediated,too.Inotherwords,trustin logoexplains(causes)thechoices(andrelativeWTPs), notwith-standingthekindofpoint-of-purchasethelogoarefound.Wecan concludethatthemediationeffectoftrustinlogodoesnotvary acrosspoints-of-purchaseorregions.

5. DiscussionandConclusions

Inthispaper,wehaveshownhowmediationandmoderation analysiscanbesimplyintegratedinadiscretechoicesetting.By ournovelapproach,onecantesttheroleofmediatingand mod-eratingvariables onconsumerchoiceandrelated WTP,using a generalframeworktomodelpreferenceheterogeneity,i.e.without imposingunrealisticconstraintsonthemodel.

Specifically,ourstudytestedthehypothesisthattrustinthe (organic)logomediatestherelationshipbetweenthelogoand con-sumerchoice(andrelatedWTP)fororganiclabelledfoodproducts. Ourfindingsconfirmthefactthattrustinlogoplaysamajorrolein consumers’choiceoforganicapples(andeggs),andthatthereason laysinthatthelogosareperceivedascuestotrustworthyorganic quality[35,36].Trustisa multidimensionalconcept [37],which facilitatesandpositivelyinfluencesconsumerspurchaseintentions andchoices[38].Inthecaseoffood,itisanessentialelementfor decision-makingandisdirectlyrelatedwithfoodsafety percep-tions[39,40].Consumersperceiveorganiclogosasanimportant sourceoftrust[41,42],whileZanoli&Naspetti[43]haveshown thattrustin logoisdirectlyassociatedwithconsumers motiva-tionstopurchaseorganicfood.Usingourapproach,sincetrustin logototallymediatestheeffectofthelogo,WTPfororganic prod-uctscouldbeinterpreted–given(9)–as“costfortrust”,sincetrust totallymediates thelogos (ASCs).Inotherwords,trust embed-dedinthelogosfullyexplains(“causes”,inmediationsense)WTP, andthehigherthetrustthehighertheperceivedvalue-for-money. Therefore,consumers’perceivedacceptablepriceorinternal refer-enceprice[44]isdependentontrustinlogo;althoughthisprice mayvaryacrosspoints-of-purchaseand/orregions,themediation effectoftrustinorganiclogo–aswehaveproven–doesnot.

We believethat thefindings of ourstudyhave implications thatgobeyondourlimitedexperimentonorganiclogos.Usingour

approachonecaneasilytesttheeffectofbrandimage(asembedded inlogosbutalsoinbrandnames)onchoiceandWTP,byusingtrust asmediator.Buttheapproachissusceptibleoffurtherextensions, wheninvestigatingtheimpactofmediating/moderatingvariables onWTP.

The advantage to use our approach is that choice experi-mentsallowarigorous–albeitveryrealistic–investigationofhow consumerchoicesaremade(rectiusstated)givenhypothetical situ-ations.Forexample,Laranetal.[45]testedthediverseprimingrole ofbrandsandslogansonWTP.Theirstudywasperformedusing asimpleANOVAapproachtoanalysetheeffectofthepersuasive intentofdifferentstimuli (slogans,sentences,brandnamesand logos)onWTP7.ThestimuliandtheWTPwheremeasuredin sepa-rate,subsequentexperiments.TheWTP–giveneachstimulus–was measuredinawaynotconsistentwithutilitymaximization.Ina discretechoicesetting,theirmediationandmoderationanalyses wouldhavebenefittedofarigorousrepresentationofutilitybased onasetofattributes(thedifferentstimuli)andtheirlevels(the persuasiveintent),whileallowingadirectmeasureofWTP.

Furtherextensionsofourmodelmayconsidertheuseof “struc-turalchoicemodelling”[46],anapproachthatintegrates choice modelsandstructuralequationmodels,andallowscombiningdata fromseparatebutrelatedchoiceexperiments.

References

[1]C.M.Judd,D.A.Kenny,Processanalysis:Estimatingmediationintreatment evaluations,EvaluationRev.5(1981)602–619.

[2]R.M.Baron,D.A.Kenny,Themoderator-mediatorvariabledistinctioninsocial psychologicalresearch:conceptual,strategic,andstatisticalconsiderations,J. PersonalitySoc.Psychology51(1986)1173–1182.

[3]P.A.Frazier,A.P.Tix, K.E.Barron,Testing moderatorandmediatoreffects in counseling psychology research, J. Counseling Psychology 51 (2004) 115–134.

[4]Kenny, D.A. (2014). Mediation, http://davidakenny.net/cm/mediate.htm [accessedonSeptember2014].

[5]D.P.MacKinnon,Introductiontostatisticalmediationanalysis,Lawrence Erl-baumAssociates,Inc,Mahwah,NJ,2008.

[6]A.F.Hayes,BeyondBaronandKenny:StatisticalMediationAnalysisintheNew Millennium,Commun.Monogr.76(4)(2009)408–420.

[7]A.F.Hayes,IntroductiontoMediation,Moderation,andConditionalProcess Analysis:ARegression-BasedApproach,GuilfordPress,NewYork,NY,2013. [8]L.Valeri,T.VanderWeele,Mediationanalysisallowingforexposure-mediator

interactionsandcausalinterpretation:Theoreticalassumptionsand imple-mentationwith SAS andSPSS macros, PsychologicalMethods 18(2013) 137–150.

[9]S.Naspetti,R.Zanoli,Organicfoodqualityandsafetyperceptionthroughout Europe,J.FoodProdMarketing15(3)(2009)249–266.

[10]K.B.Monroe,Pricing–MakingProfitableDecisions,McGraw-Hill,NewYork,NY, 1991.

[11]V.A.Zeithaml,Consumerperceptionsofprice,qualityandvalue:ameans-end modelandsynthesisofevidence,J.Marketing52(1988)2–22.

[12]Hayes,A.F.,&Scharkow,M.(2013,forthcoming).TheRelative Trustworthi-nessofPopularInferentialApproachestoTestingIndirectEffectsinStatistical MediationAnalysis:DoesMethodReallyMatter?PsychologicalSci.,accepted forpublication.

[13]M.Janssen,U.Hamm,Productlabellinginthemarketfororganicfood: Con-sumerpreferencesandwillingness-to-payfordifferentorganiccertification logos,FoodQualPreference25(1)(2012)9–22.

[14]A.Krystallis,G.Chryssohoidis,Consumer’swillingnesstopayfororganicfood: factorsthataffectitandvariationperorganicproducttype,Br.FoodJ.107(4/5) (2005)320–323.

[15]M.L.Loureiro,W.J.Umberger,Achoiceexperimentmodelforbeef:WhatUS consumerresponsestellusaboutrelativepreferencesforfoodsafety, country-of-originlabelingandtraceability,FoodPolicy32(4)(2007)496–514. [16]L.Lockshin,W.Jarvis,F.d’Hauteville,J.P.Perrouty,Usingsimulationsfrom

dis-cretechoiceexperimentstomeasureconsumersensitivitytobrand,region, price,andawardsinwinechoice,FoodQual.Preference17(3–4)(2006) 166–178.

[17]J.L.Lusk,T.C.Schroeder,Arechoiceexperimentsincentivecompatible?Atest withqualitydifferentiatedbeefsteaks,Am.J.AgricEconomics86(2)(2004) 467–482.

[18]R.Dhar,I.Simonson,Theeffectofforcedchoiceonchoice,J.MarketingRes.40 (2003)146–160.

(8)

[19]Willer,H.,&Kilcher,L.(Eds.)(2012).TheWorldofOrganicAgriculture- Statis-ticsandEmergingTrends2012.ResearchInstituteofOrganicAgriculture(FiBL), Frick,andInternationalFederationofOrganicAgricultureMovements(IFOAM), Bonn(supplementarymaterialsat: http://www.organic-world.net/yearbook-2012.html)

[20]R.Sassatelli,A.Scott,Novelfood,newmarketsandtrustregimes.Responses totheerosionofconsumers’confidenceinAustria,ItalyandtheUK,European Societies3(2)(2001)213–244.

[21]A.Gracia,L.M.Albisu,FoodConsumptionintheEuropeanUnion:Main Deter-minantsandCountryDifferences,Agribus17(4)(2001)469–488.

[22]K.Zander,U.Hamm,Consumerpreferencesforadditionalethicalattributesof organicfood,FoodQualPreference21(5)(2010)495–503.

[23]M.Wier,K.O’DohertyJensen,L.M.Andersen,K.Millock,Thecharacterof demandinmatureorganicfoodmarkets:GreatBritainandDenmarkcompared, FoodPol.33(5)(2008)406–421.

[24]K.O.Jensen,S.Denver,R.Zanoli,Actualandpotentialdevelopmentofconsumer demandontheorganicfoodmarketinEurope.NJAS–Wagening,J.LifeSci.58 (3–4)(2011)79–84.

[25]G.S.Becker,AccountingforTastes,HarvardUniversityPress,Cambridge.MA, 1996.

[26]W.E.Huffman,M.Rousu,J.F.Shogren,A.Tegene,WhoDoConsumersTrustfor Information:TheCaseofGeneticallyModifiedFoods?Am.J.AgricEconomics 86(5)(2004)1222–1229.

[27]K.Train,DiscreteChoiceMethodswithSimulation,SecondEd.,Cambridge Uni-versityPress,NewYork,NY,2009.

[28]K.E.Train,Recreationdemandmodelswithtastevariation,LandEconomics74 (1998)230–239.

[29]D.Fiebig,M.Keane,J.Louviere,N.Wasi,TheGeneralizedMultinomialLogit Model:AccountingforScaleandCoefficientHeterogeneity,MarketingSci.29 (3)(2010)393–421.

[30]D.Revelt,K.Train,Mixedlogitwithrepeatedchoices:Households’choices of appliance efficiency level, Rev. Economics Statistics 80 (4) (1998) 647–657.

[31]K.Train,M.Weeks,DiscreteChoiceModelsinPreferenceSpaceand Willing-to-Payspace,in:R.Scarpa,A.Alberini(Eds.),ApplicationsofSimulationMethods inEnvironmentalandResourceEconomics,SpringerPublisher,Dordrecht,The Netherlands,2005,pp.1–16,chapter1.

[32]R.Scarpa,M.Thiene,K.Train,Utilityinwillingnesstopayspace:atoolto addressconfoundingrandomscaleeffectsindestinationchoicetotheAlps, Am.J.Agric.Economics90(4)(2008)994–1010.

[33]K.Train,QualitativeChoiceAnalysis:Theory,Econometrics,andanApplication toAutomobileDemand,MITPress,Cambridge,MA,1986.

[34]D.A.Hensher,J.M.Rose,W.H.Greene(Eds.),Appliedchoiceanalysis.APrimer, CambridgeUniversityPress,Cambridge,UK,2005.

[35]Naspetti,S.,Hamm,U.,Janssen,M.,&Zanoli,R.(2009).Quantovaleil mar-chiobio?Un’indagineedonimetrica.AgricolturaBiologica:sistemiproduttivie modellidicommercializzazioneediconsumo,ProceedingsoftheIVWorkshop delGruppodiRicercainAgricolturaBiologica-Italia(GRAB-IT),Universitàdegli StudidiPalermo-DESA,Palermo,173-178.

[36]R.Zanoli,R.Scarpa,F.Napolitano,E.Piasentier,S.Naspetti,V.Bruschi,Organic labelasanidentifierofenvironmentallyrelatedquality:Aconsumerchoice experimentonbeefinItaly,Renew.Agric.FoodSyst.28(1)(2013)70–79. [37]C. Moorman,R. Deshpande,G. Zaltman,Factorsaffectingtrustinmarket

researchrelationships,J.Marketing57(1993)81–102.

[38]V.A.Zeithaml,A.Parasuraman,L.L.Berry,Deliveringqualityservice:Balancing customerperceptionsandexpectations,FreePress,NewYork,1990. [39]J.M.Green,A.Draper,E.A.Dowler,G.Fele,V.Hagenhoff,M.Rusanen,T.Rusanen,

PublicunderstandingoffoodrisksinfourEuropeancountries:aqualitative study,EuropeanJ.PublicHealth15(5)(2005)523–527.

[40]R.Hughner,P.McDonagh,A.Prothero,C.Shultz,J.Stanton,Whoareorganic foodconsumers?Acompilationandreviewofwhypeoplepurchaseorganic food,J.ConsumerBehav.6(2007)94–110.

[41]H.Torjusen,L.Sandstad,K.O’DohertyJensen,U.Kjaernes,Europeanconsumers’ conceptions of organicfood: areview of availableresearch, Professional Report,vol.4,SIFO-NationalInstituteforConsumerResearch,Oslo,2004, http://orgprints.org/2490/.

[42]L.Hamzaoui-Essoussi,L.Sirieix,M.Zahaf,Trustorientationsintheorganic fooddistributionchannels:AcomparativestudyoftheCanadianandFrench markets,J.RetailingConsumerServices20(2013)292–301.

[43]R.Zanoli,S.Naspetti,Consumermotivationsinthepurchaseoforganicfood:a means-endapproach,Br.FoodJ.104(8)(2002)643–653.

[44]J.P.Peter,J.C.Olson,K.G.Grunert,ConsumerBehav.MarketingStrategy, Euro-peanEdition,McGraw-Hill,Maidenhead,1999.

[45]J.Laran,A.N.Dalton,E.B.Andrade,Thecuriouscaseofbehavioralbacklash.Why brandsproduceprimingeffectsandslogansproducereverseprimingeffects,J. ConsumerRes.37(2011)999–1014.

[46]C.M.Rungie,L.Coote,J.J.Louvier,StructuralChoiceModelling:Theoryand ApplicationstoCombiningChoiceExperiments,JChoiceModelling4(3)(2011) 1–29.

References

Related documents

Investigating local elections in more than 3000 municipalities in 1995, 1999 and 2003, Bosch en Solé-Ollé (2004) find that property tax increases have a negative impact on

Studiul statistic asupra nivelului de recuperare funcțională a pacienților cu endoprotezare totală de șold; (6) Urmărirea recuperării funcționale a pacienților cu

Module 1 Paola Bellaviti | LabCoopInt - DAStU, Politecnico di Milano Module 2 Rachele Radaelli | LabCoopInt - DAStU, Politecnico di Milano Module 3 Camillo Magni |

We investigated several behavioural parameters at five different time points in the lifetime of the holometabolous mustard leaf beetle Phaedon cochleariae (Coleoptera:

Indeed, in addition to representing a story that would appeal to most young spirits, Brizzi’s use of language corresponds perfectly to the characteristics of

This study explored the perceptions of nurses in decision-making process in palliative care for patients with cancer. Eight themes emerged from this study included

Due to the continued interest and debate in this area, we conducted a scoping systematic literature review and meta-analysis of all observational studies

By applying the particle-number projection to the finite-temperature BCS the- ory to the state densities in the rare-earth nuclei and comparing its results to the SMMC ones,