Attila Yaprak is Associate Professor of Marketing, Wayne State University, Detroit,. Michigan. His research as been published in such journals as the Journal of ...
A CROSS-NATIONAL COMPARISON OF CONSUMERRESEARCHMEASURES RaviParameswaran* OaklandUniversity AttilaYaprak** WayneState University Abstract.The impactof three factorsthat affect the psychometric rigorof cross-nationalconsumerresearchcomparisonsis measured in this study. Conclusionsare also drawn about the interactions among these factors and the resultant deterioration in the psychometricqualityof the data. The cross-nationalconsumerresearchinstrumentis sensitiveto the natureof the attitudinalconstructs,the nationalityof therespondents, and the country-of-origineffects examinedin the research.The instrumentis alsosensitiveto lower-orderinteractionsbetweenthese factors.The authorsconcludeby discussingways to facilitatemore rigorouscross-nationalcomparisons.
The objectiveof the researchreportedin this paperwas to identify,and measure the impact of, three factorsthat affect the psychometricrigor of cross-national consumerresearchcomparisons.A furthergoal was to measurethe interactions among these factorsto examine the resultantdeteriorationin the psychometric quality of the data. The authors also speculate on how their findings might be used to facilitatemore meaningfulcross-nationalcomparisons. The need for rigorousevaluationof the psychometric, propertiesof cross-national consumer researchmeasureshas been intensifiedas a result of the explosive growth in the multinationalizationof business (Levitt 1983). In this context, Davis, Douglas and Silk (1981) and Sekaran (1983) have argued that, since the principalquestionin cross-nationalresearchis the comparisonof how markets fare against each other, establishingthe psychometricqualitiesof the research instrument,its behavioral/attitudinal constructs,andthe equivalenceof its samples shouldbe issuesof paramountconcernfor multinationalmarketers. * RaviParameswaran isAssociateProfessor ofMarketing, OaklandUniversity, Rochester, Michigan.Hisresearchhasbeenpublishedin suchjournalsas theJournalof Marketing ResearchandtheJournalof theAcademyof Marketing Science. ** AttilaYaprakis AssociateProfessor of Marketing, WayneStateUniversity, Detroit, Michigan.Hisresearch hasbeenpublished in suchjournalsas theJournalof Advertising, International Marketing Review,andManagement International Review. Date Received:July 1985; Revised:April 1986; Accepted:July 1986.
35
Palgrave Macmillan Journals is collaborating with JSTOR to digitize, preserve, and extend access to Journal of International Business Studies ® www.jstor.org
36
JOURNALOF INTERNATIONAL BUSINESSSTUDIES,SPRING1987
Reliableassessment ofcross-national measures is,ofcourse,of fundamental interest suchassessments andtherefore to multinational marketers; mayaffecttheaccuracy, the quality,of strategicdecisions.For example,whetherand to what extent marketing programs arestandardized oradaptedacrossmarkets maybeinfluenced considerably by the reliabilityand the validityof consumers'responsesto the measuresemployed.If the reliability, or precision cross-national interpretation, in fact variesacrossmarketsbut is not detected,conclusions of measurement drawnfroma cross-national surveymay not be valid.Whatmightappearto be a cross-national difference in, say,attitudestructures, mayin realitybe solely in the reliability of underlying measurements a reflectionof variations employed in theanalysis(Davis,DouglasandSilk1981,DouglasandCraig1983).Further, thereliability aseachmarketmaypossess"unique" sociocultural behaviorpatterns, of the behavioralconstructsused in the researchwouldneedto be examined before sound marketingdecisionsare made (Sekaran1983). Alternatively, universally validbehavioral constructs mayfacilitatetheexecutionof pan-cultural but only aftermeasurereliabilityhas been accountedfor marketingstrategies, (DouglasandCraig1983).In addition,inferences regarding nationalinnovation adoptionpropensities mayalsobe subjectto errorwithoutreliablemeasurement. of targetmarketprofilesmaybecomelessmeaningful Inter-country comparisons if the reliabilityor validityof the relevantinputsinto such comparisonsare notproperly accounted for.Also,unreliable measurement ofcross-national attitude structures mayleadto inappropriate positioningof products.Indeed,Cookand Campbell(1979) havearguedthatthe threatsto the psychometric qualitiesof in a cross-national researchmeasuresincreaseexponentially studyas the number and diversityof marketsencompassedin such a projectis expanded(Davis, DouglasandSilk 1981). Yet, a criticalreviewof the cross-national consumerbehaviorliteratureshows a surprising lackof concernfor psychometric rigorin research(Roberts1970). Whilethisis not entirelyunexpectedgiventhatsuchconcernis only of recent vintageeven in intra-national research(Kassarjian marketing 1971;Heelerand Ray 1972;Ryanand Bonfield1975;Rogers1976;Jacoby1976;Peter1979; Parameswaran et al. 1979; Ray 1979; Churchill1979; Bagozzi1980), the of thebehavioral ina multicountry complexity patterns encompassed studydictates that the psychometric propertiesof measuresbe prudentlyconsideredbefore multinational formulating strategies (Sekaran1983;SekaranandMartin1982). Fortunately,the literaturerecordsa handfulof studieswhich have explored variousdimensionsof this methodological issue.Thesestudies(e.g.,Greenand White1976;Bernhart andGroff1979;Davis,DouglasandSilk 1981;Douglas and Craig1983, Adler 1983aand 1983b;Sekaranand Martin1982;Sekaran 1983) haveexhortedcross-national researchers and managersaliketo establish construct, functional, conceptual, instrument,translation,and sampling equivalenceof researchmeasuresprior to inferringstatisticaland practical conclusionsfromtheirfindings.Somehaveevenproposedmethodswithwhich to approachequivalenceor to minimizethe presenceof measurement error (e.g.,backtranslation techniquesdiscussedby Whiting1968;methodsof coping withextremeresponsestylesproposedin Brislin,LonnerandThorndike1973). However,the authorsare awareof only one studythatexploredconceptually,
RESEARCH CONSUMER MEASURES
37
and demonstrated empirically,the differingreliabilitiesof consumerbehavior context(Davis,Douglasand Silk 1981.) In constructswithina cross-cultural that study,cross-national were madefor threetypesof reliabilitycomparisons consumer behaviormeasures householddecisioninvolvement, (demographic, and acrossfive country-markets. psychographic) three different Utilizing reliability assessmentmethods,Davis, Douglasand Silk foundthat:(i) betweenmarket reliabilitydifferentials were relatedto the type of reliabilitymethod,but not to the particularcountry-market used;and, (ii) reliabilitydifferentials varied accordingto the type of constructunderstudy,beingless likelyto occurfor "hard"(demographic), but more likely to occur for "soft"(psychographic) constructs.The principalimplicationof their findingswas that, two sources of measureunreliability methodand natureof the construct)may (assessment confoundthe comparability of cross-national researchfindings; andhence,these shouldbe accountedfor beforemeaningfulconclusionsare drawnfromsuch research. RESEARCHOBJECTIVES
This articlebuildson the groundwork of its predecessors in the cross-national consumerbehaviorarea,particularly on Davis,DouglasandSilk's1981 work. Itspurposeis to encourage psychometric rigorin cross-national consumer research by demonstrating the need for establishing the reliabilityand validityof the measuresof constructsused in the research.Measuresof a constructcan be simple(say, age measuredin years)or complex(say, measuresof a lifestyle domainsuch as fashionconsciousnessor a predisposition towardan object, say the HondaCivic).Like the Davis, Douglasand Silk study,the effectof construct-evoked in thereliability variability of cross-national consumerbehavior measures is exploredin thisarticle.Inaddition,thecurrentauthorsalsoinvestigate the potentialimpact of market-induced reliabilitydifferentialsthat are a of theuniquesociocultural consequence behaviorpatterns, valuesor expectations associatedwiththe nationalsamplesparticipating in the research(Douglasand Craig1983).Suchsample-induced differentials may affectcomparative market studies(Sekaran1983;Sekaranand Martin1982).In the only studyreporting on this effect, however, no noticeablesystematictendencyfor reliability differentials were observedin comparisonsof researchmeasuresacrossfive nationalsamples(Davis,Douglasand Silk 1981). Furthermore, the potential impactof "country-of-origin" effectson the reliabilityof consumerresearch measures(Nagashima1970 and 1977;Lillisand Narayana1974; Whiteand Cundiff1978;Yaprak1978;Bilkeyand Nes 1982;Cattin,JolibertandLohnes 1982) is also explored.In this context,the authorscontendthat differences in "made-in" imagesassociatedwith chosencountries,the natureof historical tiesbetweenthe marketandthe production countries,andcontinuities between nationalproductandproduct-attribute profilesmayperpetuate response-specific reliability differentials amongtherelevantnational-markets' consumers (Douglas andCraig1983;Adler1983;Sekaran1983). In sum, the purpose of the currentpaperis to explorethe degreeto which measureunreliability as a resultof construct-,country-of-origin-, and market-
38
JOURNALOF INTERNATIONAL BUSINESS STUDIES,SPRING1987
of cross-national induceddifferentials research impairthecomparability findings. of the interactions A furthergoal is the measurement amongthesefactorsand on how thesefindingscanbe usedto facilitatemorerigorous recommendations cross-national comparisons. Accordingto Brislin,LonnerandThorndike (1973), of interaction of research themeasurement effectscanenhancethedetermination variableeffectsin a cross-cultural study. METHODOLOGY Data Collection
Datawerecollectedas partof a largerstudyin two countries: the UnitedStates as possible,random and Turkey.To assureas close a samplecomparability samplesof 600 businessexecutiveswere drawnfromthe AtlantaChamberof CommerceDirectory,and the IstanbulChamberof Commerceand Chamber of IndustryDirectories,respectively. were obtainedthrougha Measurements mail questionnaire which,aftertwo mailings,yielded158 usableresponsesin the Atlanta,and202 usableresponsesin the Istanbulsamples,respectively. The Istanbulsamplewas somewhatmoreupscalein demographic dimensionssuch as income,educationand occupationrelativeto the Atlantasample(Yaprak 1978). Questionnaire
The questionnaire, specificallydesignedfor this study,was originallydrafted in English,translatedinto Turkish,back-translated into English,and finally, checkedfor consistencywiththe originalas suggestedby Whiting(1968). This instrument contained twokindsof measures commonlyemployedincross-national research.Respondents' attitudestowardthe peoplesand productsfrom three countries-of-origin (West Germany,Japan,and Italy) were examinedin the two marketcountries(the UnitedStatesand Turkey)wherethe researchwas tenstatements conducted. Through (seeTable1),generalcountryattitudes (GCA) towardeachcountry-of-origin weremeasured foreachrespondent in the market countries.Thesestatements weregeneratedfromiteminventories developedby comparativemarketingresearchers(Boddewyn 1981.) Through fourteen statements(see Table1) selectedfrominventoriesdevelopedby cross-national consumerbehaviorresearchers, generalproductattitudes(GPA) towardeach werealsomeasured country-of-origin (Nagashima 1970,1977;LillisandNarayana White andCundiff1978;BilkeyandNes 1982). 1974; Additionally,respondents'opinionswere sought regardingspecificproduct attributes of threeproductcategories(cars,cameras,andcalculators) originating in eachof thethreecountries-of-origin in linewithpreviousresearch summarized in Bilkeyand Nes (1982).Afterpre-testing in AtlantaandIstanbul,the brands finallyselectedforanalysisweretheVWRabbit(Germany), HondaCivic(Japan), and Fiat 128 (Italy)in the car category;Leica (Germany),Canon(Japan), andFerraria (Italy)inthecameracategory; and,Royal(Germany), Canon(Japan) and Olivetti(Italy) in the calculatorcategory,respectively.Specificproduct attribute(SPA)perceptions weregaugedusingtwenty-four statements (ninefor
RESEARCH CONSUMER MEASURES
0
~
E0
91~~~~~~~~~~-
QO I aa)
39
CO*
0 a) )0 E .
.a
0
w
E
0
Eu 0
j
C0
05
75
co~~~~~~oa
co
c ~~~~a) 3
.0
0~~~~~~~~~~*C
0
~
~
~
0
-
0
E
c~C
a)
-0
~
O
a)
co
~
~
~
~
~
E
C : C: 3
>a)=
a)a)=
-0
-
a)
0Ca)
a) -z.t
0~~~~~~~~~~~~
~~~~~~~CO~
Ea)
aI)o
_oCI
_.
-:
-_7
C:
:3 >
CD
a)
n
To
C a3
Z5c -
a)
1 a)
0
C 2
". Co co
Lug~~~~~~~~~Z 0) -
)
Coo Ca)aa)
--
E
: =
0
co
Co co