7/28/2019 Talarzyk Attitude Model

1/5

$Q$WWLWXGH0RGHOIRUWKH6WXG\RI%UDQG3UHIHUHQFH$XWKRUV)UDQN0%DVVDQG::D\QH7DODU]\N6RXUFH-RXUQDORI0DUNHWLQJ5HVHDUFK9RO1R)HESS3XEOLVKHGE\American Marketing Association6WDEOH85/http://www.jstor.org/stable/3149618 .

$FFHVVHG

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=ama. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to

Journal of Marketing Research.

http://www.jstor.org/action/showPublisher?publisherCode=amahttp://www.jstor.org/stable/3149618?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=amahttp://www.jstor.org/action/showPublisher?publisherCode=amahttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/3149618?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=ama

7/28/2019 Talarzyk Attitude Model

2/5

An Attitude Model for the Study ofBrand Preference

FRANKM. BASS and W. WAYNE TALARZYK*

Prediction of individual preference is a difficult andelusive task; it is an important task, however, since itrepresents a fundamental step in understanding con-sumer choice. Asking whether or not preference can bepredicted on the basis of knowledge of the consumer andhis characteristics is a prelude to identifying the causesof preference and the means by which it can be influ-enced.This study of consumer brand preference was an ap-plication of a model of consumer attitudes; the basichypothesis was that measures specific to the preferencealternatives, rather than more general measures such asthose of socioeconomic and personality characteristics,would lead to successful predictions. While the approachis intuitively appealing and seemingly obvious, this studyis the first to publicly present results from testing thehypothesis.PREVIOUS STUDIES

Much researchin the area of market segmentationhasused measures of consumer behavior which are notproduct- or brand-specific. The results have been lessthan encouraging in understanding or predicting brandpreference. Studies of segmentation on the basis of per-sonality characteristicshave had negative or inconclusiveresults [2, 8, 12]. Virtually no association between per-sonality, socioeconomic variables, and the householdbrand loyalty was found in a study of toilet tissue pur-chases [6]; a similarconclusion resultedfrom researchonhousehold purchases of beer, coffee, and tea [9].If brand preference is explained by attitudes-madeup of perceptions of and values for product attributes-the distributionof these variables among socioecomonicand personality segments is not necessarily systematic orregular.While socioeconomic variablesmay be related tobrandpreferences in an aggregatesense (in fact, we havefound such relationships for some product categories),and while these relationships may be useful for some

managerial purposes [1], they are not sufficiently strongto predict individual preference.The Computational Model

None of the numerous definitions of attitude and atti-tude models in the social psychology literature apply di-rectly to the issue of brand preference, since researchershave apparently been reluctant to compare a person'srelative attitude about one object with his attitude aboutanother. Social psychologists have been concerned withevaluating attitudes toward an object, rather than withrelative attitudes toward a group of objects.This study extended an attitude model to the compari-son of individuals' preference ordering of brands. Sinceprevious studies have demonstrated that demographic,personality, and general attitudevariables do not predictindividual brand preference well, it is important to as-certain whether or not attitudes, as measured by beliefsabout specific attributes of brands, predict individualpreference.The computational model applied, whose two compo-nets are beliefs about attributes and evaluative aspectsof the beliefs, was developed by Fishbein [4]. Beliefabout a concept is defined as the probability that a spe-cific relation exists between the concept and an object(e.g., toothpaste prevents decay and whitens teeth). Theevaluative aspects of a belief reflect the importance as-signed to the concepts in forming an attitude toward anobject.For the purposes of this research, Fishbein's model isrepresented quantitatively as:N

Ab = WiBib=1where:

Ab= the attitude oward a particularbrandbWi = the weightor importance f attributeBib = the evaluative spector belieftowardattribute forbrandbN = the numberof attributesmportantn the selectionof a givenbrand n the given productcategory.

A consumer's attitude toward a particular brand is hy-pothesized to be a function of the relative importance of

* Frank M. Bass is Professor of Industrial Administration,Purdue University, and W. Wayne Talarzyk is Assistant Profes-sor of Marketing,The Ohio State University. They thank theAAAA EducationalFoundation for its support of this research.93

Journal of Marketing Research,Vol. IX (February 1972),93-6

7/28/2019 Talarzyk Attitude Model

3/5

94 JOURNAL OF MARKETINGRESEARCH, EBRUARY972Table 1

PRODUCT TTRIBUTESND BRANDSFrozen orange juice Mouthwash ToothpasteTaste/flavor Kills germs Decay preventionPrice Taste/flavor Taste/flavorTexture Price Freshens mouthNutritional value Color Whitens teethPackaging Effectiveness PriceMinute Maid Micrin PepsodentSnow Crop Cepacol CrestBirds Eye Listerine GleemA & P Lavoris ColgateSunkist Colgate 100 MacleansToilet tissue Lipstick BrassieresTexture Color StyleColor Taste/flavor PricePrice Prestige factor ComfortPackage size Container FitStrength Creaminess LifeAurora Hazel Bishop Penney'sDelsey Max Factor PlaytexNorthern Avon LovableScott Coty MaidenformCharmin Revlon Sears

each of the product attributes and the beliefs about thebrand on each attribute.METHODOLOGY

Consumers were asked directly about specific attri-butes of products and beliefs about brands, and this in-formation was used to predict the preference order forthe brands. Five attributes were specified for each of sixproduct categories studied, on the basis of informal in-terviews. Weights for the evaluative component were de-termined from each respondent's forced ranking of theimportance of the attributes in the selection of a brand.Beliefs about the attributes for individual brands weremeasured by having respondents provide a scaled valuefrom 1 to 6 from "very satisfactory" to "very unsatis-factory" on each attributefor each brand.An alternative methodology is suggested by theemerging literatureon nonmetric multidimensional scal-ing. Some exploratory examples of this approach haveused a basic hypothesis similar to ours [5, 7]. However,we believe our direct approach provides a stronger ini-tial test of the underlying hypothesis.Ideally, it would have been desirable to study the re-lationship between attitudes and brand purchasing be-havior, but this posed methodological and financial dif-ficulties. Preference clearly does not convert directly intopurchasing behavior, but the two are related [1]. Fur-thermore, since attitudes should be a weaker predictorof purchase than preference, the result is conclusive withrespect to actual behavior if the hypothesis that attitudespredict preference is rejected.

DATAA national sample of 2,000 female heads of house-holds from the Consumer Mail Panel of Market Facts,Inc., was used to provide a balanced sample parallel tocensus data for geographic divisions and within each di-vision by total household income, population densityand degree of urbanization,and age. Preferences and at-titudes toward individual brands were measured for six

product categories: frozen orange juice, mouthwash,toothpaste, toilet tissue, lipstick, and brassieres. Five at-tributes were specified for each product category on thebasis of informal interviews (Table 1); preferences weremeasured for the five leading brands in each category.Respondents were also asked whether or not their favor-ite brand was included in the set of five brands and tospecify theirusage rate for each product. Standard socio-economic data were also available for each householdsurveyed [11].Of the total sample, 78.5% responded to the ques-tionnaire; 63.6% returnedquestionnairesusable for theentire analysis. Socioeconomic segments were repre-sented in the final sample in approximately the sameproportions as represented in the panel. The predictiontest of the model focused upon the subset of respondentsfor each product category who were users of the productcategory and whose favorite brand was included in theset of five brands studied.

PREDICTION RESULTSIndividuals' preference orderings of the brands werepredicted using relative attitude scores derived by theFishbein model. In cases of distinct attitude scores for

each brand, the predicted preference ordering wasunique. Ties were resolved by randomization or marketshare. Table 2 is a sample confusion matrixwhich showsthe conditionalprobabilityof actual rank given predictedrank for mouthwash.Equally successful predictionswereobtained for the other five product categories.While the market share scheme for ties producedsomewhat better predictive results than the randomscheme, prediction was quite good in both instances.Table 2

CONDITIONALROBABILITYF ACTUALRANK GIVENPREDICTEDANKFORMOUTHWASHaPredicted Actual rankrank 1 2 3 4 5

1 .687 .174 .084 .035 .0212 .177 .472 .206 .097 .0483 .082 .210 .387 .205 .1164 .038 .105 .217 .419 .2215 .015 .040 .106 .244 .595a Users whose favorite brands were included. Ties wereallocated on a market share basis

7/28/2019 Talarzyk Attitude Model

4/5

AN ATTITUDEMODEL FOR THE STUDY OF BRAND PREFERENCE 95Table 3

PROBABILITYF CORRECTLY REDICTINGMOST PREFERREDRANDS FOR VARIOUS MODELSProduct

Model Frozen orange Mouthwash Toothpaste Toilet Lipstick Brassieresjuice Mouthwash Toothpaste tissueAttitudes .67a .69% .75a .63a .70a .72aMultiple discriminant analysis (using .58 .58 .63 .56 .63 .58beliefs)Multiple discriminant analysis (using .52 .49 .57 .42 .53 .56demographics)Market share .53 .44 .46 .36 .46 .39Random .20 .20 .20 .20 .20 .20

a Significant at the .01 level when compared to all of the other models.

Notice that in Table 2 the main diagonal is dominant,indicating an ability to predict the entire preference or-der. Also, there was a greater chance for successful pre-diction of the most and least preferred positions than forbrands in between: a predicted first-choice brand wouldactuallybe ranked firstor second with probabilitiesvary-ing from .75 to .90, depending on the product categoryand the tie rule. Furthermore, although not broken outseparately here, predictions for low market share brandswere about as good as for high market share brands.There is, therefore, substantial support for the basic hy-pothesis. Consumers' beliefs and values for product at-tributes,measured for individualbrands, do substantiallyexplain brand preference. (Confusion matrices for allproduct categories and brands are in [11].)

ADDITIONAL FINDINGSTwo additional questions were explored in the analy-sis: how did the attitude model developed here comparewith other models in predicting individual brand prefer-ence? Are there certain consumer characteristicsassoci-ated with the ability of the attitude model to predictpreference?Table 3 compares the predictive results of the attitudemodel with two multiple discriminantanalysis models, amodel based on market shares, and preference predictedon a random basis. The first MDA model utilized thesame information as incorporated in the attitude model,while the second one focused on standard demographiccharacteristics. The naive market share model simplyimplied that the brand with the largest market sharewould be preferred by all consumers. With five brandsin the analysis, the random basis model had a .20 proba-bility of correctly predicting the individual's most pre-ferred brand. For all product categories the attitudemodel was significantlybetter than the other models inpredicting the most preferredbrand.The following generalizations from contingency tableanalysis can be made regardingthe predictive ability ofthe attitude model as compared to that of consumer

characteristics:1. A respondentwho perceivedseveral brandsas simi-lar (equal attitudescores) in one product categorywas likely to perceiveseveralbrandsas similar forthe otherproductcategories.2. If a respondent'spreferredbrand for one productcategory was incorrectly predicted the model wasalso likely to make incorrect predictions for theotherproductcategories.3. Age (older respondents)and education (less edu-catedrespondents)were found to be positivelycor-relatedwith the probabilityof an incorrectprefer-ence prediction or someproductcategories.4. There appeared o be no relationshipbetweena re-spondent'susagerate of a productcategoryand theabilityof the model to predictpreference.

CONCLUSIONSThe research and results reported here strongly sup-port the hypothesis that brand preference is related toattitude measurements based upon beliefs about andrelative importance of product-specific attributes. Theattitude model was shown to result in a greater percent-age of correct brand preference predictions than othermodels tested. While much work needs to be done beforeit will be possible to make strong statements about theimplications of this study, attitude theory does appear tooffer considerable potential as a basis for studies of con-sumer choice behavior.This research has not dealt with the cause-and-effect

relationshipbetween attitude change and change in pref-erence. Festinger and others have raised questions con-cerningthe nature and direction of causation in relationsbetween attitudesand behavior [3]; our view is that eachprobably influences the other. While such conclusionsabout directional causality must await studies of the dy-namic process, we do think that these results are suffi-ciently strong, particularlyin comparison with other re-ported studies based on socioeconomic and personalityvariables, to suggest a basis for a study of dynamics.

7/28/2019 Talarzyk Attitude Model

5/5

96 JOURNAL OF MARKETINGRESEARCH, EBRUARY 972REFERENCES

1. Bass, Frank M., Douglas J. Tigert, and Ronald T. Lonsdale."MarketSegmentation:Group Versus IndividualBehavior,"Journal of Marketing Research, 4 (August 1968), 264-70.2. Evans, Franklin B. "Psychologicaland ObjectiveFactors inthe Prediction of Brand Choice: Ford versus Chevrolet,"Journal of Business, 32 (October 1959), 340-69.3. Festinger, Leon. "BehavioralSupportfor Opinion Change,"Public Opinion Quarterly, 28 (Fall 1964), 404-17.4. Fishbein, Martin. "A Consideration of Beliefs and TheirRole in Attitude Measurement,"and "A Behavior TheoryApproach to the Relations between Beliefs about an Objectand the Attitude towardthe Object," n MartinFishbein,ed.,Readings in Attitude Theory and Measurement. New York:John Wiley & Sons, 1967, 257-66, 389-400.5. Green, Paul E., Frank J. Carmone, and PatrickJ. Robinson.Analysis of Marketing Behavior Using Nonmetric Scalingand Related Techniques. Cambridge: Marketing ScienceInstitute, 1968.6. Kildegaard, Ingrid and Lester Krueger. Are There Con-sumer Types? New York: Advertising Research Founda-tion, 1964.7. Klahr, David. "Decision Making in a Complex Environ-ment: The Use of Similarity Judgments to Predict Prefer-

ences," Report No. 6806, Center for Mathematical Studiesin Business and Economics, University of Chicago, January1968.8. Koponen, Arthur. "PersonalityCharacteristicsof Purchas-ers," Journal of Advertising Research, 1 (September 1960),6-12.9. Massy, William F., Ronald E. Frank, and Thomas Lodahl."BuyingBehavior and Personality,"working paper, Gradu-ate School of Business,StanfordUniversity,June 1966.10. Pessemier, Edgar A., Philip C. Burger, Richard D. Teach,and Douglas J. Tigert. "UsingLaboratoryBrand PreferenceScales to Predict Consumer Brand Purchases," InstitutePaper No. 221, Institute for Research in the Behavioral,Economic and Management Sciences, Krannert GraduateSchool of Industrial Administration, Purdue University,October 1968.11. Talarzyk, W. Wayne. "An Empirical Study of an AttitudeModel for the Predictionof IndividualBrand Preference forConsumer Products," unpublished doctoral dissertation,PurdueUniversity, 1969.12. Tucker, W. T. and John J. Painter. "Personalityand Prod-uct Use," Journal of Applied Psychology, 45 (October1961), 325-9.

13. Westfall, Ralph. "PsychologicalFactors in PredictingProd-uct Choice,"Journalof Marketing,26 (April 1962), 34-40.