31 553335.} Y ‘5‘.“ - ' . mil". 53‘3‘3 szrlijké”; v . ’ J. ~ w dag-.2, 2,); .'. 7 . ‘1'. r ‘ .1"- , ‘at'._ _, '. 3, . ... ”n“: .., , a 5 1 .. €..,.»'-_.‘ - .u ., -. ‘45 \N, . uuv‘ucu .51", ~_ [7) , ”‘9 . x ‘ .. .u. 3; ‘3‘ . «.3 a ~~m.afi~3~'3:~ .. w k , , ,.3 - x . 1 ~~ u". n‘*-M"‘ul ‘ .. \ , .59. . v..-.: 1 .r _ .,. .3 3‘2. ...— . n T ' Mimi. . Al "N‘J u v {_ vuqfigzg. g3 . 'wi” . . . ‘t?‘\..": ‘rt" “ 5* Mm “m 3 an 9.2% 7 fi. . “fijgflgé‘? 3‘ fitém‘fi w: 5 .- 2.; 'L.,~ -- “"1 "M“? r "9-155%, ‘& t ' . . u ‘ “ v ’ ‘« "V , @ATwJ".L..§¢75h¢' ”*3 5',“ I . ‘ ' 1.. ' ,1) u --:.;'i‘ i-fi; ,, ‘ " 1;, ' V’:I.;}."\’~ ‘3 ‘fir'f Jaws-w , > ' x - m an . 1.; .n u .3, m; , 1 My ‘ “ i .3 ~. .. . ,. a A ,5 a; .- u 5 ' g. , - ' “A . ‘ 1' ' ‘- w ' - ., '35- ‘J-hll-r . ‘ u . - .. ‘ n .A' 13'.’ I ‘ ' " as u. f,g. rJ'l-‘v'l ;. u. .. . «at $2"?§F:‘ ‘. , 'v“ n} “0' u :3 ’. “-‘Priafi‘ ‘ a». - 28C; . \v l '1 ,~. 1" ’ 3. .,1., J. .,... -.....,..,., ....- “a .4..." v- ., /_ . ,.H,..,.. ..,..,., 1‘" “1. ‘ew-n Q ‘ V .- ' ”.17“. '42" .. ~. .. ..- .. ,, m...” "5‘; - .‘3- 'E J . , 3 . ' mm -.... .—.. .3 -.... . ,_P_o "c“. .. ..“'...";...‘.‘.~7.L..7 mum,» " 3.. "Wt. ., ...... -. ,. -- ‘ .3. -xv,,.....!,’..',._xb.:._ PIHM ("Inc ’PMOV . -. -...__. 7 . 3,». M . 3.....- "“3... .,. .. "v... .x 3 “my“.-. -.. flu--.” . J. o-;-~ .m-b—u: '4'. .- W . «a m ":74? '.'._.':... 1 . m... .. '3‘... w .4... 0. _. 7 4:3,.” 21-. ‘5 vm~ . "at.“ "4:2,.“ 'u .ww: N "ya“. =7.-- - , :wrmm. Lafithlfiiit‘xf‘ 2'. This is to certify that the dissertation entitled EXPECTATIONS AND CONSUMER SATISFACTION: SOURCES AND INTERACTIONS presented by Sharon Vondra Thach has been accepted towards fulfillment of the requirements for Ph.D. degree in Marketing m Major professor Date M MSUi: “Affirm: iAw n/qu aplOportu nilylnm': 0-12771 PLACE ll RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date duo EATE DUE DATE DUE DATE DUE WM p v » 3 , . A; ‘ fl 4‘3""‘119 2 J '7 1 p ‘ ~ - 2 5 v TI: MSU I: An Affirmative AetIorVEqunl Opportunity Institution EXPECTATIONS AND CONSUMER SATISFACTION: SOURCES AND INTERACTIONS By Sharon Vondra Thach A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Marketing and Transportation 1986 ABSTRACT EXPECTATIONS AND CONSUMER SATISFACTION: SOURCES AND INTERACTIONS By Sharon Vondra Thach This study explored facets in the dominant expectancy model of consumer satisfaction. Two specific issues were addressed: the effect of previous experience on both expectations and satisfaction with a new product offering, and, the relative importance of benefits and attributes of products on satisfaction. One hundred ninety-six consumers were surveyed prior to and following use of a service offering. Analysis of the aggregate patterns was effected through testing. of alternative models as well as null hypotheses related to the model specifications. Results indicate that experiential norms are stronger determinates of satisfaction and future intentions than expectations specific to the new offering. Benefits and attributes are not separate in total effect, although results indicate that information for each is processed differently, suggesting that pre-usage and post- usage evaluations may differ in criteria and weighting of individual factors. This study strongly supports continuation of research into the context of decision-making, both before and after purchase. Understanding consumer behavior as a dynamic process should affect both managerial action and academic research. FOR MY MOTHER AND FATHER ACKNOWLEDGEMENTS There are many who deserve thanks and recognition, but space will allow only the mention of a few. To all those not named, my gratitude is nonetheless heartfelt. The members of my committee should be recognized for whatever is worthwhile herein, as the good is a direct result of their insistence and persistence. Dr. F.S. Carter, my chair, was always positive, helpful, and full of necessary good humor. Dr. Ronald Savitt and Dr. Bixby Cooper made many useful suggestions and were always enlightening. Dr. Arieh Goldman, who directed me to this area and guided my first research, deserves special gratitude. Paul Lane and Catherine Axinn provided support and always listened. Members of the Marketing faculty at Cleveland State University were enormously patient, and Dr. Terrence Parridon was especially helpful. My special love and thanks go to Sarunh and Alistair, who often must have wondered if I still belonged to the family, but cheerfully loved and took we of me anyway. List of Tables List of Figures Chapter One Two Three Four TABLE OF CONTENTS INTRODUCTION Issues LITERATURE REVIEW Theoretical Foundations Consistency-Theories Attitude Implications and Conclusions Marketing Models Research Summary Concepts Summary MODELS, HYPOTHESES, AND DEFINITIONS Constructs and Definitions Structural Models RESEARCH DESIGN AND THE MEASUREMENT MODEL Product Sample Questionnaire Design Development Analysis Procedures Description Construct 312858‘ 37 Five Six ANALYSIS: TESTING HYPOTHESES Discussion Comparison of Research Summary SUMMARY OF RESEARCH, LIMITATIONS AND IMPLICATIONS Findings limitations Implications and Future Research Theoretical Research Appendices A-1 Computer-Generated Data B-l Survey Instrument List of References 100 101 105 107 108 111 112 113 116 116 147 4.1 4.2 5.1 5.2 5.3 5.4. 5.5 5.6 5.7 5.8 5.9 LIST OF TABLES Sample Demographics Scale Reliabilities Model D Model C Model B Model A (dual-factor) Model A (single-factor) Chi-Square Test Model A Path Coefficients Difference in Impact Individual Comparisons, Model A 2.1 2.2 2.4 2.6 2.7 2.8 2.9 2.10 3.1 4.1 LIST OF FIGURES Theorist Oliver’s Model Cadotte, Woodruff, and Jenkins (1982) Cadotte, Woodrufff, and Jenkins (1983) Hypotheses for Experienc-Based Norms The Purchase/Consumption/Evaluation Process Researchers A View of the Development of CS/D Research Summary of Research Approaches Research-Based Causal Model Final Questionnaire Items Final Questionnaire Items 835:883883 8&3 67 CHAPTER ONE INTRODUCTION Consumer satisfaction is the foundation of the marketing concept, which is generally presented a the pursuit of organizational goals through the satisfaction of consumer needs and wants. Academic inquiry into consumer satisfaction/dissatisfaction, designed to elucidate the relationships of marketing activities and consumer evaluations of them, covers a diverse set of interests and theories. Sources of this diversity include the differing aims of theoretical and practical motivation for such research. Academics have embarked on these investigations in attempts to unify a number of behavioral observatiom in a systematic framework with a strong theoretical base. Practitioners desire improved guidelines for new product development, promotional activities, and beneficial after-sale investment. This research has both theoretical and managerial implications. Within marketing, investigation of consumer satisfaction has grown as the paradigm developed showed promise as a way of integrating a number of consumer behavior concepts into a predictive model. It has also served to focus research attention on consumption, use and other post-purchase behaviors and attitude issues. These post-purchase phenomena are important since consumers act through time, with existing attitudes about a product class affecting future choice and consumption activities. This dissertation studies questiom which provide a framework for explaining how existing attitudes and norms affect new consumption evaluations and the implications for future consumption behavior of that single consumption experience. Issues This dissertation will explore the existence of, and, the effects of, product class on consumer satisfaction with a single consumption eXperience. Alternative models for ways in which product class may affect satisfaction will be tested. These models rest on the following argument. Two processes, purchase choice and post-usage evaluation, are similar in that certain conditions must be met for choice and for satisfaction to occur: the belief that the given product is at least as good or better than other choices, and the belief that the benefits sought can be achieved by the use of the product. The two processes will differ in that the information available at the two time periods will differ, particularly with respect to actual comparison of the chosen product with others in its class and with respect to the achievement of the desired benefits. This formulation is appealing because it integrates the two evaluation processes within a framework with parallel evaluative processes It enables us to understand the idea of satisfaction as a summary affective evaluation, the outcome of several separate evaluative conditions. The theoretical arguments and experimental results pertaining to these issues are presented in Chapter II. The second area of investigation concerns the differential impact of characteristics (attributes) and goals (benefits) in models of consumer satisfaction. Generally, most expectations have been treated as applications from the multi-attribute tradition. There is, however, no reason to assume that this is satisfactory if post-purchme judgments are included in models as they are for consumer satisfaction] dissatisfaction. There are two arguments against the use of simple multi-attribute approaches to expectations and satisfaction. First, as Levitt (1980) eloquently argued, all products have tangible and intangible features. Further, the appreciation of those attributes will differ before and after experiencing the product. The use value will be apparent only after consumption. This might be even more true of the intangible aspects. For even the most tangible of products are before they‘re bought, largely promises...Satisfaction later in consumption or use can seldom be quite the same as earlier in trial or promise. (96) 3 To the extent that this is true of a product, therefore, one would expect that the tangible and intangible aspects would differently affect pre-purchase and post-purchase judgments. One would, through time, learn to connect certain observable properties with function and use. Thus, in purchasing, say, a new coffee maker, one would look at size of pot, color, number of dials, automatic timer, and brand name as characteristics (attributes) observable and valuable at the time of purchase. It is only after use, however, that one can judge goals (benefits) such as the quality of the coffee, or the accuracy of the timer. In addition, there may be outcomes not considered at the time of purchme which will have a bearing on satisfaction: ease of cleaning, or amount of counter space occupied. Each of the above issues has been raised although experimental investigation is sparse (e.g. Swan & Travich 1982). This dissertation will examine the issues in a limited way by looking at the possibility that characteristics (the more easily perceived aspects at pre-consumption times) and goals/benefits are different in evaluation and impact at various stages in the purchase and use process. Only two limited areas will be investigated: the existence of a separability in benefits and attributes, and then any differing impact on the constructs of expectation, post-purchase judgement, and satisfaction. The expectation paradigm contends that the confirmation or disconfirmation of expectations prior to a purchase is a major determinant of satisfaction. In its most recent form, this paradigm rests on the argument that expectations form ajudgmental base and that satisfaction is a consequence of the extent to which a product performs with respect to those pre-purchase beliefs (Oliver 1980, Day 1976, Engel and Blackwell 1982). Although this approach hm been well-received, and popular with researchers, a number of the basic theoretical and empirical issues are the subject of much debate. 4 One of these issues is the entire notion of context: from where do expectations arise, on what basis are they calculated, what are the types of expectations which consumers use. Although each of Me alone could constitute a stream of research. one important area of context is the source of expectations. One published study (Codotte, Woodruff, & Jenkins 1982) has explicitly examined the effect of prior experience of a product type (herein referred to as a product class) on the pre-purchase and post-purchase evaluations of specific product choices, integrated into a theoretical model. Day (1982) and Woodruff, Codotte, and Jenkins (1982) both refer to these prior experiences and have integrated them into conceptual models of consumer satisfaction] dissatisfaction as norms. Day sees these norms as related to actual performance and contribute [ing] to the direction and magnitude of ’ ' the confirmation/disconfirmation effect which in turn influences the nature and degree of feelings of satisfaction/ dissatisfaction. (Day 1982, p. 113) Woodruff, Codotte, & Jenkins, on the other hand, indicate that these norms show better predictive validity than expectatiom for satisfaction] dissatisfaction, but both correlate with satisfaction “probably because predictions will be correlated with normative standards“ (Woodruff, Codotte, and Jenkins, ,5. 119). Wright and Rip (1980), in a study unrelated to satisfaction] dissatisfaction, report that criterion features, norms, and attribute ratings for products are established at the product class, not brand, nor single item, level. Characteristics and goals are also related to context and product class. Three types of expectations appear in the empirical satisfaction literature: predictive, normative, and comparative. Predictive expectations are consumer beliefs about how a particular product is likely to perform (see Olson and Dover 1976, Swan and Trawich 1979). Sources of these beliefs have been ascribed to past ' experience, advertising. and word-of-mouth from others. Normative expectations are beliefs about what 5 ought to be, or a standard (see Granbois and Summers 19W). These may derive from all of life experience as well as more immediately relevant product exposure. Comparative expectations are beliefs about a particular product as compared to other, similar products (Latour and Peat 1980). Clearly, all three types rest on some attitude about or experience with the product class prior to the current purchase. Little research comparing the three types has been conducted, but the Woodruff et. al. study cited previously and research by Swan and Trawich (1979), both finding that product specific norms were better correlates of satisfaction than predictive expectations suggest that prior attitudes and experiences are determinates which need further investigation. Prakash and Lounsbury (1984) directly comparing all three types found that normative and comparative expectations were both better than predictive as determinates. The idea of normative standards, however, raises the issue of what norms (or product class beliefs) are composed. One might logically expect that in most cases (where any degree of choice exists) product class norms would affect the set of possible product alternatives considered, the attributes evaluated, and the relative acceptability of the product chosen. It is reasonable to assume that consumers buyaproducthopingtobenoworseoffthantheywould have beeniftheyhadchosen anotheritem ornotboughtanyitem intheproductclass. This willbetruenomatterwhatthe specific buying motive-novelty, curiosity, dissatisfaction with other offering, a search for improvement, etc. This assumption is consistent with economics and marketing. Norms may, as Woodruffet al, sugest be very important to the post-purchase evaluation where the consumer looks at benefits obtained, and presumably matches that benefit to previously obtained benefits as part of this judgement process. It is also to be expected, therefore, that product class norms would affect satisfaction which is a summary feeling about the product and the purchase. If a consumerlooksatbenefits obtainedandusesthosebenefitsasthe basisfornorms, thencharacteristics of the product which are observable at purchase time really serve as cues to effective satisfaction of 6 needs. Predictive expectations may be predictions about the utility of characteristics as guides to desired goals as well as predictions about the actual characteristics themselves. The post-purchase judgement is based more on the experience than the promise. Thus, the judgments are different in both time and information. Since benefits resulting from use or acquisition of a product are the usual purchase motivation, one can conclude that benefits promised and then delivered enter into purchase choices as a major decision factor, whereas tangible features of products serve as cues about the quality and quantity of benefits prior to actual use. For many types of products, the attributes may also be desired in themselves a well and so the quality of those features along with benefits obtained will be evaluated after purchge with some level of satisfaction or dissatisfaction with the purchase resulting. If this research shows some indication that this is so, then some of the problems raised by types of expectations may be reson by further explorations of context. The proliferation of terms and concepts with relatively little consistent definition and operation contributetotheunsettledstateofinquiry. Thisresearchisbasedonwhatappeartobethe commonalities of the proposed schemes supported by previous experimental treatments. A product, composed of tangible and intangible properties, is purchased primarily for the benefits it is supposed to confer. The evaluation of the properties and benefits issue in expectations which arise from several sources: experience and information of various sorts, and some degree of unique assessment. Experience and information with the generic product form a judgement base or norm. This norm may issue in beliefs about specific features, benefits to be obtained, and the context. This product norm is predicted to affect the specific product expectations, evaluation, and satisfaction. (Experience with a specific brand or product probably generated a brand norm as well, but that is not a consideration in this reseuch). Similarly, the expectation “types" proposals are, it is argued, a reflection of the 7 structure proposed here. Further discussion of the theoretical concerns and impact are contained in Chapter 11. Research into consumer complaint behavior discovered that as much as 20% of purchases are unsatisfying (Andreason and Reck 1977). Westbrook and Newman (1978) found that those satisfied with previous durable purchases were more likely to have positive expectations about new purchases, more likely to evaluate new purchases favorable and more likely to resolve complaints satisfactorily. Latour and Peat (1979) report similar finding. Given the widely reported research that word-of-mouth is the most influential source of consumer information, dissatisfaction, is significant for managers of consumer goods and service marketing. High levels of discontent can be expected to affect long term success though effects on repeat purchase and brand choice. Conversely, satisfied consumers are likely to have positive affects in the long-term and likely to generalize positive feeling across a range of consumption experiences. Consumer groups and governmental responses to organized dissatisfaction continue to affect business practice. Thus, better understanding of what determines satisfaction, how consumer evaluations occur, and likely responses to both satisfaction and dissatisfaction are important practically and academically. The dynamic process of marketing and consumption means that managers, with control over marketing activities for their products, would potentially benefit from a model which is oriented toward process and the links between managerial activities and consumer preferences. Consumers are exposed to a variety of stimuli which serve as preference cues. These cues are interpreted in light of previous attitudes and may be reevaluated as a result of new experiences. Improved understanding of how consumers conduct their evaluations of marketing stimuli and products should enable managers to 8 improve or design products more efficiently, determine message content more effectively and conceive of delivery systems in a more integrated manner. The two areas of investigation, to recapitulate, are the influence of product class norms on pre- and post-purchase evaluatiom of specific products and the possible import of a distinction between product characteristics and product goals on consumer satisfaction and future purchase intentions. As research questions in this report, they are formally stated as: QI Does an attitude about a product class partly determine consumer satisfaction by affecting expectations and evaluations of specific product choices? Q2 Are product characteristics (attributes) and goals (benefits) separate conceptions and, if so, do they have differing effects on other elements of the consumer satisfaction process? Chapter II will review theoretical and experimental literature. Chapter III presents the models and hypotheses. Chapter IV contaim information on the research design. Chapter V reports results of the investigation, while Chapter VI discusses the results and implicatiom. Chapter II LITERATURE REVIEW This chapter contains three sections: (1) a review of the theoretical foundations for the expectancy approach to consumer satisfaction; (2) a summary of major research related to expectations. and consumer satisfaction; and, (3) a discussion of issues in the methodology and modeling of consumer satisfaction. Each section summarizes the arguments and discusses the relevance to this research. W All the discussion and research into consumer satisfaction from an expectancy view rests heavily on theories in social psychology. As this area has developed in marketing, the theories used as the foundation have changed, but all are variants of the basic expectancy model of cognitive behavior where the distinctive characteristic of this class of models is their attempt to relate action to the perceived attractiveness or aversiveness of expected consequences. That a person does is seen to bear some relation to the expectations that the person holds and the subjective value of the consequences that might occur following the action. (Feather 1982, p. 1). There are many subclasses and variants contained within the above framework. The ones reported herein are only those which are relevant to the substantive issues in this area of marketing. All derive ultimately from the work of Gestalt psychologists (e.g. Kohler 1927), but developed in fullness from Levin (1951) and Tolman (1955). 10 ' te 'e The earliest marketing research into consumer satisfaction rested on theories hypothesizing that people behave (make choices) in ways that maximize the consistency of their cognitive systems. All depend on Heider’s (1946, 1958) theory of balance: an imbalanced set of cognitions is associated with tension and the person will be subject to attempts to attain or regain balance. The emphasis was placed on a person’s perception of the relationships among himself, another, and an event or object. Balance exists whenever all relationships are positive or if two are negative and one is positive. While Heider’s approach has not been directly incorporated into the applied areas of marketing, several important aspects of his perspective reflected in others’ variants are significant: valence (positive or negative) of beliefs, the ideas of association between an object and some related object, and the role of both belief and feeling in attitude. Belief valences reflect the notion that external objects (or behaviors) have multiple properties (or consequences), some of which are desirable to the perceptor and some of which are not. The source of the desirability valence is the set of values or beliefs of the person which are used associatively to evaluate the object. The multiple factors, in turn, mean that the relative strengths of the beliefs to each other are important in determining the overall evaluation of the object as good or bad. Association also extends beyond the person and the object to some set of goods via objects and behaviors, producing affective responses. One of the earliest developments from Heider was used extensively in all areas of social science including marketing. 11 W. Most often associated with Festinger (1957), dissonance looks at elements and their relationships. The essential claim is that dissonance is a state occuring whenever a person holds two cognitions (ideas, beliefs) which are inconsistent or incompatible. The theory assumes that dissonance is unpleasant, so people strive to reduce dissonance. Aronson (1968) argues that essentially dissonance arises from violations of self- concept: predictions and consequent behavior violate an idea of the self. The first explicit research into consumer satisfaction followed directly on these ideas (Cordoao 1965). Calder (1973) suggests that, while dissonance as a theory lacks sufficient specificity and breadth to be completely satisfactory, it does explain observed consumer behaviors and rating well. Supporting this contention are studies reporting that committment to a product (purchase or choice) results in a strong tendency to regard the product either much more highly or significantly worse than comumers who had not made such prior committments (Cohen and Goldberg 1970; Doobs et. al. 1969). The prediction in dissonance is, of course, that perceptions will be re-evaluated until balance occurs, and that the greater the committment (selfcinvolvement), the greater the strength of the effect. Objections to dissonance as an explanatory and predictive approach include the lack of specificity in predicting either the magnitude or direction of adjustments, lack of situational context, and the absence of falsification conditions. Thus, the observation that high, confirmed expectations resulted in satisfaction fit the paradigm, but results were otherwise difficult to explain within the context of dissonance: very high expectations produced satisfaction, but at lower than predicted levels (a non- linear result). Second, there was no prediction about the direction of the resolution with conflicting belief confirmations. Finally, disconfirmed expectations did not invariably lead to satisfaction or dissatisfaction, but often rather neutral responses. Comparable results in other fields led to the expansion or adjustment of this approach. One of these, assimilation/contract, formed the basis for a second set of investigations into consumer satisfaction Dissimilation (Howland, Harvey and Sherif 1952; Sherif, Sherif, and Nebergall 1965) phenomena have been noted and explained as adaptation All these theories (Helson 1964, Thibaut and Kelley 1959) differ from the simpler Festinger model in that 12 the dissonance model is based on the notion that people seek to eliminate the noxious stimulation resulting from inconsistency. This notion [contract- assimilation] implies processes which are recurring...[and] presumably gratifying. (Upshaw 1968, p. 213) The basic claim is that persons have a reference scale Which contains a point of indifference. All stimuli are evaluated with respect to this pre-existing scale. In the case of discrepant information, either the scale or the information must be adjusted. Contrast claims that information within the indifference range will assimilate while, outside that latitude, the disparity between elements will be magnified. The mixed research results, unhappiness with the level of explanation, and the emergence of more appealing variants have led to a general abandonment of both dissonance and assimilation—contrast as bases for the observed effects on expectations on satisfaction (See, in particular, discussions in Anderson 1973, Olshevsky and Miller 197?, and Oliver 1980.) Attitude The period marked by dissonance type interpretations of cognitive behavior also saw the development of alternative approaches to motivations which were more grounded in supplying explanations and which were more specified. All the major variants attempted to connect preexisting beliefs, attitudes, and experiences to the judgement and evaluation of new situations in ways which allowed for the introduction of situation-specific and individual factors. Him Helson’s adaptation level, strongly championed by Oliver (1980, 1981), states that stimuli are perceived only in relation to existing standards. Unlike the previously mentioned theories, the emphasis here is on perception, and the source of the standard, as well as the effect of perceptions on subsequent judgements and attitudes. The theory predicts that only stimuli which differ significantly 13 from what was expected will yield any but a neutral result. Based on studies of physical phenomena, Helson has been used by a number of researchers as a basis for refinements in expectancy models (Kahnevan and Tversky 1979, Cofer and Appley 1964), but the weaknesses are the continued lack of emphasis on the context in which expectations develop and the outcome of judgements on future events. The first difficulty is that the source level radically affects the outcomes. The purely physical perceptual phenomena studied which formed the original basis for the theory, tended to limit further consideration of the source. Thus, although the standard is defined as a function of perceptions of the stimulus itself, the context, and psychological and physiological characteristics of the organism (Oliver 1981), there is no consideration of the possibilities for changing the standard. More significantly, the theory rests on the assumptions of operant conditioning which, to a great extent, presuppose fairly stable, continuous conditions. The second major shortcoming is the short-shrift given to the role of affect (see Feather 1982). Cofer and Appley (1964) see the discrepancy between expectations and events as the antecedent condition for the development of affect, however, this rather beg the question of the role affect plays in expectations, does not specify how affect relates to the persistence (or lack thereof) of adaptation level, and generally ignores the potential of affect as an influence on belief strength and direction Given the concept of affect as central to satisfaction as a complex emotional response, this is a problem. It also makes it somewhat difficult to link satisfaction research to other areas of consumer behavior when affect has been posited as a major influence, e.g., impulse goods shopping, advertising appeals. It should be noted that Oliver, the major proponent of Helson’s adaptation level approach in consumer satisfaction research, has tried to demonstrate the utility of this theory by redefining satisfaction as the evaluation of the degree of ”surprise inherent in a product acquisition and/or consumption experience and attitude as an affective response (Oliver 1981). As he proceeds to claim that satisfaction has an impact on attitude, the connection with affect is still indeterminate in this context. 14 Two other researchers, Rosenberg and Fishbein, have proposed theories more formally concerned with attitude and affect. mtg. Rosenberg proposed a theory of attitude where attitudes are defined as positive or negative feeling toward an object with affective significance. Persom want to maintain a balance between affect and belief as the importance of an object is instrumental with respect to values. While not used directly in satisfaction research, one prOposition (Rosenberg and Abelson 1960) directly concerns the limitations of the early consistency based approaches: imbalance is predicted to lead to change in affect or belief mm the person thinks about the imbalance. Thus, if an object, or beliefs associated with it, is not sufficiently important, adjustment and change will not occur. If this be true, then disconfirmation of expectations, sufficient to affect satisfaction in Oliver’s terms, would occur only When the object or the most salient of beliefs wereviewed as especially irnportant. In the context of expectation/satisfaction generally, it calls affect into play as well as beliefs. A second proposition, which might be of some importance, is that adjustments follow the path of least resistance - a variant of the maximum principle. If true, this suggests that goal primacy might be the framework from which to predict the least and most likely paths of adjustment, as well as the likelihood of disconfirmation effects on satisfaction Rosenberg's concept has not been used with any frequency in consumer research largely due to the emphasis on the primacy of core values as the base. Most marketing situations are seen as fairly removed from these values and close alternatives without such restricting conditions have been available. Lutz (1981) notes that Rosenberg is potentially useful in examining highly involving conceptual situations or for looking at behaviors and attitudes at the product class level. The other major difference which links Rosenberg and Fishbein in opposition to what has preceded is the detailed attention both give to the concept of attitude. In fact, Rosenberg attempts to show that object evaluation is related to the expectation (set of beliefs) that the object will facilitate (or 15 not) attaining a goal and that the evaluation thus has both cognitive and affective elements. His model is a specification of the relationship of affect and belief with respect to the object. The components and their relationships are: Attitude . F ( Vi P|) Where attitude - favorability or unfavorability toward an attitude object VI - value important of the ith value Pi - perceived instrumentality of the attitude object with respect to the ith value it - number of value 16 55mm. By contrast, Fishbein, whose major interest is the link between attitude and behavior, devised a similar expectancy model Whose original form is: Attitude - F (bI el) Where Attitude - affect for or against the attitude object b| - strength of the belief that the object possess the ith attribute el - evaluative aspect of the ith attribute n - number of salient attributes As Fishbein describes this model, it is a way of determining an attitude about an object or behavior based on the attributes salient to the particular object and is not composed of a set of evaluative attributes (Fishbein and Azjen 1980). To that end, then, the more traditional uses of expectancy formulations in marketing research are criticized as often inappropriate and seldom validated. Thus, looking for consumer preference or satisfaction without using both a saliency measure and a behavioral tendency is seen as inadequate as there may be two attitudes underlying consumption: attitudes toward the product and attitudes toward buying the product. For this reason, a normative component wm added to the above model. However, that addition is, in a sense, a move closer toward Rosenberg’s position in that the purchase choice is now seen as somewhat instrumental. In Fishbein’s own examples, attitudes derived from salient beliefs about the product and the situation plus beliefs about the social factors of the purchase together are good predictors of buying intentions, which, in turn, are good 17 predictors of purchase. Consumer research has used a Fishbein approach both directly and in severely modified form (e.g. Wilkie and Pessemier 1973). The multi-attribute choice models are ngt, however, studies of satisfaction in consumption. Thus, those satisfaction studies which have loosely referred to a Fishbein approach or adopted a multi—attribute measure for expectations have not really examined the implications of the intentions underlying the model. What satisfaction research is intended to accomplish is an understanding of what occurs at purchase choice and afterwards. It is important, then, to look at the gig]; beliefs for the product chosen and the instrumental features those beliefs may represent in order to understand the consumption evaluation. Two examples in Fishbein and Azjen (1980) may clarify this. One example was the British study of detergents where it was found that one brand was purchased where it was believed good for woolens and another not purchased where it was believed to cause scum. Attributes of detergents in general elicited from informants did not include these characteristics. Another way of looking at this is to observe that a desired (satisfactory) detergent must have the relevant properties of detergents generally plus one additional instrumental benefit (good for woolens) and no serious negative effect not present in the whole product class. A first purchase of any given detergent would depend on beliefs about detergents plus some specific quality while satisfaction with the product after use would depend on confirmation of the good reasons plus some possible additional benefit. Dissatisfaction would result from inadequate performance on the general qualities or some unlocked for negative quality. This is essentially similar to Herzberg’s two factor model of satisfaction. The second example of toothpaste/ car purchasing introduces situational and normative components, where the purchase choice varies depending on the perceived consequences as well as the attributes of the product in itself. This goal/instrumental focus, subsumed into subject norms, ignores the potential relationship between the attributes and the situation which may affect both purchase Choice and satisfaction. Thus, floride may be important in a toothpaste only if it is purchased for a 18 child’s use. Floride would be a salient belief and the child's use a subjective norm, but the approach omits the relationship. m. Vroom (1964) examining motivation and job satisfaction faced essentially the same set of problems, and, in doing so, prOposed two different expectancy models for the prediction of job satisfaction. The first formulation treats activities as instrumentalities to the achievement of desired benefits and is operationally defined as: V - f‘ (Vkikl) k- 1 Where Vl :- valence of outcome j Ijk - cognized instrumentality of j for attainment of k Vk . valence of outcome k n . number of outcomes 19 The second, still emphmizing desired benefits, looks at the probability about effort required to obtain benefits desires: F, - " (E'le) i-l Fl - force on individual E" - strength of belief that i leads to j V’ . valence ofj In both forms, the goal is a choice of actions and the models are designed to predict the likelihood of choosing some end. Later researchers extended the inquiry into satisfaction with the results of the choice. For many of the researchers in this area, however, the relationship between expectations and satisfaction is moderated by two factors: goal setting (Carroll and Tosi 1973, Kim and Homans 1976) and perceived equity of outcomes (Ilgen 1971, Ilgen and Hamstra 1972, Hamner and Harnett 1974). This whole line of research finds, therefore, that satisfaction is dependent on t_wg comparisons: actual performance to expected, and, comparison of actual performance to a referent person’s performance. Results of experiments along these lines have produced evidence for both pure expectancy m goals and expectancy plus equity. This essentially combines both models proposed by Vroom. These efforts derive more directly from proposals about expectancy from motivation theorists like Rotter and Atkinson. (See Figure 2.1) They are indirectly related to the consumer issue at hand, 20 but are important for two reasons. First, the expanded framework is similar to Fishbein’s in that motivation is introduced. Second, all the more recent work on job satisfaction and motivation emphasizes the continuous nature of expectancy processes; expectancies arise from a background of experience, information and observation Equity. Equity theories, which have become increasingly important in job satisfaction research, have been examined in only a limited fashion in the consumption literature. Two forms of equity theory, Adams (1961, 1965) and Thibaut and Kelley (1959), have been used as a basis for empirical research in consumer satisfaction With roots in general dissonance theory, equity can have two comparison bases: equity as compared with a significant other and/or equity as compared to some more general sense of equality in exchange. Both approaches specify that greater satisfaction occurs when there is greater perceived equity and vice versa. To date, results of limited testing show that equity approaches provide lower explanatory power than disconfirmation paradigms (Swan 1982, Evans 1982). However, this may result from two factors. First, experimental testing has not included examination of real purchase events and significant others, but has relied on simulated experience in laboratory settings. Second, the Thibaut and Kelley framework, designed to organize data about interpersonal relationships, contributes little to specify the type and influence of cost/benefit analysis consumers may apply to purchases. Cordom’s (1965) finding that shopping effort significantly affected satisfaction supports the idea of some cost/benefit consideration 21 Determinants of Theorist Impulse to Action Subject Tolman (1932) Expectancy of goal, Maze behavior demand for goal Lewin et.al (1944) Potency x valence Level of aspiration, decision making Rotter (1954) Expectancy, reinforcement value Social learning Edwards (1961) Subjective probability x subjective utility Economic decisions Atkinson (1964) Expectancy x (motive x incentive) Achievement-oriented behavior Vroom (1964) Expectancy x (value x instrumentality) Organizational behavior Dulany (1968) Hypothesis of the distribution of Verbal learning reinforcement x value of the reinforcer SOURCE: Lutz, Richard J. Contemporary Perspectives in Consumer Research, Kent publishing Co., 1981, p. 246. Figure 2.1 The success of equity approaches in the job satisfaction literature and the intuitive appeal of a framework which includes some form of consumption attitude do support further work to refine and adapt a notion of equity. Perhaps, as in job satisfaction research, it will enter 8 an additional set of factors which partially explain levels of satisfaction, particularly in longitudinal studies. mm A number of expectancy formulations attempt to explain and predict motivation effects on outcomes. Atkinson (1978, 1982) has examined risk taking and action In the latest formulation of his expectancy model, he sees action as a consequence of evaluations of both instigating and inhibiting tendencies. Some of the conclusions he draws from this stream of research which have a bearing on consumer satisfaction include: (1) preference may be greater for less ideally satisfactory activities [products] if the negative risks are greater for the potentially more desirable outcome; (2) time is a significant factor as individuals more conscious of risk will be initially more resistant to risky choices but moderate that anxiety over time; (3) individual characteristies have significant effects on choices and evaluations. Feather (1982) in summarizing 20 years of research on actions as related to expectations, he found that difficulty and risk significantly affect the evaluation of the goal [benefits] itself. Additionally, it has been found that past experience (or reports of others’ past experience) is a significant mediating link between expectation and personal aspiration - e.g. a systematic tendency to over-estimate success with low reported probabilities and vice versa. A second area of interest in the importance of a goal: the more highly valued the goal, the more the effect of success or failure impacts affect with respect to the experience or goal. Related to this is his proposition that goals and expectancies have affect (valences) which are determined by values; the more closely a goal is related to some terminal value, 23 the more impact a set of instrumental values becomes. The implications of this for consumption is the importance of considering the benefits to be derived as the set of properties which are evaluated. As a corollary, if satisfaction is directly related to affect, then reconsideration of values as demonstrated through purchases and consumption is called. for. The intermingling of tangible properties (attributes) and benefits derived from consumption of the product needs further investigation Rotter, whose interest is social learning, notes that the assumption, supported by a considerable body of research is that: Expectancies in each situation are determined not only by specific experiences in that situation, but also, to some varying extent, by experiences in other situations that the individual perceives as similar. (Rotter 1982) He follows by noting that the amount of prior similar experience has a significant effect on the importance of specific expectancies versus a more generalized set. In consumer contexts, that means that the newer the product type is to a consumer, the more general consumption experience will affect the type of specific expectancies the person will hold with respect to that product. With greater experience, expectations will be more product specific. (Although I know of no research in this area, it does suggest that both persons and products need to be carefully controlled in satisfaction and brand choice experiments). Rotter also emphasizes the importance of personal and situational variables. Despite the extensive work in psychology which has generated such diverse ways of dealing with beliefs and expectancies, the borrowing into consumer satisfaction research has been very general. Real consideration of the theoretical sources for empirical work has been quite limited (See Hunt 1982). Much of the difficulty, of course, lies in the difference in the range of behavior and the explanatory motivation considered. Nevertheless, the insights from social psychology can be used in developing a better basis for consumer behavior research and explanation Despite the many differences in the approaches briefly discussed, there are some similarities all share which seem well supported by empirical research. First, expectations about objects, behaviors, or outcomes are derived from information and/ or experience. The effect of expectations are moderated by the characteristies of the individuals forming these beliefs: values, motivation, self confidence. The difference between expectations and actual outcomes does affect the overall evaluation a person gives to the object, behavior, or outcome, although the degree of affect does seem dependent on the extent to which the moderating factors play a part. Moderating factors affect summary outcome judgments depending on the degree to which outcomes are attributed (to oneself or to some outside actor), by the relative importance of goal or instrumental values, and by the consistency/ consensus basis of the expectations. (Consistency is experience derived and consensus is indirectly derived via others) (Wiener 1980, Abramson, Seligrnan and Teasdale 1978, Feather 1982). It has been noted also that the degree to which an outcome is perceived as related to self esteem magnifies the outcome evaluation Thus, one would expect contingent theories of consumer satisfaction to emerge which are parallel to the contingent approaches to consumer choice, e.g. low/high involvement. The base for this type of approach would specify a limited number of components with specified relationships. Those relationships would, however, be modified by contingent factors. To date, marketing research has indicated that the type of good (Churchill and Suprenant 1982), the existence of prior consumption experience (Westbrook and Newman 1968), and the ease/difficulty of shopping (Cardozo 1965) all may be significant modifying elements of a simpler expectation outcome satisfaction/disatisfaction model. There are, also, strong indications in the disputes over disconfirmation as an element of satisfaction models that expectations may modify post-usage evaluations in a contingent manner: contingent on the type of expectation (this is discussed in more detail in a later section of this chapter). 25 Three models have been presented heretofore in the marketing literature. Each derives differently from the social psychology framework so far presented. W The first formalized model is, also not surprisingly, the one most often used as a basis for research and discussion by marketing researchers. Oliver (1980) developed his model on the basis of Helson’s Adaptation Level Theory and the results of his own empirical work (see Figure 2.2). As he presents it, a person forms expectations about a product from summed evaluations of individual beliefs about attributes. This evaluation prior to use results in the formation of an attitude toward the act of purchasing the object which issues in the intention to purchase or to not purchase. After usage, disconfirmation occurs: an evaluation of the degree to which perceived performance deviates from what was expected. The outcome of this comparison is an affective judgment about the product which is a judgment of satisfaction/dissatisfaction In Oliver’s terms, this is: an evaluation of the surprise inherent in a product acquisition and/or consumption experience. In essence, it is the summary psychological state resulting when the emotion surrounding disconfirmed expectations is coupled with the consumer’s prior feelings about the consumption experience. (1981, p. 27) This emotion is then posited to form part of the attitude toward the product and purchase after consumption with an effect on intention to rebuy. Over time, satisfaction is expected to decay and general expectation for future purchases returned to a homeostatic level unless disconfirmation was so strong that attitude change has occurred. 26 DISCONFIRMATION SATISFACTION ATTITUDE ATTITUDE INTENTION INTENTION TIME DISOONFIRIMTION “"52 1 PEnIoo EXPECTATION KEY: Expectation-Z blot Dbconfirmatton- degree to which actual performance deviates from expected be belief e =- evaluation Souce: Oliver (1980) FIGURE 2.2 Oliver'sModel 27 PRIOR EXPERIENCE WITH PRODUCT PRODUCT Rm cuss TYPE B D v , V PEIu=ORIrANcE B D PERFORMANCE NORus ‘_ PREDICTIONS PERCENED BRAND COMPARISON OF BRAND PERFORMANCE —" WITH NORM I SATISFACTION! BELIEF ABOUT CONFIRMATION] DISCONFIRMATION I SATISFACTION FEELINGI I v I SATISFACTION OUTCOMESI Source: Codotte, Woodruff, Jenkins (1002) FIGURE 2.3 28 PRIOR PRODUCT [BRAND ’ EXPERIENCE I I If PERFORMANCE BRAND BRAND NORMS ATTITUDES EXPECTATIONS I I L. PERCEIVED BRAND PERFORMANCE 1 CONFIRMATION OR DISCONFIRMATION PN NEGA‘IWE POSITIVE I FEEUNGS "W'- flEEUNGS I I V I I ' 1 ! OUTCOMES I FIGURE 2.4 mmwmnm H13: 29 Selection of a standard for evaluating brand performance depends on the consumer’s perception of the relevant use situation. Selection of norms as standards for evaluating performance is more likely when the relevant product set has several brands than when the set has one or very few brands. The position of the reference norm along the performance dimension varies directly with the average performance of the brands which are evoked by the use occasion. The more similar the performance across brands, the more likely a product- based norm will be used. Unless the consumer has extensive experience with the focal brand, experience—based norms serve as a better basis of comparison than focal brand expectations. If the consumer has had extensive experience with the focal brand, the brand norm is equivalent to focal brand expectations. If the consumer has limited experience with the focal brand but extersive experience with a second brand, the norm for the second brand serves as the basis of the comparison. If the consumer has experience with an assortment of brands and no one brand dominates, the norm for the product category serves as the basis of comparison for the focal brand. The perception of focal brand performance is influenced directly by prior focal brand expectations and attitudes. Only perceived performance outside the zone of indifference elicits positive or negative feelings about the consumption process. The initiation of unusual satisfaction outcomes (i.e., recommendations to friends, letter writing, seeking out the manager, legal action, etc.) is associated only with brand performances outside the zone of indifference. The frequency of performance disconfirmation varies inversely with the width of the zone of indifference. The width of the zone of indifference varies a. inversely with the breadth and depth of a consumer’s experience with a particular brand. b. inversely with the degree of personal and/or situational involvement with a brand, and c. directly with the variability in performance of the brands in the consumer’s experience set. Source: Cadotte, Woodruff, Jenkins (1983) FIGURE 2.5 Hypotheses for Experience-Based Norms This model depends on the existence of three factors in the purchase situation: (1) a distinct difference between expected and perceived performance in order to produce either satisfaction or dissatisfaction and (2) stability of relationship between beliefs and purchase attitude. (3) that the question of interest in this line of research is the effect of disconfirmation on future purchase intentions. The third factor is the most fundamental and its significance will be discussed in the measurement section of this chapter. Research based on this model (and similar, reduced form, variants which generally exclude the attitude constructs) has centered essentially on three themes. The first, of course, is the overall explanatory fit of the model to obtained data. The disconfirmation construct, as a construct, has been the most problematic of the proposed portions of the structure. Essentially, the issue has two parts: the demomtrable existence of disconfirmation as a construct, and, the other on problems of measuring disconfirmation (The relevant research is discussed in the next section of this chapter.) The second issue centers on the nature of expectations and their measurement. The most common approach has been the conceptualizing of expectations as beliefs about attributes or performance. However, expectations have been treated as product specific attitudinal scales (that is, with beliefs weighted by some evaluative measure) and as summary judgments prior to use. Sources of beliefs have been scrutinized more recently as the Cadotte-Woodruff-Jenkins model discussed next 31 indicates. Equity and cost/benefit expectancies have also been proposed although research support for these types of expectancies has been weak. It also has been proposed (e.g. Miller 1977) that expectations should not be conceived m a unitary set of beliefs, but rather as a multiple set of constructs antecedent to purchase and post-purchase judgments. The Miller proposal, supported by limited experimental work, parallels recent proposals in social psychology (Heckhausen 1977, Bandura 1977, Feather 1982). Such multidimensional sets imply, in turn, more complex post-purchase judgments. The third issue is the relative importance of prepurchase and post purchase effects on satisfaction The arguments center on the relative importance of post-purchase evaluation versus disconfirmed initial expectancies, the role of previous experience and product norms, and the change or modification of the belief set between pre- and post-purchase judgments. F'mally, the proposed model, as an expectancy model, shows the limitations noted for all models of that type. Among the more serious concerns is the issue of falsification Many critics note that manipulation of measurement models can significantly alter experimental results, but as there is no theoretical standard specified within expectancy theory, falsification conditions have not been established. A second major criticism, particularly applicable to the research in marketing areas, is that behaviors and motivations exist in continuing streams while most expectancy approaches look at events as discrete, thereby obscuring the possible importance or lack thereof of insights from expectancy theories. A third criticism concerns the distiriction between object, action, and outcome expectancies--it has been a frequently blurred distinction although there are some potentially important implications, especially in the area of consumption For example, tradeoffs among product, opportunity, and benefits may be central to understanding the satisfaction, intention and future behavior links so important managerially as well as academically. 'fifirfi E3 COHJ aver are i With 35a anUn Rab Stung 8km“ 32 A second, but distinctly different expectancy model has been proposed by Cadotte, Woodruff, and Jenkins (see Figure 2.3). Incorporating suggestive information from experimental research with acknowledgement of some of the criticisms delineated previously, they have proposed a model which specially incorporates performance norms and experience into the antecedents of satisfaction Drawing on concepts from Morris (1976) about normative deficit, Swan and Mercer (1981) on social equity, and from Thibaut and Kelley (1959) on comparison levels (with related research [Latour and Peat 1977, Swan and Martin 1981]), they posit that a normative comparison level exists, just as Oliver does, but that its source is not specific product expectations, but rather a standard derived from previous experience or information about the product class, shopping experience or other relevant variables. The perceived performance of a particular brand, then, is evaluated relative to those norms as well as the predicted brand specific performance. The outcome of disconfirmation is simIiarly posited to be complexly dependent on the intensity and direction of disconfirmation The norm can be understood as several different comparison standards: brand based, product average based, or best brand/favorite brand based. The number and range of items in the norm set are individually determined. The number and use of norms in confirmation/disconfirmation may vary with involvement, risk and investment associated 'with the purchase. The general pattern is expressed a a median/mean norm in a bell shaped distribution across brands and in a summary judgment rather than attribute form. The presentation of the model is accompanied by a set of hypothesis and measurement specifications. (See Figure 2.5 - The measurement issues are discussed later). The hypotheses indicate the contingent dimensions of this proposed formulation which takes into account a number of other consumption behavior concepts, and thereby strengthens the applicability of research. It also recognizes research findings indicating that product types (durable, service, etc.) may have a strong impact on information processing which suggests limits on the utility and reliability of a more general model like Oliver’s. 33 Other than the limitations pertaining to all expectancy models, the utility and informativeness of the model appear promising. To date, only one limited research test based on this model has been conducted. Not specifically accounted for are the questions about the nature of expectations and the readjustments of criteria throughout the process. The implication of norms is that all criteria will be included in past purchase judgments; whether a difference in criteria affects satisfaction is still unexplored. The third major paradigm proposed is Day’s (1977, 1982). As shown in Figure 2.6, this model is more inclusive than either of the previous two. Similar to Cadotte, et.al., it posits a set of expectancies (brand and product norms), but there are three categories of antecedent norms and expectancies: performance, social benefits, and total cost all of which enter into cost consumption evaluation and issue as multiple responses. There are explicit conditions specified, including the definition of satisfaction as an emotional response, the differentiation of norms from expectancies (predicted product performance), all measures are summary judgements, and the model is applicable to one specific event. The emphasis on social and economic benefits is a natural consequence of the intent 'of the model: to provide a method for explaining and predicting post-satisfaction behaviors, particularly ”complaining behavior". The problems lie first in the antecedents to satisfaction Although much discussion in marketing has centered on equity and cost/benefit, researchers attempting to operationalize and test for it in satisfaction paradigms have had little success (see Swan 1982). Given the unsettled nature of agreement on expectations and product norms, this model is unlikely to assist research in the near future. 35 A second problem lies in the strong position on measurement. Irrespective of other considerations, the summary nature of the measurement lessens the potential for uncovering specific features which may causally link or correlate with other elements. As a diagnostic process, it is weak. A third limitation is the specified intent to look at single events only. While this does maintain focus on consumption versus life satisfaction, it also ignores the longitudinal aspects of many consumer situations, e.g., health care, and is highly time-of-occurrance bound . The positive features of all three models are several. First, all represent an improved maturity in the field; earlier research in this area was neither systematic nor programmatic. Second, attention to the context of decisions, in time, experience, and marketing efforts has been progressively added. Third, the development of models addressing the issue of measurement specification allows for better and more rigorous research. The issues raised by the models (and their presentation) are many, but two theoretical questions are emerging as central. The first is the nature of the effect of previous experience (or surrogates for it) on specific product expectancies, post-purchase judgements, or satisfaction It was prOposed by two of the theorists that previous experience enters into the satisfaction process as a norm which establishes the center of the “zone of indifference” for post-purchase judgements and disconfirmation with effects on satisfaction Cadotte, et.al., also indicate that brand performance may affect product norms (Figure 2.3) or that performance norms may affect brand attitudes (Figure 2.4). The effect of norms on specific brand (product) choices is an open question both theoretically and empirically. 36 A second consideration is the nature of judgements both pre- and post-purchase. As previously mentioned, the set of evaluative items may differ in content and focus at different points in the process. The tendency toward summary disconfirmation measures and the varying approaches toward defining and measuring expectancies have resulted in few guides to the nature of product norms: do they include summary judgements, judgements across all salient attributes, and in what ways do they affect the attributes for the choice object? Based on available indicators, previous experience may provide a ground for baseline performance and may determine the attributes salient for the single purchase. An alternative view, which is implicit in the Cadotte (Figure 2.4) model is that product norms are affective in influence on a chosen brand. An unresolved and undiscussed issue is the nature of performance expectations. The multi- attribute formulation of expectancies typically may include features which are desirable because their presence partially determines the benefits conferred by the product: e.g., softness in pajamas because it aids sleep, prevents unpleasant feelings, etc.. Which sort of performance is to be the focus of concern? It is possible that both sorts of conditions, attributes, and benefits need to be evaluated as satisfaction and disconfirmation may depend on evaluations of both types of beliefs. The same may be true of product norms. Attributes may be desired (or baseline levels set) since previous experience has demonstrated their mediating utility. All of the models accept some form of adaptation theory while including elements from other social psychology formulations: values, equity, motivation What is still missing is a consideration of the peculiar features in the domain of marketing and the essential questions which the various proposals aredesignedto address. These are broad issues and the subtleties are reflected in the major differences of each of the proposals. This thesis is designed to examine only a few of the issues in a highly limited way. First, 37 several alternative models incorporating product norms will be examined. Each of the variants looks at product norm effects on each of the other constructs which the models have in common: specific product expectations, post-purchase evaluation, and satisfaction Second, the nature of expectations will be explored in order to see if indicators (or observable performance features) and derived benefits are separate or interdependent expectancies. A third intent is to examine the overall causal structure: expectancies -> post-purchase judgements -> satisfaction -> intention The importance of this third objective will be highlighted in the next section. KW There are several important features of previous research which have a bearing on this examination This summary, not intended as an exhaustive, but rather a selective survey, of research to date will consider three areas where current experimentation bears on the issues of this research. These are: concept development and testing, product and marketing variables, and measurement issues. Cam W. The first experimental research was derived from the long tradition of expectancy and the varying dissonance theories prominent in the 1960’s (see Figure 28 and 2.9 for summary of major experimental research). A specifically marketing oriented tradition has emerged over the past 20 years although there seem to be more disagreements than agreements as to the paradigm, research methods, and limits. Swan (1982) has summarized a consensus view of the field in the two charts shown in Figure 2.8 and Figure 2.9 ' ..c.:.fs\.l.l.:.-J.... 3.....dl-Iqundi51 [Mam\ u...e::- :Aflmfers ..:....~.:.. s s. a» .e... ..s .. o 5.3.. 1.15.? 5,54: _. c.7131: 38 mco_uauueaxo nuance ouau_uue + oocmEcoeeon ~.~ wa:o_n BEBE on poem.oc .em_ummm_u new uaonm mcomuouoenxo ee_nc_ u.eeen nec_moc Co aces co>oeoe ACACF. xo_3oc» Occmmoc + e>_uu_p0cn -_.m»..3e on co~e.oE ..m_umm uoOm mccnaocm eta cmzm cecammoe ommgoean new OEOuuao oucoc_ecoo .eocec_ecoo .co_uauuonxo use ecoum modcum eo_.on anon + and Ow neum.ec >.o>_u_won .wmmuam .unoo memnnozm Aumopv cmzm .cozuo uoc e.aEem oco no. .cm_m-cox_E muueeea .Oxo.oEO co_uc0uc_ ~m_m>.mCm op co_uumem_uem co_umEe_~ cued. .CCOOm_O >emEESm -ccum_O Ou mcomucuuonxo cusp muceuaum .»e_u u_uanomv noumdme >.zm_z ocoE on:u_uuo co_u + >u_o .o>e. .nep< o.eum .uue masseuse .ecoom_c + mco_uouoenxo we -c.:oocc_ we .non coeds: anon» mee_.on ma .axo co_uuc:e cmvc_. m_ »m_uom ade Codnsem Anuopv ec>_.o .umm we .o>o. “Ocean om_m mco_uouoonxm .ecoum_c .mw_u + >ccEE3m .mco_uceuc_ uce unseen ccExO_ca mecamacs Ocamonxo-umon Ou noua.oc o>_ep amen + co>__o .nxo ou:n_cuue + >LmEEJm >.mc0cum Hume‘s .ecOum_O o._nOEou3e ucmpaum “Asa-v cm>_.o co_ucoOc_ + op:u_uue.umon + «an mo_ucmuooaxo o» pounce cannons nuancoee_n .E_mme .umm x mucoecdo eo_.on “CROPS co_ua._e_mmc acec_ >9 team.3n_cms .nxo on On c30cm mo_ocmuoenxe oeeeou mm>_:om:cc ew>oo + cem.o oucecomm_o .ecod >.co anon .econ oucadcn :daoc_: Ca Ca Occogoce .nauom we mmodccemec mco_u mcoccoOOE Ammo—v no..mz >u_:cmcou .weon + axe .n_cmE .muuenxo on Oeum._E_mmm »m_umm one“ mucouaum + >xm>ozm.o oocccomm_u umOu unamonxe.oen em_ummm_0\wm_ucm no» acme: can “ooucooxsmmmo Co»c_:n_cee aco_uouuonxo >u_cmnmmu we .o>o. Odonmoccu muc_on..mn mucouaum Anna—V comcouc< .comueuuonxu gum: noncsoecou meeameoe-0cn yucc_u cc co_uue.m_uom cecm_c x uneven and: o» x .em_uem cauoocm “.o>o. co_ueuoonxo >e~EE§m oumcn + o»c_ >5 .e_u_c_ can uneven mc_ndo:m .cemm_n .uoo couo.:n_an - co_ueuuoaxo >2 nanomeem mm co_uumem_uom muc_oa._cn mucocaum Anne—v Onopceu docoxuxcoo;h mcomu_c_»oo\yco66c3mmox mmc.nc_n acceded muuoqeaw mcocuceomoa voacmucou ~.~ wane—n mcogao we ouco_eenxe no 39 co_umEc0ec_ uoc.co,uumem_umm cu cocoo.u unaccouc_ Aoworv ouco_eonxo .>u_:oo mco_uauuonxo mo mon>u m cmue.oe eucoweonxo ce_cn >.co o._u _.mE anon ecu enouwn mco_uauuoaxe mco_ueuuonxo O>_uo_p0cn can» mon>u pooco_cenxe comuumemmuem Co ecuu_coen Arno—v co_uouuonxo co.._z ace o>_uu_cocn comc0cuu we: ooce_cmnxo «men e._OOEOu:e mconnozm cmucex use cmzm omen: Oeum.:E_m Oeuoonxo me: yes: EOE» ww_u we deemed no» n_cmo >9 nauseous com» -aEL_»coum_p m_m>.oc< cauoae came m_m>.mcm umOmCOLum use comumec_ecoommn echwaodoE sumccnem cu_3 co_uumem_umm meomueuoonxo co_uooe unseen >Euceu_»_cm_m co_ue mmEc_ea cednozm good—v cow.o= -m_ucm new eu:n_euuc -Ec_*c0um_c new mco_uouuonxe m.ceE ..eE enema new cu>_.o .mco_uouuonxo O» >.co pounds; co_uumem_ucm emo3u00e mudcum “accuomw o>_uuo*em co_u:n_cuue seam o.n_u.:E m we emcee new: cu Ocua.mc coozu00» mcomuouuonxo + noes co_mmocmoa o.nmu.ae co_uooemmuom mo._nOE0uae mo._COE0u:e mucouaum Aomopv goodnumo: co_uomem_ucm ommgocnn as: c_ ecuuew quu_w_cmmm we: co_uoaem_umm ouco_eoaxa co_uocem_umm Euu_ o— mao_>mcn mu_um_co~uecezu >o>cam ococno.uu .mEOOQOEm we co_uuc:» a mco>an Ammo-v cases: co_uoe~m_uom o»_n no_< ..ucoa m_ co_uoaemmmummm_c mouce_.nna .oauua + x00cnumo3 mmmmmuxumumm mco_u_cmeeo\unmawcamoox mumwmaflu mmmmmum mmmmqmmw meococaemoz 40 ucao_e_cm_m >domco>c_ omde mm: menace new >umuccad co_umEcO.c_ menu new: no» uoenEm was. >.uzm_.m .nmzmcoczc modaum emu soc gum: co_uocem_umm ecuum.\50u_ o.n_u.:E nanomeeo >Eucmumw_cm_m cacao ouam “weary .dmxc_gh nso— >mo >e>c3m ._mE oucomconxo m:O_>Ocn me__noeou:m ce_Occcu new >coccox >33. 35339.3 \o>_u_uonoE - acomucuoonxo mauvewm >.ucaome_cm_m ooco_cenxo amen co_umEc_wcoommO o» mmm>.ucc uuemnmoe ;u_3 acou_»_cm_m umoE acmc_E_eummu co_uoaem_uem m_ co>eommc no co_umEc_vcoum_c mon>u Ou Omgmco_ue.0c ace mco_ueuoonxo we mon>u mc_co*e_c Ammo-V >ccem co_uouuonxo m.eo.._x ce_unuoonxo we mon>u e *0 eucomum_xo ueonnam moon momcaou mucouaum pee cecu .>..mo co_umco.nxo Ocean gauges moc_>0cn mauan_cuuc no oocm_.nnm Auuce_ ooce_umn can .Coom co_uoewm_ucm .dem -co>coOv \e>mmmoOnxo new uncanam oEOm eceu mucouaum mOEOOu use cczm enamczuou >.uceu_w_cmmm >cm> Encomcon -coc can no comuoum_noE “coc_uc_ .mu_u_cu mon>u Hostage "mu_:moE pox_E ac_zuc.u euconaum propv Rocco: co_uueemmucm o» e>_mmoenxo “co_uueem_unmmmu o» doucosacumc_ "meEoOuao ucee0wemu >Eo>_ueu__eao Ou OOOoEOE co_uuo~m_uemm_u ecu co_uoowm_uem ucon.< ace «come: meannmcuue o>mmmoEnxo acme..mw.:m comueuoonxo Aesopu unonneoz new .eucosaeumc_ cu neuadmc co_uoo~m_umm mc_guo.u mucouaum mOEou new coxm .opox\>coocp mco_umcm.oa\ucoeoc3meox mmc_uc_n nuances muuomnam mcogueeomoz 41 Anm¢m_ne m_m>.~ca seas «cameos comumEc_wcoum_v o>_uueOeuec odoc_m we no:c_ucou A.~ w¢:o_a Oeuo.oc >.omo.u on3u_uue use mco_ueuuonxo uo_e_cn.u uoc nauseouoe a mo comumEe_chOm_c we maueum acoo_e_cmmm uoc .oncE ..oco>o co_uo.0ccoo OEom u_n_cxo co_uauoonxe ace co_uaEc_wc0um_O comuue~m_uam Ou modeum Eou_ o.n_u.:E paua.0c >de>mummon can co_u annopv deep .oCOE eo>_.o .mcmncummco. -aEemecoum_p new aco_ueuuenxo e_on0e dune dosed uco concede m_m>.eco co_u .ouoe .amaau nm¢m_n -ae._ecoum_n new co_saeuuaxo unease o>_uoe0eusc ceezuen co_ue.0ceou oEOm - ce_umEc_ecoOmmp use co_uumem_uam nuanceee -o.:n_cae oozeEcO»con can oucmEcOeeon >dco "no> acedn 5:5 co_uauuonxo muuacumcou cc_uomemmuom -ozucem>czu dunes .eucos .do no» wodoum uu:n_euua cu ouu.0e p_u co_ueEc_ecoum_c ce>o.n unaccouc_ “Nae-v uCQCOensm -_conxo eu_monEou ace .mno.m neon use mco_uauuuonxo “ucodn om_u Oou_> ..nE CCo .dmgocacu ao.oom no» mo_uw._ne_.oe so. meanness aco_uc0uc_ cc_uucwm_uam COHOOQOE assume .oco_>ecoa .eeo>om e_num ace omaCOLuon mc_:c_ucou odnEem >anEJm so», o.nmu.:E cu nee. co_uooem_uam o_unc ococno.uu a~oopv unmade» >ao we cc_uomem_uom mao_>oLn nee ucoec>do>c_ u__n:n sauces ace cozm acmumOu >u_u_.e> new >u_._no_.oL e>_mceuxm comuoeem_uem co .eodaoa eocuELOweon menswwo accedevmn saw: so.._: pecan uo>_oocon .mcomu mace: oc_c0m~_u accnnza o.nEem .eo a~oopv u_u_eoo .auoonxm .uceem .aacox eo_eon:a ._snocmoeu ac_xc0s ace «Swansea: peace umom nee pecan comuaEcmecoommc panacea, aucocaeuoOE Eocene .enanuco: .oauocou Ill / Mummmuxumumm muaqmaflu sonnets mummqmmm j 42 no:c_acoo u.~ maze—n oococ_wc0u aco_uouuonxo use .e.sm m_ we co_uumemmumm mun3u_uue mc_anozm-0cn Ou nope.0e euco_conxo “meme care 223% £92.95 £8. .88. 3 38 so a: 23m 288% 38: fits: ecu co>_.o co_uoa»m_ucm .cno.m acmo_e_co_m we: use .ucco ucoEuceneO macaw concoo .cexm .m>o. co_ueunacc an we mco_uauuonxo uEOQQSm uoc moon comuumwm_umm me: me anecm ucoEo>.o>c_ so. no. co_uaEc_ecoum_c o» Oeuo.oe muco_umee00u >.uceu_v_cmmm on:u_uua.oEn uo_~_oonm >_cc . ascem acme - .OCOE so umeu .dmco>o -o>.o>c_ nm_c c_ n_gwco_ua.oe on use .m_m>.aCe guns eeoE .enau_uue pee comuauoeaxo odeom acoec>do>c_ coozuon co_ue.oEeOo so. o.mom co_u mco_ueuoonxo acomm0ennam -oEe_ecoum_O e>_uom0cu0c we we: .>.co co_umEc_ecoom_c mu_uonnm cavemen deco: co>mdo mco_ueuucaxu eEouuao no co_uoc3e e we: co_uuoem_uam - uo_c .oCmn new ce>m.o pseudoEcou uoc mco_ucuomnxo ace ccmueEe_ecoum_n meanness co_uouooaxo .mno.m >u_eenm_n ucnu_*_cmmm up: «noueon - cede) om.a .._um use Lennon >._e_uceumnsm nm¢m_n deuce >umecdmmu unneeon-o:.e> u>_uoc0cuoc - co_u co: mco_ucuoonxo eczu_oc -mEcmecoom_n ooco_conxo ou:u_uua an >cOu Anwopv >.._o¢ dope: cu>_.o maom>ocn we ..auoc .du -umem_uam an ace mco_ueuuenxo mo__OOEOO:u mucopcum ecu xOOECEOHmon mmmmuuxummmm mam_u_c_eao\unmawcammox mmawmmwu mmmmmum mmmmqmmw mummmummmwm 43 Oe3c_ucoo n.~ mung—n madame; acmume_cm_m egos pou300cn o.mum co_uueem_uam Eo»_ -2322. of Ba 822:: -coum_u no» m_mum oucocewe_n e "madame; as“ we Oucmu_e_cm_m new on3u_cmee on“ c_ ucmucone_ m_ uoacumcou ecu no» acceaeammoe ecu .eo>ezo: co_umEc_~c0ummc m:.O mc_>_u-oen co_mm0emoO co_uuo»m_uom cc co_uucaw a w_ co_uumem_uem mcomumpc:o* new co_umEc_ac0um_p “guacamCOO acouconouc_ zu.ooc “Snowy Oc_eunagm co>_.o Co monumeos o>_uccEeu.e cm uoc m_ co_uaEcmec0ummu u_~0cn-coc .ecmn new deco: u_umEo_Oeen om.a mo>_uecEOude use .mo_u_._mne__oL so. rage magnum oococo~w_u co_mm0emec e.n_u.:E cow m_m>dece co_um.occou coeo_pacn amen one >.uc_o_ ouceeoew_u Co O>mumccOEou new o>_uoELoc >u_._ne_.oL no. mc_umo~ co_uumwm_uam comuuewm_uem .cno.u no. ecuuee accume_cm_m umOE comuoade>o omogocan.umon mm co_ua3.m>e amazocan.»mon we.:co~ ace co_quc_ecOu peccmvcm pauo.0c euuoceu mco_ucuoenxe .Em zmaozu.o .cc_uoaem_uem .eezenooz o>_uecmnEOo new cu pseudo; >.ucco_._cm_m coon. Remap. meanmCSOn use eo_._x .o>_umEeoc .e>_au_00cn umoE mcomueuuonxo o>mumEcoc poem anew mucenaum use nmexecn mmmmmuxummmw wcomu_c_eeo\uccawcamaoz m _uc_n mmmmmum mmmmammw mummmummmmm A major area of concern among researchers continues to be the nature of expectations. They have been conceptualized as beliefs about specific attributes (Olson and Dover 1976, Swan 1977, Churchill 1982, see Figure 2.7); beliefs about outcomes (Oliver 1977, 1980, Moore and Shuptrine 1984, see Figure 2.7); as value in use (Thirkill 1981); and as beliefs about anticipated satisfaction (Oliver and Linda 1980). Thus far, the major emphasis has been the evaluation of the varying definitions in terms of how well the procedure fit the measurement process. The use of attribute ratings stems from the long standing view of the nature of products and the predominance of multi-attribute models. The influence of Fishbein, who tendered a perspective and method well suited for the investigation of marketing phenomena, further solidified this as a preferred method. The approach has been criticized by Day (1977, 1982), in particular, as inappropriate for all goods and services. Within the context of satisfaction, a consumer cannot be assumed involved, although that assumption underlies the full attribute approaches. It may also be inappropriate where the product is familiar and frequently purchased. Studies directly addressing these issues are few. Swan and Trawick (1987) did find that involvement was significantly related to satisfaction Churchill and Suprenant (1982) found that for a durable product only post exposure measures affected satisfaction, while a less valuable item showed the disconfirmation effect to be significant. Westbrook (1928),'1980, 1981) has demonstrated that previous experience is an important factor in satisfaction for durables, but disconfirmation was not significant for automobiles. He also established a relationship between emotion and evaluation at the global attitude level for automobiles. The indirect evidence, therefore, tends to support the idea that involvement and/or type of good may partially determine the information processing and relationships within the full consumption process. 45 Information processing research itself reflects some of the same dilemmas: Troutmen and Shantiari (1976) raised the. question of whether consumers average or add attribute information Cohn, Miniard and Dickson (1980) argue that situational features affect the choice of averaging or adding attribute ratings. An extended purchasing process, level of confidence, and involvement all shown to affect satisfaction would appear as factors affecting information processing strategies as well. The definition of expectations as beliefs needs to be examined, therefore, for the type of beliefs encompassed. A contingent definition seems to be indicated: summary belief for some products and extended attribute evaluation for others. The interesting complication is the fairly consistent findings that expectation disconfirmation is less significant for durable products and some services (Bearden and Teel 1983). That is, in precisely those situations where one would expect extended information processing, the pre-purchase criteria do not seem to affect post-purchase judgments to the extent that they do less valuable products. The data are somewhat limited, however, since most of the research involving durables and services has been retrospective or laboratory. Thus, real time and usage conditions have not been examined (the exception is Bearden and Teel 1983). The use of beliefs about the product versus beliefs about the outcomes similarly reflects dilemmas in consumer research generally. The low predictive results obtained in purchase research based on Rosenberg, Fishbein and hybrid attitude models where purchase was predicted by attitude toward the object has led to the interest in Fishbein’s extended model which incorporates outcomes and choice probabilities. (Fishbein and Azjen 1975). The corresponding change in satisfaction research Ins been the development of expectations as beliefs about outcomes. No testing of differences in effects in the context of satisfaction research are known to this author. Oliver has used both types of measures with similar results (compare Oliver 1977a and 1977b). Additionally, many of the outcomes measures reported are measuring expected levels of attributes inherent in the product after purchase rather than effect of product usage (e.g. softness of pajamas, size of plant). The trend in most research has been to sum belief statements with great variation in the inclusion or exclusion of evaluative components where it is assumed to be incorporated into beliefs or stable (see Bearden and Teel 1983). To some extent, the statistical procedures have influenced the ways in which the expectations have been treated. Only two causal modeling studies have been published to date where the summative versus singular treatment issue does not arise (Churchill and Suprenant 1982, Bearden and Teel 1983). Summative measures are another question of interest. It has been proposed that a summative measure of overall expectation be included. Where used, the measure tends to fall within the scale for the other items. Conceptually it may be important if the intent of the research is to indicate a proposed difference in information processing between high involvement/high complexity goods and those for which a general expectancy may be a more accurate direction It is not an issue in this research. The use or nonuse of evaluative weights reflects the definitional dilemma: studies which have operationalized expectations as beliefs and also included an attitude measure have shown very strong correlations between the two with similar effects on satisfaction (Oliver 1980, Bearden and Tiel 1983). It may be argued that the expectancies are simply the beliefs portion of the underlying attitude (Day 1977, 1982). Alternatively it has been argued that since affect is stable during the disconfirmation processes, only belief changes are of interest (Bearden and Teel 1983). The introduction of normative influences and different types of expectations has further clouded the issue of what expectations are. At this point, there seems to be a consensus only that expectations include beliefs about outcomes with no resolution on any of the other issues. 47 5.88:. Emoo .o €38.26“. 2.. .o .65 < NQQF agave: 2.3m 958» 5.2.8:. 958» 28.: 3 e838 5 .3 .8228ch 3:88.50 25 5398:. 5323.383. EOEqu—uflo: :0 gasp—I ...... c. It u... mafia \ \ Gd wm—DOE . £23239; .2 53:00 .0 3392.5. 30.535 583.50 85> 2 8.523.? 5535 8 5.38:. 285950". 30:28:... .0 COB—EGO. 8:35.53. 58:00 eggs-done... 5.88: 35.880. 5326 u: wO the relative magnitude of difference score unreliability is increased as the correlation between the two measures on which it is based increases.“ (p. 3). Considering that Moore and Shuptrine found significant correlations between the expectancy and disconfirmation measures, the argument for use of difference scores and the role of disconfirmation, 57 generally significant in research where difference scores were used, is weakened. Alternative measures have their problems as well. Asking for retrospective judgements of disconfirmation raises problems of bim and face validity. It is clear, however, that measurement per se has a significant impact on results of research conducted in this area. SJIMMABX The status of research into consumer satisfaction from an expectancy perspective was well summarized by Swan (see Figure 2.8 and 2.9). Most current research rests on some version of learning theory from psychology although the differences important in psychology have not been as explicitly examirned in the consumer context. General agreement may be said to consist of the following: 1. Consumers use information from experience, other people, and goods providers tO form judgements which lead to product choice. 2. Evaluation of products after use occurs. Judgements, both cognitive and affective, are made about products which affect future intentions with respect to the product. There is no consensus on: 1. How judgements and evaluations are formed. 2. The basis of post usage evaluations 3. The strength of relationships of judgements across products and situations 4. The nature of judgements: predictive, idealistic, normative 5. Proper measurement techniques 6. The effect of individual affective states and personality characteristics on judgements in this consumer context. S8 The study proposed here will examine these aspects of the problem of consumer satisfaction and its antecedent: l. The relationship between general product class norms and specific product choice expectancies in impact on satisfaction and intention. 2. The differential and interactive effect of benefits and attributes on satisfaction and intention. The research is, as most in this area is, based on aspects of learning theory. It assumes that behavior is purposive and represents accumulated knowledge about what behaviors are likely to enhance goal attainment. It strongly suggests that cumulative learning establishes a base norm of what ought to occur in new situations and the range of acceptable deviations. CHAPTERIII Models, Hypotheses, and Definitions Intrsslusn’cn This research investigates possible antecedents of consumer satisfaction The analysis is conducted through the testing of models which specify both constructs and the relationships among them. The constructs and relationships have been developed on the basis of existing theory and research, which was discussed in the previous chapter (Chapter II). The models, in diagamatic form are contained in Figure Ill.1. The models represent ways of analwing the two research questions presented earlier in Chapter I. The first question centers on the existence of a product class attitude and its effect on the other constructs in the consumer satisfaction process. The essential premise is that expectations and evaluations of specific products are both grounded by reference to beliefs about the nature of the generic product class. The existence of a product class effect of this type provides for a fuller contextual exploration of consumption The research question can be stated as: 59 QI Does an attitude about a product class partly determine consumer satisfaction by affecting expectations and evaluations of specific product choices? The second question concerns the nature of product attitudes and expectations. This interest is derived from the notion that consumers seek products for the benefit use of such products may confer even though the benefits may not be observable at the time a product is chosen This concept is common to marketing definitions of product (see Kotler 1984). Over time it is plausible to believe that consumers may associate observable properties (characteristics) with desired benefits (goals). Whether these are separable at any point in the purchase process, and, if separable, exert different influences at different stages is an open question Thus the second research question is: Q2 Are product characteristics arnd goals separate conceptions and, if so, do they have differing effects on other elements of the consumer satisfaction process? W As reported in Chapter II, a variety of definitions, conceptual and operational, are found in the expectancy] satisfaction tradition The theoretical and operational definitions for this dissertation are presented here, along with the reasons for these choices. There are six major constructs in the models, as well as a postulated division of three contructs intotwoparts. 61 mm This is the consumer’s evaluation of the product type’s characteristics ahd benefits. A product type is a set of similar offering differing largely in non-essential features or levels of quality. Consumers may rate the product in its essence as their ideal, their vision of the average, or their view of the best possible. Operationally, in this study, the consumers were asked to rate all previous products in a way that implied lthe average] performance. Further, consumers were asked to rate the product class for the degee to which certain characteristics were present in the product, and the degree of salience each of those characteristics had for satisfaction or dissatisfaction with the product category. This operationalization differs from that of Woodruff, Cadotte and Jenkins (1982) where three different product class measures were obtained: best brand, product, and specific brand ratings. The average expectation rating in the same study were closest to the product rating and in post-usage ratings also the product norm emerged as more significantly Correlated with all other measures than the other two. Thus, the interest here is in pursuing this more general product norm. Finally, in the operationalization, consumers were asked to rate the product class for both characteristics of, and benefits obtained from, the product. The product class used in this study was business school classes. Finally, this construct is termed a norm although it is related closely to attitude 'a learned predisposition to respond consistently in a favorable or unfavorable manner with respect to a given objective" (Fishbein and Azjen, 1975, p.6). The learned predisposition requires information, experience 62 or some combination which is used to construct an image based on a set of beliefs and a judgement of the favorability of an object as it is believed to be. Wanna: (SPE) Specific product expectations are predictions about the degree to which a product will possess some characteristic or confer some benefits. These judgements are made prior to the use of the product purchased this time. There has been geat debate over what expectations really ought to be conceived as Miller (1977) suggested that four types of expectations may play a role in consumer selections: ideal, desired, minimum tolerable, and expected (predicted). Only desired and predicted have been supported (Gilly 1982). The efforts of LaTour and Peat (1980), Prakash (1980), Swan and Trawich (1980), and Wooer et al (1982) to expand expectation measures beyond simple predicted levels of attributes have indicated that some sort of normative standard is incorporated into prepurchase expectations and does correlate to satisfaction and post-purchase evaluation This corresponds closely to Miller notion of desired expectatiom. (Prakash and Lounsbury (1984) found that comparative expectations were also important, that is, specific product rated against competing alternatives with respect to the same set of attributes were significant in effects on satisfaction) In this study, many of the features of normative and comparative expectations were incorporated into the product class construct. Operationally, consumers were asked to predict both the degee to which an attribute or benefit was likely and, again, the salience of that feature to positive or negative evaluations of the product. Thus, the operational construct includes some degee of desired or normative expectations as well as the predictive expectation notion 63 The designed inclusion of benefits as well as characteristics also draws together some of the information implied by desired and comparative expectations, as the conceptual definitions of these constructs include the idea of benefits to be obtained. W (PE) Post-evaluation is the consumers judgement about the degee to which a product possessed certain characteristics or conferred some benefits subsequent to the consumers use of the product. Operationally, consumers were asked to rate the product after use on the same set of characteristics and benefits as they rated the product class and the specific product prior to consumption. 52mm (SAT) Satisfaction is a post-usage summary affective judgement as to the adequacy or inadequacy of the product. Swan (1982), in a review of the various conceptions of satisfaction, notes that the central dispute in defining the construct theoretically revolves on the question of whether satisfaction is a cognitive or affective state. In this study, satisfaction is conceived of as a summary affective state based on cognitive judgements but which may also include feeling or judgements which are not captured directly by looking at expectatiorns, attributes or benefits of the product per se. Feeling about other type of products not purchmed in order to purchase this offering may be present. (A vacation may be evaluated positively, but satisfaction levels may be lower than they would be because the vacationer is also regretting the VCR he cannot now afford. Day (1979) discusses this at length.) 64 Satisfaction is also conceptually different from attitude in that satisfaction represents a judgement about a specific experience whereas attitudes may represent judgements which are non-experiential. Clearly, however, there is a relationship between the two, which is explicitly represented in Oliver (1980) and Bearden and Teel (1983). Satisfaction was operationally measured by a set of questions asking for judgements about the overall product and the happiness/unhappiness the consumer felt as a result of using the product. The set of items constituted a satisfaction scale. Intenticma) Intentions are consumers’ expressed willingness or desire to repurchase or recommend the purchase of the same product. Operationally, consumers were asked to judge two types of intentions which correspond to intention to repurchase the brand and intention to repurchase the same product type. mg“) Attributes are the perceived features of a product. These are the aspects of a product which a consumer can assess most wily. Examples would be color, style, size. These were assessed directly by asking consumers to rate 10 characteristics of the product in question The ten items were selected from a larger set of potentially significant features (see Chapter IV for details.) These characteristics may be important in themselves as sources of satisfaction, or they may serve as cues to the desired benefits which are the object of the consumer’s purchase. M (G) Goals are the benefits consumers desire from use of the product. Goals are the benefits which the use of the product is expected to confer. They constitute the motivations for choice of the product class and the selection of the particular offering within that category. Operationally, the goals were represented by a scale consisting of five items chosen from a larger initial set. The development and refinement of the scale is contained in Chapter IV. The final items used in the arnalysis are shown in Figure III.1. (See Questionaire in Appendix for exact statement and order): W The models have been constructed so that each represents a set of relationships among the constructs which are relevant to the research questions posed at the beginning of the chapter. The hypotheses underlying the models are contained in the presentation of the structural models. The hypotheses and models are presented in a hierarchial testing order, that is, logically sequential. Some sections are dependent on previous sections. The use of a limited information causal modeling technique was specically chosen so that re-specifications could be made on the basis of information obtained at each stage. 66 The first model represents the null hypothesis of no distinction between goals and attributes. Model Single PCN --> SPE --> PE --> SAT --> INT Factor 67 OBSERVABLE ITEMS Teacher prepared and easy to take notes from Teacher enjoys teaching. Teacher respects and understands students Teacher is enthusiastic Teacher stimulates thought and hard work. The class is enjoyable. The class is not boring. Gives competence in the discipline. Give competence in business. Improves understarnding of my major. See progess toward graduation The class was similar to what was expected. Theclasswassimilartowhathadbeenheard. Extent to which feel satisfied with the class. Extent to which satisfied compared to other courses. Extent satisfied with what was learned. Ladder scale - Best to Worst class. Would take another course from this instructor. Would recommend this class to a friend. FINAL QUESTIONAIRE ITEMS Figure 111.1 The second model represents the converse:. PCNA --> SPEA ---> PEA-- Model D SAT -- > INT PCNG --> SPEG -> PEG Ho1 There is no difference between a model with product attributes as dual constructs and a model with product attributes and benefits as a single construct. Operationally, the best performing model with the distinction between attributes and goals will be compared to a single-factor version of the same model. As discussed earlier in this chapter, a central marketing concept that consumers buy products primarily to receive benefits,which are not entirely tangible at the time of purchase. Therefore, products are chosen with expectations about benefits - that they will occur and the level of occurance - but product choices often are made at least in part, on the basis of the more visible attributes which may serve as cues about the benefits. In some cases, particularly, but not only, service products, those attributes are also desirable in themselves and are in some sense, therefore, also a class of desired benefits - pleasantness of experience, etc. It also may be true that the more tangible attributes and goals are desired, but have unequal impacts at different stages in the consumption process, i.e., the more tangible aspects being more importantatthepurchasestageandthebenefitsatthepcst-usagestage. Inthecaseofserviceproducts, one might expect the experimental and tang’ble attributes to be more important at all stages than for hard goods which are primarily utilitarian. Thus, the atmosphere, decor, pleasantness of a dentist, may affect satisfaction as well as choice more significantly than the appearance, color and ease of ordering mightinthecaseofapowersaw. On the basis of this logic derived from theory and previous research, a reasonable expectation is that attributes and benefits should be related but independent concepts. It is plausible, however, that a singular relationship (essentially a multi-attribute conceptualization) may be better. 69 There are two types of testing for this hypothesis. The first type, whose results are reported in Chapter IV, requires the development and testing of scales to obtain measures of the proposed latent variable. Determining reliability and validity will, in itself, constitute a partial test of the relationship between attn'butes and benefits. The second test is the comparison of models 1 and 2 to determine which is a better fit to the data. The outcome of the test for Ho, constitutes a partial response to research question 2. Hypotheses two and three also concern that research question, and are dependent on the rejection of H01. Ho2 There is no difference in the impact on satisfaction between attribute and goals. H03 There is no difference in the way information is processed between attributes and goals. These two hypotheses are similar, but reflect two aspects of the issue regarding the status of attributes and goals. The first examines the overall effect whereas the second direct attention to marnrner in which such an effect may occur. The models which will be used to test Ho2 will include: Model A PCNA — SPA —-> PEA SAT—p INT PCNG _. SPG __> PEG I I I I I PCNA — SPEA —> PEA Model B 8 AT _, INT PCNG — SPEG -'———> PEG 70 The models represent a way of obtaining results which can be tested for an unequal impact on satisfaction The results of this will, in turn, be reflected in additional models designed to investigate other aspects. The second hypothesis requires testing the significance of differences in the variances between the beta coefficients for attributes and goals obtained from the test of the full model to the total varience. A statistically significant difference would result in rejection of H02. The next set of hypotheses are generated from the first research question All sets of relationships are reflected in the models contained in Figure 111.2 The test for rejection of the following null hypotheses are two. The models will each be subject to path analysis in order to determine the overall goodness of fit for the models. The models will each, in turn, be compared for significance to each other in order to determine which best represents the causal relationhips, if any, of all the constructs. Second, the individual links, as indicated, will be tested for their strength and direction The first 3 of this set represent the specified relationships among latent variables which are strongly supported by previous research as reported in Chapter 11. Ho3 There is no linear relationship between Specific Product Expectation and Post-Evaluation Ho4 There is no linear relationship between Satisfaction and Intention Ho5 There is no linear relatiomhip between Post-Evaluation and Satisfaction 71 The next of the hypothesis set are of great importance to the question directly raised about the possible causal influence of a product class on expectations, evaluations and satisfaction Hoo There is no linear relationship between Product Class and Specific Product Expectation Ho7 There is no linear relationship between Product Class and Post-Evaluation Hoa There is no linear relationship between Product Class and Satisfaction While the models allow for testing of H03, H04, and H05, the differences between specific models which will be subject to path analysis represent the direct tests for H05, H07, and Hoa Model D is a gaphic representation of the relationship specified in Hos, model C adds the information relevant to HO7 and Model B indicates the link referred to in H08. Again, if any of these specified relationships are not supported (the null hypotheses above are not rejected) the proposed models will be respecified to reflect those results. Theninthhypothesisis,inessence,asummaryofthewholethrustofthisresearchproposal: Ho9 There is no difference between any of the proposed models in the amount of variance explained. The results of investigating this hypothesis are important even if none of the proposed models is, in itself, significant for the identification of the 'best" model. It directs attention to those areas which are most relevant to theory and/ or model respecification Conversely, if any or all of the models yields significant results, the testing among models allows us to determine (with geater confidence) what elements may contribute to the better explanation 72 All models, presented here in simplified gaphic form, conform to standard formal path equations. Chapter IV contaim the details of sample, research design, and construct measurement. CHAPTER IV RESEARCH DESIGN AND THE MEASUREMENT MODEL The research design for this study takes into account the criticisms of previous research in consumer satisfaction/dissatisfaction Day (1979), Olson and Dover (1979), LaTour and Peat (1979), and Westbrook and Oliver (1980) have written useful critical commentaries. Summaries of past research with sugestions for future study by Hunt, Swan, and Woodruff, Codotte, and Jenkins (all 1982 at the Seventh Annual Conference on Consumer Satisfaction, Dissatisfaction and Complaining Behavior) were also useful (See chapter 2 for details). Essential elements of the research design are reported in the appropriate subsection However, the overall intent of the research - the exploration of previous product class experience effects on satisfaction and the relation of benefits and attributes - dictated certain essential elements of the research program. These included choice of a product that would minimize the effect of marketing elements not germane to the study (e.g. brand image), choice of a natural, self-selecting consumer group with some previous product experience, and unobtrusive testing measures. 74 This chapter reports the product, sample, and data gathering methodologies. The last section of the chapter contains a detailed discussion of data refinement and testing in light of the models and hypotheses of interest as presented in Chapter III. PRODUCT The product, a service offering, chosen for use is a set of marketing courses generally taken by a fairly wide range of majors. The classes selected were mid-level university courses, so that students had had other business school classes and had heard, presumably, enough about future classes to have formed some expectations regarding the new courses. The previous exposure allows for the formation of product class attitudes, and the preview information for erqnectation formation for the new offering. The courses also were selected to increase the chances of obtaining a range of product expectations; this was accomplished through analysis of instructors’ experiences and advisors’ impressiom from past student commentary and behavior with respect to the classes. The basic research design requirements for the product used were: 1. The specific product should be a new offering to the consumer (any respondent repeating one of the courses was eliminated from the final example) so that results would be generalizable to other new product offering situations. 2. The consumers reasonably could be expected to have prior experience with the product class to preserve the claim of an experience basis for product class attitude. 3. The product be one considered a high involvement item for the majority of consumers, again to improve uniformity of conditions for the research. 75 4. Brand name effects be controlled to reduce the number of marketing variables affecting results. 5. A remonable chance for a variety in attitudes and expectations for the product class and new offering expectations 6. A product for which prior evaluation and research were available for comparison of scaling and item results. 7. A product for whom those surveyed could be natural consumers that is the research would be conducted on those who would be chosing and using the product even if the research had not been conducted. 9. A product chosen, consumed, and evaluated in its normal setting. (The fact that course evaluations are required made even the evaluation portion of the research more natural than intrusive). A service was preferred, because service products provide benefits whose impact is often long- term. The impact of some services, restaurants for example, generally is immediate and not long- lasting, but other services - medical, legal - have outcomes which extend far beyond the time the service is performed. A product whose impact would be intermediate and less life-determining then the latter example, but of longer duration and more important than the former would be chosen as ideal for the purposes of this research. One of the criticisms of previous research is that the consumption and measurement periods were too short to reflect “real“ consumption experience. Use of a long duration service may increase the naturalness of an experimental situation The drawback is, of course, that the post-purchase evaluation would not reflect the truly longer term effects since the evaluation still would be solicited at the end of the service offering. 76 SAMPLE The classes used were one large section (200+ enrolled) of introductory marketing, four small sections (30-35 students each) of business statistics and four small sections of a marketing case course (50-60 students each). All classes are required for all business majors and for several majors from other colleges. These classes were taught by 7 different instructors. The fact that all courses used are required for business majors is a limitation of this research. Choice is limited to the term, time, and instructor rather than content. Thus, this research does not represent a number of purchming situatiom where the product is truly a discretionary choice. On the other hand, a number of purchase situations are, in effect, constrained choices for most consumers analogous to this one: detergent, new tires, medical or legal assistance. The required classes were selected because they guaranteed a larger range of students and because the number of factors affecting choice could be limited to a more constrained set. The total number of students to be sampled was 470. It was estimated on the basis of a pretest analysis that an initial sample of 450 would yield about 200 usable, complete, and non-duplicate respondent surveys, an amount deemed necessary for reliable statistical analysis. The final samme of 197 was checked using the formula suggested by Bagozzi for LISRE procedures as a guide for adequate sample size (Bagozzi, 1981) and, therefore, adequate for the path technique chosen for this research. The pre-usage instrument was ministered during the initial sessions of the spring 1984 quarter. The post-exposure instrument was admirnistered in the last meetirng of the courses. The researcher administered the surveys to ensure consistency. Demogaphic information (presented on Figure 1V.1) shows a reasonably wide distribution of 77 respondents by experience, sex, gade point, and major. Complete responses from the three courses were fairly proportional to the total sampled overall. Respondents were asked to fill in self-report questionnaires during class time . As similar course evaluations are routinely gathered in every course, every term, the testing situation appeared a normal procedure to the respondent. QUESTIONNAIRE DESIGN The survey instrument was developed in accord with the following criteria: 1. following a recommended format for attitude research 2. items were product specific 3. case of use for the respondents 4. reflecting the experience and recommendatiom of researchers in both satisfaction/ dissatisfaction and course evaluation The attn’bute and goal items for all three latent variables (Specific Product, Post-Usage, Disconfirmation) were rated in two ways: likelihood and importance. For Product Class, the rating were agree/disagree and importance. All items were rated on 5 - point anchored sales. The choice of agee/ disagree and likelihood combined with importance follows standard marketing research procedures, which, although often criticized as a departure from psychological practice, has been shown to work well in multi-attribute preference studies and attitude/purchasing intentions research (Wilkie and Pessemier 1973, Lutz 1981, Lutz and Bettman 1977). 78 No. 5 No. % Sex: Male nos 52 Female 95 ‘3 Course: Statistics: 30% Principles: 40% Case: 30% Number of Previous Courses: 5 or more: 72% 3 - 4: 19% less than 3: 6% Grade Point Average: 3.5 - 4 5 - 1 (4.0 SCIIO) 3.1 _ 3.5 33.8 2.6 - 3.0 46.1 2.1 - 2.5 14.2 less than 2.0 5 Source of Payment for Course: Parent 50.2% Self 35-295 Scholarship 53% Loan 5.5% Other 2.7% Major: Marketing 16% Finance 9.1% Management 16.4% Accounting 20.1% Packaging/Engineering 1.8% General Business 11.4% Advertising/Journalism 12.8% Retailing .5% Other 14.2% Table IV.1 SAMPLE DEMOGRAPHICS (FINAL SAMPLE) ‘ May not total 100% due to rounding or omitted answers Fishbein and Ajzen’s (1975) objections are not fully applicable to this research as the preliminary scale development included obtaining and refining a set of item/beliefs which were salient across the sampled population However, the particular saliency to the differing courses and persons differ. The issue of whether these self-rated weiglnts correspond to “empirical” measures is somewhat difficult to determine sirnce all judgements in the study are psychological and not empirical. Respondents were asked for student number, not name, and assured that data would not be retrievable by use of student number nor made available except in summary form, to instructors. DEVELOPMENT Two stages of pre-testing were used to develop the final survey instrument. First, 62 attribute- related items and 16 goal items were generated on the basis of previous research in course evaluations and from analwis of approximately 1200 open-ended course evaluation forms. (Wilson, 1982, Bertsch and Peek, 1982). Statistical analysis of questionnaires gathered in a basic marketing class resulted in a second imtrument which contained a reduced set of 30 attribute-related items and 16 goal items. Analysis of the second survey administered in a different course than the first resulted in the selection of 10 attribute-items and 5 goals for use in the final survey. The post-usage survey also contains 8 satisfaction and 2 intention items which were tested in both goups for reliability and consistency. ANALYSIS PROCEDURES Ninetycnine usable questionnaires were obtained in the first pre-test administration and item- to-total correlations, coefficient alpha, and factor analysis techniques were employed in the second pre-test, 839 questiormaires were obtained. In addition to the above procedures which were replicated 80 on the final data, a systemic sample of 10% was used to determine the preferred goals by importance and desireability measures where forced choice respondents were elected. Results were consistent with the first pre-test but allowed further refinement of the instrument. The attribute areas identified were: instructor, course content/atmosphere, and grading practices. These results are consistent with previous findings (Wilson L982). The goal items are related to benefits expected, which reflect, to some extent, the degee to which the content area itself is valued. This is also consistent with findings on courses and service products requiring extensive consumer involvement. The derived scales all had alphas above .75. (See Table IV.2 for full array.) DESCRIPTION The final instrument has two parts. The first sectionwasadministered durirngthe initialclass sessions and elicited responses to items reflecting product-class attitude and specific course expectations. The instrument contains 10 attribute items, and 5 goal items. The items were rated on a S-point anchored scale following standard attitude measurement procedures and wording. The second section was a post-exposure survey, asking for post-exposure beliefs about the same 10 attribute-related items and 5 goals, with 8 satisfaction items (7 verbal sales and 1 gaphic scale) and 2 future intention items. In addition, 5 demogaphic items were contained in the post-exposure questionnaire. (See Appendix for copy of the final questionnaire). W The primary focus of this dissertation is the possible effect of prior product class attitudes and beliefs on subsequent satisfaction for a single product choice. Refore testing of structural models relevant to the propositions regarding product class could proceed, the measurement models had to be developed and tested. 81 The measurement model, which specifies causal relations between the theoretical variables and responses to observed variables as directed by the theory underlying the structural model, requires four steps in order to detect both random and specific error. The first stage is the creation of scales and testing them in order to obtain undimensionality, reliability and validity. All scales in this research were so tested. Scales were constructed for Product and Specific attributes, goals, satisfaction and intention While all portions of the measurement model were so treated, the portions regardirng product-related and goal-related benefits constitute a secondary set of hypotheses which are central to the structural models. Procedures for examining the reliability and constructs will also serve as some of the tests for that hypotheses: Ho1 There is no difference between a model in which goals and attributes are part of a unidimensional construct and a model in which they are separate constructs. The tests are those recommended by Kenny (1979) and Anderson and Gerbing (1982). Reliability was assessed through Cronbach’s coefficient alpha applied twice: to each separate proposed sale and to calculated scales using confirmatory factor analysis. Split-half reliabilities for both the proposed scales and for the sample were computed, both as an additional source of information and asapartialresponse to criticismsbyBagozzi(1980, p. 128)inthisdiscussiononthelimitations of reliability measures as components of construct validity. Bagozzi’s major argument against both types of reliability testing is the problem of systematic, situational error. The use of two sample and several situations combined with both types of tests should give good, robust measures reliability. A similiarity coefficient index (Hunter 1973, Anderson & Gerbing 1982) was used to assess the unidimensionality of each the constructs. This is a measure of both internal and external consistency and also a check on interpretational confounding. For the purpose of H01, it provides additional information necessary to evaluation and confirmation] rejection of the hypothesis. (Full data is contained in the Appendix-The reliabilities for all tests and both samples were similar.) Summary results on the final item sets show: 891: W Plenum Product class attributes .78 7 Product class goab .72 4 Specific product attributes .95 7 Specific product goals .88 4 Attn'butes .88 7 Post-Use Evaluation .82 4 Goals .82 4 Satisfaction .84 7 2 ‘Intentions .61 ‘Less than three items, so Alpha scores are unreliable (SPSS, 1981). SCALE RELIABILITIES Table IV.2 (The similarity coefficients for the same scales are presented in full in the Appendix.) Thus, the individual scales reliability and validity were sufficient (Nunnally 1978) to proceed in testing hypotheses related to the models of consumer satisfaction The final items used in the analysis were (see Questionnaire in Appendix for exact statement and order) are listed in Figure IV.1. Attributes: W Teacher prepared and easy to take notes from Teacher ernjoys teaching. Teacher respects and understands students Teacher is enthusiastic. Teacher stimulates thought and hard work The class is enjoyable The class is not boring. Gives competence in the discipline Give competence in business. Improves understanding of my major. See progress toward graduation. Theclasswassimilar towhatwasexpected. Theciasswmsimilartowhathadbeenheard. Extent to which feel satisfied. Extent to which satisfied compared to other courses Extent satisfied with what m learned Ladder sale - Best to Worst class Would take another course from this instructor. Would recommend this class to a friend. FINAL QUESTIONAIRE Figure 1V.l The items which were discarded included all questions related to testing and gading. Maddox and Smart (1983) showed that students tend to discount those issues during a term, unless there is obvious unfairness, and accomodate to any prevailing system. The satisfaction items discarded also included questions relating to equity. There was an insufficient number to determine whether this should have been an additional, post-usage scale as some researchers suggest. However, the two items (numbers 5,6) did not intercorrelate highly with any other items or scales. Copies of computer sheets containing reliability testing, similarity coefficients, and factor analysis are in Appendix. (All other results related to hypothesis testing and models are reported in Chapter V.) Parameters for the measurement models were estimated by confirmatory factor analysis using PACKAGE (Hunter and Cohen 1969). This centroid multiple group analysis was chosen following Burt (1976) and Gerbing and Hunter (unpublished, 1980) who argue that the risks of interpretational confounding are fewer than from alternative metlnods. This procedure fits with the usual causal research paradigm wherein first one constructs a theory of causal procemes among a set of specified variables. This theory generates the predictions about the relationships which are stated in the form of a causal model tested by use of path analysis. Valid use of causal measurement techniques depends on unidimensional measurements, both internally and externally consistent. When this is not the case, interpretational confounding may occur. Burt (1976) notes that confounding happens 'when the sources of empirical meaning used to interpret an unobserved variable are different from those used to estimate parameters in terms of which the unobserved variable is interpreted” (p. 8). To prevent such an occurrence, both internal and external comistency conditions must be met. 85 In applying meuurement techrniques to models, one can only hope that the model is correctly specified. Sincethisisunlikelytobetrue, always,onemustchooseamethodofanalysiswhichwill be useful and adequate for respecification Two suggestions for information on respecification have been put forward: use of residuals (difference between predicted and observed variable covariances) or the use of first-order derivatives of the likelihood function Residuals from a maximum liklihood procedure have been deemw unsuitable for such purposes (Costner and Schoenberg 1973). Joreskog (1978) shows ways of using maximum liklihood information, however, these prOCedures are not applicable for multiple indicator models. A measurement procedure which does permit respecification in a grounded way and at the same time is sufficiently irnformative is most desireable. Following the argument of Burt (1976) a limited information confirmatory technique is what was needed. Multiple goups centroid analysis is just such a procedure as only the covariances of the Variables in each equation are used to estimate the parameters in the equation Maximum liklihood techniques, by contrast, use all covariance. Thus, a multiple goups technique limits the effects of any partial misspecification on the other parts of a model, while the maximum liklihood procedure does not. Therefore, the parameter estimates for the former procedure can be used to detect the misspecification Anderson and Gerbing (1982) provide examples of superiority of multiple goup techniques over maximum liklihood for precisely this sort of problem, while both techniques yield comparable results after respecification The tests for consistency - internal and external discussed above - also contribute a major portion of the information required for model evaluation This information, together with l’esults of reliability testing, allow one to determine whether one can proceed to test the full model. If both the reliability and validity are adequate, the measurement model can then be evaluated by Comparing observed correlations to the correlations predicted by the model. In this dissertation, path analysis is used to obtain that information The statistical results will be evaluated with reference to the causal models’ predicted relationships. 86 Since all the models presented in Chapter III are based on multiple causation processes, the path coefficients will be beta weights. All the models are recursive with one or two exogenous variables: either both product class attributes and product class benefits or a single product class latent variable. Path analysis is a systematic combining of partial and multiple correlation in order to investigate the causal relatiom among a specified set of latent variables. The parameters are estimated from the correlation matrix of latent variables constructed from the confirmatory factor analysis. As path analysis is now a fairly standard analytical technique, further information is better obtained from Blalok (1971), Hunter and Gerbing (1982), and Kenney (1979). The final steps in this research will consist of cross-model comparisons and testing of individual coefficients for significance as required for testing of the hypotheses. Results of the analytic procedures are reported in the next chapter and discussion of the implications lies in Chapter VI. CHAPTERV Analysis: Testing Hypotheses The hypotheses proposed in Chapter III were tested usirng the path analysis progam PACKAGE. This chapter contains the results and conclusions. Implications of these findings are contained in Chapter VI. HYPOTHFSIS TESTING The first set of hypotheses examines propositions relating to model specification The second set examines the significance, if any, of specific relationships in the best fitting models. The models described in Chapter 111 each represent a set of relationships derived from theory and previous research. Each differs in at least one respect reflecting plausible alternative descriptions of the satisfaction process. The essential differences reflect the two basic research questions: 1. The possible effect of a product class norm on the evaluation of specific product offering, and, 2. theunidimensional/multidimensional nature of the constructs antecedent to satisfaction The full correlation matrices, similarity coefficients and confirmatory factor analyses are contained in the Appendix. The models and path coefficients for the five models tested are depicted in Tables 5.1, 5.2, 5.3, 5.4, and 5.5. Ho, There is no difference in a model in which goals arnd attributes are parts of a unidimensional construct and a model in which they are separate constructs. The significance of differences between models was tested using Chi Square. (Table 5.6 contains the set of comparisons). The differenCe between model A (dual-construct) and Model A (single-construct)wassignificantata - .05. Thenthenullhypothesiscannotbeaccepted asthesingle factor model was a significantly better representation of the specified linear relationships. The display of Chi-square results also indicates that all models are significantly different from each other and thus Ho9 is also rejected. H09: There is no difference between any of the proposed models in the amount of variance explained. The full array of models tests this proposition Model A, which specifies both relationships, is significantly better than any of the models which specify no, or only one, of the relationships at Chi Squarea - .05. Thuathenuflhypothesiscannotbeaccepted. The results in combination indicate that model A (single construct) represents the best specification of relationships for this data: Model A (single construct) Produ S cciflc—b fl “—5 Product Post Evaluation —> Satisfaction —> Intention 89 Product _, specific Product —> Post-Evaluation Class mug“ Attributes Attributes Satisfaction _.p Intentions Product —> Specific Goals ——p PoapEvalnation/ Class Goals Goal Rgroduced Correlations LN .76 .19 .09 -.14 .04 -.01 .-.01 LN .14 .12 .11 .(5 .N .00 LN .02 -.76 .01 -.18 -.14 LN -.01 .38 .22 .17 LN -.01 .24 .19 LN .59 .46 LN .78 LN Observed Minus Predicted Correlations PCA PCG SPA SPG P-EA P-EG SAT INT 0 0 0 .10 39 .61 .71 .57 0 -.(B 0 .33 .38 .50 .89 0 -.72 0 .19 .50 .38 0 .82 0 .14 .19 0 .25 .15 .13 0 .06 -.07 0 0 2 X =3 222.5 with 19 d.f. Table 5.1: Model D 90 Class Attributes Aim‘b tes Attributes ' \ ' Satisfaction —> Intentions Product -——§ Specific Goals ——> Pupfinlufion / Class Goals Goals 1 Reproduced Correlatives 1.00 .76 .19 .09 .24 .33 .25 .2 1.00 .14 .12 .19 .43 .30 I! 1.00 .02 -.76 .06 -. 15 -.12 1.00 .02 .38 .23 .18 1.00 .08 .29 .23 1.00 .61 .48 1.00 .78 Observed Minus Expected Correlations 0 0 0 .10 0 .32 .45 .36 0 -.03 0 .03 0 .20 .65 0 -.72 0 .14 .47 .36 0 .79 0 -.15 -.20 0 .17 .10 .09 0 .04 -.09 0 0 x2 a 180. 71 with 17 d.f. Table 5.2 Model C 91 I Ch” Attribute- Attributes Attributes \ Satisfaction —> Intentions Product ——-§ Specific Goals —> Post-Evaluation / Ch” Goals ”°"' 1 Reproduced Correlations L00 .76 .19 .09 -.14 .04 .44 .34 1.00 .14 .12 -.11 .05 32 .25 1.00 .02 -.76 .01 -.07 -.05 1.00 -.01 .38 .15 .12 1.00 -.01 .14 .11 1.00 .32 .25 1.00 .78 Observed Minus Predicted o o o .10 .39 .61 16 32 0 -.03 0 .33 .33 .18 .64 0 -.72 o .19 .38 .29 0 .82 o -.07 -.14 o 35 .25 .21 o .33 .14 0 0 2 X = 156.58 with 17 d.f. Table 5.3 Model B 92 P‘l‘mnfl -—> Specific Product —> Post-Evaluation Mgfim Attribute- Attributes \ Satisfaction —> Intentions "04““ —> 3M5“ 60‘“ —'> Post-Evaluation / Class Goals “i“ t Reproduced Correlations 1.00 .76 .19 .09 .24 33 .60 .47 1.00 .14 .12 .19 .43 .50 .39 1.00 .02 -.'I6 .06 -.05 -.04 1.00 .02 .38 .16 .13 1.00 .08 .34 .27 1.00 .46 .36 1.00 .78 1.00 Observed Minus Predicted Correlations 0 0 0 .10 0 .32 .10 .09 0 -.03 0 .03 0 .01 .50 0 -.72 0 .14 .37 .28 0 .79 0 -.08 -.14 0 .17 .05 .05 0 .19 .03 0 0 0 x 2 = 95 with 15 d.f. TABLE 5.4 Model A (two factor) 93 I Class Attributes Attributes —> Satisfaction—p Intentions Attributes Reproduced Correlations LN .21 .59 .67 .53 LN .86 .66 .53 LN .86 .68 LN .80 LN Observed Minus Predicted Correlations 0 0 0 0 .20 0 0 -.29 -.3 0 0 -.01 0 0 0 x 2 = 22 with 44.1: TABLE 5.5 Model A (single - factor) 94 Model D 5.99 Model C 3.84 Model B 5.99 Model A (Dual) 19.68 Model A (Single) Alldillerencessignificantata< .05lowestvaluesreported TABLE 5.6 CHI-SQUARE TEST FOR DIFFERENCE BETWEEN MODELS Produ: Clns Specific Product Post-Evaluation Satisfaction lntenslon Expectation Product Class Attitude .240 -.538 843 -.966 Specific' Product Expectation .326 .339 -1.25‘ Post-Evaluation .341 1.133' Satisfaction -.989 7-7507 0' PATIO 005m SINGLE FACTOR MODEL WI -.-1 new moo-t m not. PM sees an. a seam mo. sou as .«7 w as 1.00s m. a ' as an nu ma- an. as "I. I I .0212 «to us as: has “a", nut .731 .m .232 W .m as: m- "I am em- “ asst west or am coemcam wanna-t n 0M or up two rseron nooen. Reproduced Correlations Satkl’action t-value PCNA .60 985 PCNG .50 1.773‘ SPEA -.05 .0357 SPEG .16 .173 Significant at a < .05 TABLE 5.7 Model—A(dual-coutruct)pathCoeffidenr 95 All of the models were tested for goodness-of-fit using Chi Square as an indicator. All models were significant and therefore, following causal modeling procedure, must be rejected as representations of a linear relationship among all elements in accord with the data (See Tables for values). These results accord with the two other investigations which employed causal techniques (Churchill and Suprenant, 1982, Bearden and Teel, 1983). The best fitting model had a Chi Square of 22.4 at 4 degees of freedom, 2 < .005). The improvement in fit, significant for each additional relationship specified, does indicate that the more complex formulation represents a markedly better representation of the set of relationships. Investigation of the significance of the individual coefficients indicates areas where further investigation may be warranted. A Bentler and Benett (1980) suggestion for determining incremental fit by changing the null hypothesis to a form which tests the proposed model(s) against a completely independent model was deemed unnecessary as the progessive testing of the models proposed demonstrated essentially the same pattern. The next set of hypotheses concern individual latent variable relationships which were examined by testing the significance of the path coefficients generated by the best performing models. The t-test was computed following Dillon and Goldstein (1980, p. 226), where path coefficients are treated similarly to beta coefiicients: t-b,-13l sci. 96 Where bI - path coefficient Bl - 0 S - unbiased estimate of c" - element in ith row and ith column of the inverse matrix. This was appropriate even though the true distribution is unknown since the sample size was large and the t-test is robust for departures from normality (Churchill, 1983, p.528). The full set of t values for all coefficients in the single-and dual-construct model A are contained on Table 5.7. Those relevant to the stated hypotheses are discussed below. H02 There is no difference in the impact on satisfaction between product-related attribute and goal—related attributes. Table 5.7 displays the relevant path coefficients and their significance test results. The results indicate that Product Class goal-related attributes are significantly related to satisfaction and post- usage evaluation. The product class attributes are not significantly related to any other construct. None of the paths for specific product are significant, although the goal-related attributes are closer to significance than the product attributes. Thus, the evidence suggests that there is a difference in the impact on satisfaction (and post-usage evaluation as well). 97 Thus, overall the null hypothesis cannot be rejected on the basis of this research. However, there is evidence of a more complex relationship which should be investigated further. The remaining hypotheses concern the individual pairs of relationships embedded in the best fitting models. H03: H04: H05: H06: H07: H08: There is no linear relationship between Specific Expectation and Post-Evaluation. There is no linear relationship between Satisfaction and Intention. There is no linear relationship between Post-Evaluation and Satisfaction. There is no linear relationship between Product Class and Specific Product Expectation. There is no linear relationship between Product Class and Post-Evaluation. There is no linear relationship between product Class and Satisfaction. Testing for each of these was accomplished through a t-test for each of relevant pairs of path coefficients. The results are presented in Table 5.7. The table presents the testing of path coefficients for the two best models - model A (dual-construct) and Model A (single-construct). There are two points of interest here. First, although the single-construct model had the best overall fit, there are more individual pairs significant in the dual-construct model. Inspection of the individual path indicates that one aspect of the specific product was problematic for both models. It exerts more influence on the dual-construct model since there are more relationships dependent on this construct (specific product-related attributes). A fuller 98 consideration is presented in the discussion section of this chapter. Second, the arnalysis indicates the strength of individual relations. Analysis of the coefficient indicates that for the best fitting model, only the relationship of Specific Product to Intentions is significant at any reasonable level, although the relationships of Product Class to Intention are close to significance levels and in the correct direction. For the two-factor version of the same model, six relationships are significant, including Product Class goal related attributes to Satisfaction, both product-and-goal-related attributes to Post-Evaluation, and Satisfaction to Intention. Impection of the full array leads to acceptance of the null hypotheses except in one clear case: the relationship of Satisfaction to Intention. There is also a strong indication that Product Class is related to Satisfaction and Intention. This latter point will be considered in the discussion section. A further test for the seventh and eighth hypotheses lies in the relationship of specific models to each other. H08: There is no linear relationship between Product Class Attitude and Post-Evaluation. This hypothesis was tested by comparing Model C and Model 0 where the difference in models was the relationship of interest. Model C was significantly better than Model D at a :- .05. Thus the null hypothesis cannot be accepted as the model including a direct linear relationship is significantly better in overall goodness-of-fit to the one without such a specification. Hog: There is no linear relationship between Product Class Attitude and Satisfaction. The difference in Model B and Model C is a test for this hypothesis. Model B which specifies such a relationship is better at a Chi Square a - .05. Thus the null hypothesis cannot be accepted. 99 However, testing for the significance of the difference of magnitude in impact yields somewhat mixed support for rejection of the null hypothesis. For the product class constructs, the difference between goal-related and product-related attributes with respect to satisfaction is not significant, although the difference for the relationships to Post-usage evaluation is significant. The actual results for the difference testing is continued on Table 5.8. Neither specific product differences is significant. Ell. l' I' I“ [In Product Class and Post-image Evaluation 14.75 Product class and Satisfaction .354 Specific Product and Post-image Evaluation .244 Specific Product and Satisfaction -.594 ‘Significant at - .001 TABLE 5.8 DIFFERENCE IN INPACT Dual Factor Ilodal Single Factor Modal um m let m PCN b NT .20 OPE b SAT -.29 - SPE 8) INT -.23 §§§§§§ ézggéé 383338 Mania-WW: INDIVIDUAL COMPARISONS. MODEL A TABLE 5.0 1N One further point of analysis lies in the inspection of the matrix of Observed minus Reproduced correlations. A maximum acceptable deviation limit was calculated and each path coefficient’s 'difi'erence m inspected for diagnostic information (See Table 5.9). For the best-fitting model, the three values of interest are the relationships of Product Class to Intention, and Specific Product to both Satisfaction and Intention. See the discussion section for an explanation and suggestion for further research. Also of interest is the marked deviation from the predicted relationship between specific Product goals and Post~evaluation product-related attributes in the two factor version of the best model, as well a the relationship of the two Specific Product variables. 1215321831913 The conclusions derived from the foregoing analysis suggest that the two research questions posed in this research do deserve further investigation. The significant incremental improvement in the models provided by the extension of the range of Product Class on the other latent variables indicates that the inclusion of a product norm component provides a better explanation of satisfaction than an explanation without such a structural variable. The second research question asked whether the benefits attained from a product and the more observable product attributes were separable, affecting post-purchase evaluation and satisfaction in different ways. Fnrst to what extent does this explanation accord with explanations of a similar nature? Second, to what extent are some unique factors contributing to the results? 101 W The basic premise that experience and/ or information about a generic product should affect the evaluation of a specific product offering rested on three considerations. First, research by other investigators indicated that purely predictive expectations about specific product performance were less satisfactory in explaining satisfaction than more normative measures and, that irn combirnation with normative measures, better explanations still were derived from data analysis. Second, the goup of learning theory and other cognitive theories in psychology predicting that previous experience serves as a gound for new situation evaluations has been supported by a long and rich research tradition. Third, there is evidence from marketing research outside the satisfaction area which suggests that irnitial learning establishes the conception of a product category and a pattern of affective response. Chapter 11 contains a more detailed discussion of these issues. Satisfaction research into types of expectancies has shown on a consistent basis that something other than predictive expectations for specific product operates to mediate satisfaction with the product. La Tour and Peat (1980) concluded that satisfaction was an affective judgement comparing a specific product’s performance with experience of similar brands. Although only those with less satisfactory previous experience than with the focal product expressed high satisfaction, the results indicate that previous experience was governing the response to the new product for all subjects. Swan and Trawick (1979) examined the effects of two types of enmectatiom: desired and predicted. They found that desired prediction confirmation was a better predictor of satisfaction. Westbrbok ahd Newman (1978) determined that satisfaction with products previously purchased were highly determinate of satisfaction in new shopping experiences. Gilly, Cron and Barry (1982), testing among all four of Miller’s proposed expectation types, concluded that previous experience had a strong impact on predictive expectations while 'deserved' expectations had a stronger effect than predictive. Swan and Martin (1980) reported that confirmation of past experience in a new product as more strongly predictive of satisfaction than 102 new product expectancies. Cadotte, Woodruff and Jenkins (1982) found that the confirmation/disconfirmation of the product norm (and the best brand in category norm) was a better predictor of satisfaction and intention than the brand norm. Prakash (1984) in a direct test of predictive, normative (desired) and comparative expectations, found. that both normative and comparative expectations were more significant than predictive in expqu satisfaction. Finally, there was some indirect evidence from both Churchill and Suprenant (1982) and Bearden and Teel (1983). In the former it was found that only performance and'not prior expectations had an impact on satisfaction for a video disc, but the opposite was true for the plant in the same experiment. Although they concluded that durables may be evaluated differently than non-durables, it is noted that the VDP was a new product generically as well as specifically, while the same is not true for plants. This may indicate the difference in having and not having a normative standard of reference, and thus performance alone is used as the judgement base. Bearden and Teel, using automotive repair, found they could not support disconfirmation of predictions as a predictor of satisfaction. However, there were strong indications tlmt a disconfnrmation effect is less powerful than norms, suggesting again that predictive expectations for a single offering are insufficient. Thus, this research tends to continue to provide support for this general view that evaluative judgements are made on the basis of general product norms and are thus consistent with proposed models herein and in Cadotte, Woodruff and Jenkins (1982). Such a view is strenghthened by examinirng the sources of the lack-of-fit. Product class coefficients were better performing indicators than Specific Product relationships. Additionally, only two other causal modeling tests of expectation] satisfaction paradigns have been conducted. In both cases (Churchill and Suprenant 1982, Bearden and Teel, 1983), the overall models had to be rejected despite indications of good incremental fit. As a similar pattern was in evidence here, examination of common sources of problems appears warranted Two areas appear 103 to be important. First, measurement of constructs continues to create problems. Each previous study reported problems in individual measures. This study is no exception. The single most problematic variable was Intention. Examination of confirmatory factor analysis and the similarity coefficient index, after viewing the reliability for the two item scale, strongly suggests that two different outcomes were being associated in one factor. Examination of the raw data suggeststwosources. Fortinisproductmsersmadeadistinctionbetweentheserfice productandthe person performing the service. Thus, the goals which were closely associated with the product do associate more strongly with intention to recommend the product, rather than the offerer. Product-related attributes were more closely modated with the person. Since a number of different instructors were incorporated, sharp disfike of the instructor in two clmses strongly affected the overall pattern. A charnge in the measures might have made a significant difference. However, this is a question of importance for many types of services. The implications will be discussed more fully in Chapter VI. The second area indicating the reason for the problem in goodness-of-fit lies in the relationship of specific expectatiom to satisfaction. From the single-factor model, only limited inference is available. Inspection of the two-factor model provides more information. The source of the problem, looking at the Observed minus predicted matrix (see Appendix); is the specific product goal-related attributes. First, these attributes have a very low path coefficient for relationships with product-related attributes and also with post-evaluation on product-related attn'butes. Conversely, it is a better predictor of satisfaction than the product-related attributes. The strongly negative coefficient for product-related attributes and the corresponding post-evaluation is another piece of evidence. It is suggested that since approximately one-third of the respondents were evaluating a course which is, in student folklore, a stressful one and looked upon as only a necessary evil (statistics for Marketing majors), it may well be that a number of students predicted that the service would be unenjoyable but would provide necessary 104 benefits. Thus, this may be a measurement problem specific to the research. Unfortunately, it was not possible to refine the sample to exclude responses from the courses in order to verify this belief. This observatiorn, however, does fit with the results indicating that within the two-factor model goal-and product-related attributes were processed differently. The high correlation of the two variables, evident in the product class area, may suggest that attributes are in fact evaluated in terms of benefits their presence seems to promise. The mixed results and superiority of the simple-factor model should not be understood as conclusive. There are also two other measurement issues. The first is the question of the reliability of different scores. Although this research did not use difference scores per se, the effect of using item-byoitem post-ratings is the same. Prakash and Lounsbury (1984) raise the issue as it has been a concern in psychology research for some time. The argument is that the low inter-correlations produce some of the observed low correlations between confirmation and satisfaction, although in many cases the post-purchase variable alone is a good correlate of satisfaction. In this research, the relationship pattern was different for Product Class variable and Specific Product variable where the ”typical” pattern was evidenced only for the product class variable. The reliability of difference was computed using Lord (1963) as suggested where for the Model A (single factor) variable, the results were: Product Class - Post Evaluation: .683 Specific Product - Post Evaluation: .467. 105 Both of these are higher than the results reported by Prakash for his study and higher than most of the reported reliabilities for the multiple measures in Cadotte, Woodruff and Jenkins (1982). While these are not as high as the individual scales, they are reasonable given the measurement issues already discussed. The second measurement issue is related: how should the variables be operationalized. The theoretical issues will be treated in Chapter VI, but there is one consideration relevant here. It has been suggested that a summary difference measure for post-evaluation be employed rather than an attribute post-rating or as a supplement. No such measure was used in this research. However, the pattern of responses ternd to indicate that method probably did make a significant difference in the overall model outcome that could be explained in terms of the issues of direct concern to this research. The slight superiority for the difference scale found by Moore and Shuptrine (1984) raises yet another issue about prepurchase and post-purchase judgements. SHMMABX The implications of these findings will be discussed in Chapter VI. The findings may be summarized as: 1) Each of the proposed models is significantly different from all of the others. 2) The role of product class as a determinant of satisfaction through its effect on specific product evaluations and satisfaction directly is supported. 3) The proposed differing effect of benefit desired and product attributes on the evaluation and satisfaction processes are indicated, but not fully supported by this research. 4) Only six path correlations, and none in the best model, were individually significant. However, most of the significent paths were between product goal-related attributes and other constructs. 106 5) Examination of the coefficients for the best performing models indicates that some of the results may explained by some measurement properties unique to the product and this study. 6) The hierarchical testing procedure, similar in conception to incremental fit procedures, strongly suggests that the two research questions were well-founded. 7) The results are consistent with other causal modellirng investigations of expectations/satisfaction and also generally consistent with thrust of the Cadotte, Wooer and Jenkins (1982) model. Chapter VI Summary of Research, Limitations and Implications This study investigated some factors which bear on consumer evaluations of consumption. The research was conducted within the framework of general expectancy theory, examining some variables which had not previously been investigated into a marketing model for consumer satisfaction. The two research questiorns posed at the outset were: Q1. Do norms for a product class partly determine consumer satisfaction by affecting expectations and evaluations of specific product choices? Q2. Are product characteristics and goals separate conceptiom, arnd, if so, do they have differing effects on other elements of the consumer satisfaction process? Models were developed and data collected to investigate these propositions. Results indicate that models incorporating product class effects on expectation and post-usage evaluation are better than those without their explicit incorporation, although no model was significant statistically. These results are similar to previous causal modeling efforts in the research tradition. The study also showed that consumers evaluate both characteristics and goals jointly, although there are indications that the information is processed somewhat differently. 107 108 FINDINGS The following contains the more important information gained from this researcln. Limitations and implicatiom will be discussed at the end of this section. F'mding One Product Class norms do have an impact on past-usage evolution and consumer satisfaction. The analysis of results gained through hierarchical model testing indicated that product class norms do affect all success elements in the overall model of consumer satisfaction. The primary implication is that consumers base product choices on beliefs about the generic product. Further, they evaluate a chosen product’s performance by comparing it to the expected generic product performances and the level of satisfaction is determined by the actual performance and the comparative performance. This supports the general proposition advanced by Codotte, Woodruff and Jenkins (1982) that consumers, from their experiences, develop expectatiom about the level of performance they should receive... This standard or norm maybe different from what a consumer predicts he/she will actually receive...If brand performance predictions are used in a disconfirmation process, then experiences with competing brands and product types will not influence resulting satisfaction. Since the test results are markedly improved as the range of product norm effects are increased, it must be concluded that predictive, specific product expectations are influenced during both choice and post-usage evaluations by learned norms, which leads us to the second finding. Finding Two Product class norms do affect Specific Product Evaluatiom. 109 The implication of this conclusion is that consumers do make choices of products based on beliefs and feeling about the general product class. Whether, as we often assume, consumers attempt to choose the "best“ product or not, they clearly do make choices against the backgound of other available products. The results of testing differing types of experiences, reviewed in Chapter III, suggests that consumers may, in general, use satisficing strategies. That is, they choose products which are, they hope, no worse than the generic norm, but do not “affect“ the ideal. Rather, they appear to hope for better than average. The Codotte, et. al. (1982) study of normative levels support this interpretation of the results. There, comumers asked to rate the Best Brand, Brand Chosen, Product Norm and Brand Norm chose products rated slightly above the product norm but significantly lower than the hypothetical Best Brand. In this study, there was a strong relationship evidenced between product norms and specific product expectations, demonstrating a similar pattern. Finding Three Benefit and Attribute information is processed differently although they jointly affect post-Choice Evaluation and Satisfaction. The high correlation between desired benefits and attributes resulted in a better model of satisfaction’s antecedents when the two sets of indicators were considered as a unitary latent variable. However, the difference in their relative contributions to other variables as revealed by testing models which separated the indicators into two sets showed a difference in the contribution toward satisfaction. This suggests that there may indeed be shifts in consumer’s evaluation criteria during the total consumption process. The major implication is that further research is required and this is discussed later in this chapter. It tends to support the notion raised in research question two, where it is argued that attributes serve as cues as well as ends in themselves. As learning occurs, there may be a shift in 110 relative strength. Alternatively, it may be a product-linked characteristic: for some products, outcomes are more significant overall, whereas for others, the attributes have a geater value. Finding Four Post-usage evaluations are a more significant contributor than expectation/disconfnmation to satisfaction. One of the major points of dispute is the role of disconfirmation as a moderator of satisfaction. This study did not directly test disconfirmation, using inferred disconfirmation measures by comparing pre-usage to post-usage ratings. Examination of results shows that while disconfirmation/confirmation combined with post-usage ratings are important predictors, it is post-evaluation which is the significant factor. It also is similar to patterns in Suprenant and Churchill’s (1983) data for VCR’s. This lends support to the view that involvement may be .a significant moderator of the entire evaluation process; contingent models may be required. Finding Five Specific product offerings are evaluated within the context of a product class standard. Model data in all formulations shows clearly that specific performance is strongly associated with satisfaction, but the second geatest associationiswithtlne product class norm. Ifthis is so, we cantentativelyinfer that whatever the specific offering expectatiom are, the norms for actual performance are those derived from all previous experience and not merely some induced expectations. This finding is tentative, however, as expectations were not manipulated. If through advertising word-of-mouth, brand halo, etc., comumers felt they had been promised a geat deal more than they received, the discrepancy and resulting disappointment] anger might produce a more marked effect. 111 LIMITATIONS Any research is limited by the time, place, and situation. However, specific conceptual and methodological considerations introduce other constraints on generalizability. The choices will be considered here as they bear on future research directions. This study investigated only one product, which, for the majority of consumers, necessitated a high level of involvement. Less involving and stressful products may show quite different patterns; indeed the research summary reported on research indicating that expectations may be more important for less involving goods. Further, this was a service product which required effort from the consumer to use. A less demanding situation, possibly less affected by motivational and attribution issues might show different effects from experience and benefits on satisfaction This research also looked at a product not “freely" chosen While there were good reasons for designing the study this way (Chapter IV), there are limitations on generalizing the results to products where choice is less constrained. The lack of choice may have had depressor effects on both expectation and performance. However, Cadotte, Woodruff and Jenkins (1982) showed that specific expectations for a new experience with known products were higher than brand norm levels. Thus, expectancy ratings may not be influenced as much by choice a by eternal hopefulness. Nevertheless, a study of the same product with choice might produce differing considerations and results. A final conceptual issue is the reality that alternative models specifying different linkages might have been proposed. For example, one could argue that knowledge of the general product benefits may affect expectations about product attributes or consumers have learned to use the attributes as cues to benefits. Thus, one might plausibly have proposed a model with a linkage between product benefits and specific expectations. However, as the intent of this research was to explore the general plausibility of product class effects on satisfaction and expectations generally, this was not tested. The mixed results concerning attributes and benefits certainly should direct attention to exploring relationships as a separateissueaswellmwithinthecontext oftheparadign. 112 There are several purely methodological issues which should be noted. First, the items and scales used to gather data for analysis were tested only for reliability and used only to test the models irn that sense. They were not intended to predict choice or repeat intentions per se. Because these scales are not predictive measures, the research should not be interpreted to mean that satisfaction results from specific product conformity to some general product norm. In fact, this is a major issue for further investigation Second, the measurement models and procedures followed resulted in underpredicting the later conceptual links. Thus, the results should be interpreted with caution A future reanalysis looking at interaction effects and corrections for underprediction may generate results more strongly supporting some of the proposed models. Examination of the data tables in the Appendix shows some clear indications of where attention should be directed in exploring linkages not proposed, which may have led to the underprediction One in particular was the previously discussed product benefit-specific attribute connection A third limitation, methodologically, is the non-parallelism of results. In two of the three offering, the ratings for future intentions were quite consistent. For the third offering, there was a geat difference in the responses to the two intentiorns items: The respondents clearly differentiated between the service and the service provider. This ”inconsistency” not only affected the link between satisfaction and intentions, but has implications managerially and conceptually. Finally, there was no investigation of motivation and interest with respect to the product. Although some of the goal items indirectly assessed motivational aspects, this was not specifically addressed. Implications and Future Research Although earlier research examined expectations and satisfaction for advertising implications, the potential utility of this approach is much broader. The notion of product norms as base guidelines for new offering supports the concept of product positioning, but allows for an understanding of the essential elements of a product as well as the distinctive elements of a brand. A focus on core 113 properties as seen by consumers might provide a method for looking at core products which would reduce new product risk. The research clearly supports the notion that product categories are defined by post learning the limits ofproduct change and innovation/diffusion are not yet areas of research focus. Discussions of what makes a new product “new" might be more productive if this line of approach were developed. Pricing and advertising strategies designed to conform to existing notion of value and cue/benefit concepts, too, might be more successful in new product launch. Finally, understanding of the existence of differences in what is evaluated differently at different consumption stages is essential for better management, particularly in the services area. Marketing communications, retail support, and follow-up activities might be better designed and engender higher levels of satisfaction. Public policy makers, continually searching for improvements in consumer information and complaint servicing, need to attend to the implications of this research as well. Many refinements prOposed in pricing, information presentation, and changes in the practices of various industries might be more effective if normative standards and evaluation phases were taken into account. Thus, consumer education projects should focus on the relationship between cues and benefits to find ways denhancingpeoples’ abilityto detect andusecuesinawayconsistent withexistinghabits, ratherthan attempting whole new habit formation W This line of research necessarilly touches on other lines of investigation. One of the major arguments in favor of continuing this approach is the potential for integating other finding into a more diagnostic and predictive model. Inquiries into information processing, perceptual mapping, and choice strategies all represent areas closely connected to the processes expectation] satisfaction models attempt to describe. Progammatic integration in research from these areas into model testing might produce much broader and better models of buying behavior. 114 To date, little attention has been directed at the process of evaluation in this conceptual context. Research supporting the existence of differing types of expectations, higher expectations than norms for known brands and offering, and generally neutral levels of satisfaction all indicate the need to examine information and attitude strategies employed by consumers. Qualitative research into consumption is underdeveloped in this area. While surveys and scales produce manipulable data, we really have very little insight to guide the development of these measures. In particular, attention should be given to differences in cue response, certitude, and value considerations by consumers with differing amounts and types of experiences. The same effort should be conducted to understand differences in evaluative processes throughout the time of consumption Heretofore, the dominant assumption has been that choice criteria are employed throughout the consumption process. This ignores the dynamism of consumption processes and leaves us unabie to explain how shifts in tastes and preferences arise, affect, and determine products and marketing mix strategies. In the current absence of theory and research in this area, qualitative studies seem called for. Related to that issue, the discussion about scale appropriateness and scale effects on model testing disguises a serious issue. The scales, including those in this study, are highly reliable tests for common elements across aggegated individuals. They are not predictive scales and, thus, may obscure some important issues. It is also implicit in the idea of product areas. Without knowing what people useasanorm,we reallycan’t addresstheissueofdistinctivechoiceaspectsofparticularofferings, much less the explanation for a total evaluation process. Attention to predictive elements may have to arise from more qualitative approaches followed by programmatic quantitative investigations. Here, to, the information and decision strategies employed may be significant but not captured adequately by current saliency measures. Finally, although this research looked at norms by asking respondents for an average performance rating in the product class (which is consistent with previous research), examination of the norms actually employed by consumers is important. It may even be the case that several norms are used with varying impact on satisfaction and future consumption behavior. A perception of high 115 quality/high price products 3 the norm for someone who cannot afford the 'standard' may produce much broader dissatisfaction then we has as yet addressed. We do need to examine individual factors such as self-confidence, sense of control, belief in actual ranges of choice, and situational factors as affectors of normative standards. APPENDICES APPENDDK A-l NO. OF SUBJECTS I 2N 117 APPENDICES N0.0FVARIABLESREADIN I 41 IJOGICALUNITI 1 NO. OF FORMAT CARDS I 1 UPIO VARS 0-41) FORMAT 0.1113(3) READ . VAR VAR (1X, 7F4.0. /. 2X. 11F4.0. /. 3X, 12F3.0, /, 4X, 11F3.0) FIRST OBSERVATION IN ANALYSIS 1 . 3.NON 2 I 3.00N0 3 I 3.00000 4 I 3.00000 5 - 3.000(1) 6 I 311701!) 7 I 3.000“) 8 a 23.000“) 9 I 25.00000 10 I 20.00000 11 I 2503000 12 a 15.000“) 13 I 12.00000 14 I 15.00000 15 I 12.00000 16 I 12.00000 18 I 2.1000(1) 19 a- 12.000“) N I 151“”) 21 I 150”“) 22 I 1511!!!) 23 I 1501)“) 24 I 12.000“) 25 I 15.00000 26 I 1611”“) 27 I 16W 28 I 3.00000 29 I 251!!!” 30 - 15.0N00 31 I 12.00“!) 32 I 15.00000 33 I 12.00000 34 I 1511”“) 35 I Z).00000 36 I 16.0““) 37 I 20.(I)000 38 I 20.(X)0N 39 I ZINOOO 40 I 200!!!) 41 I 25.00“!) CORRELATION MATRIX 118 APPENDDI A-2 FINAL OBSERVATION IN ANALYSIS VAR 1 I 5.00000 VAR 2 I 50!“) VAR 3 I 4.00000 VAR 4 I 4.00000 VAR 5 I 4.000(1) VAR 6 I 4.00“” VAR 7 I 3.000“) VAR 8 I 25.00000 VAR 9 I 25.000!) VAR 10 I 2501)“) VAR 11 I 2.5.000“) VAR 12 I 25.00000 VAR 13 I 20.000“) VAR 14 I 15.000“) VAR 15 I KIOOON VAR 16 I 25.00000 VAR 17 I 20.000“) VAR 18 I 25.00000 VAR 19 I 15.00000 VAR 20 I 15.00000 VAR 21 I 15.000“) VAR 22 I 15.00000 VAR 23 I NIKKI!) VAR 24 I 5.000» VAR 25 I 5.00011) VAR 26 I 25.00000 VAR 27 I 251“!» VAR 28 I 10.00000 VAR 29 I 25.00000 VAR 30 I 250!!!) VAR 31 I 20W VAR 32 I 15.00000 VAR 33 I 20.00000 VAR 34 I 151!!!» VAR 35 I 25.030“) VAR 36 I 15.00000 VAR 37 I 10.00“!) VAR 38 I 25.00000 VAR 39 I 20.00000 VAR 40 I 15.00000 VAR 41 I 25.00000 NO. OF OBS. IN ANALYSIS I 197 119 APPENDDK A-3 MEANS AND STANDARD DEVIATIONS 3.056 1.379 eeaseanxceeeeeeauceeeeeacaczcc=s°““a“sw- r: be 5‘. LII F. APPENDIX A-4 CORRELATIONS 15 13 I4 11 12 910 uaamnwmewcmeecmnwanmnununaeemnc unn7nmum4 27mew$nnnnxunmsnus7r7nom7nmnznorunmcmn9ao aannenea9aasmne9cummnnm2an mN4ntumnn9umoo4 enuncseecsnm7usnn nnnnasmsznan a2nnnxnmumma awnuuowennmgnawenmuanonunuum7xumnuucunmus “was sssnmns eucese2xnnzn nnnmnnnnnncnnnnz anenauosmnnss nmunnnuuaznnnn n9 cannun9nnnns nnaz enemas“nonexnnsnnnuunnum9nawnuuass744 «ennmamsosm4emmn4499snsnnuunnmumuumrnnnn4 «same menun9can“nnas9umuununs9nannuu9nuum9 an4 4nmamacu92am9 249294osnnutm4w42o744n2119 nunmamaanmuunewwun9stnonmunusus79n74n9924 sumnnnnacz wunmaas usu4unnmn7umumn99num99 18 15 11 6 1 nmunxas7n9mnn7 anaemmununnenmsuscnmu9 11 10 5 7 -3 ma ranaaunaaueaaus4 an un9n9nnn9m2m999n94 123456789M“nuuumnflwmflnnufi“flunflflnflfifixflfiflw“ H1 APPENDIX A-S CORREIATIOINB 19211212223247.5267] 2829” 17 18 16 9&18:Muwflfi Uflfiuflflunfl % H”%Wflfiflmflfifi M snfinnnfi7mwflm unnananannmnnnna«munxusuumnum 42mwma9mnnumunammumsuunummnnw nnaauunun mmuaun9uunuuunmgaz nunuuuusuun nnaaunn9nnu4 7mnan2uuusmunnnnu yuanunnxnaaumnxauanunwunwnnnu nnumumnanwnmmuunmnmuususauu99 smnnmnmmunnm«nasmnncunuwumuxnunnunn nuunannnumumaasmumannnnnnnwnnmmmnx9 mmxanmnxnnnmnnmnnmnunn nnuausmwununnumnwuun 6 umn2999uuxsumm gugmnanaxnsumaunuwmnmnmnu n99259nn2unm 7mm u7mmguaunnmmn9nsnuunnmmu o 5 x4 4u9u2unxmnun2nnmmum ammu4 zuuw4n4 s«mama7 n7 52 62 64 59 52 58 36 44 55 45 5 59 1m 65 $2 11 15 mmmnnfluaafinunmmnuvfiu unfluu4flmnl 7Nfinsn9fi9 9mn9n£$6wwfi7vn363§4n2 1nnu3ununm3n3usxnanw4 CORRELATIONS 31 32 3 3 5 6 10 11 6 7 -1 -2 8 15 12 10 8 19 6 17 12 21 aasasauraaaszusaa:msusszsaa:::=3°~gau‘w- 8338m288$§88§8§858688362d°°° H on “888863=8§88°3§833888$BS“S:”=8833 axaaaa:agsxxrssuaaassazassxaau 35 K 6-..- Nd 36::K3:gsasrxsxasaxsssuuussszuxmxasafl 833688§t=8838§3838388 88K8$§8$BBB§38§583836353“3“°‘5°“”é”°&A““$ R B$88§Sflfi§68§86338338333§835888 fiS8§888£883£8$$83888 no “~szu 3 6-0 H Hill. Gang“.— 86§$§38$BSSSSGt$35§B§8285$2”°8 $§888883838883635°868$flfi 8 gssaususa'zssssszw INPUT R—MATRIX 31 32 33 31 1m 39 36 32 39 IN 37 33 36 37 1m 34 30 41 41 35 31 27 41 36 25 23 31 37 so 25 28 38 27 I) 32 39 27 29 38 40 36 32 26 41 18 9 15 1 3 3 3 2 5 6 15 3 10 11 10 4 6 7 9 5 -1 -2 0 6 . 8 15 19 7 12 10 11 8 8 19 12 9 6 32 13 10 17 27 19 11 12 20 12 12 21 29 25 13 6 11 16 14 0 1 15 15 6 14 14 16 25 17 29 l7 l4 7 29 18 25 17 15 19 5 22 18 20 26 39 26 21 20 29 37 22 23 33 I) 23 28 I) 25 24 16 25 26 B 27 18 32 26 24 26 28 27 28 24 39 28 21 14 29 29 20 6 14 M 33 29 36 41 KRSRfiSfiKSfiSEBES 88§§8¢3388 B58§8888fi8553°5Q°°3°°“#“°&#““#88&88§88386 8 8&fi838338333fi383888 azmsaagasauus 3 1% APPENDIX A—‘l 39 40 41 1 2 3 27 36 18 3 5 10 29 32 9 3 ‘ 6 11 38 26 15 3 15 10 41 33 27 13 19 17 35 29 25 3 14 9 15 30 20 -4 3 3 28 29 23 11 18 18 53 29 32 10 15 11 1G) 48 37 5 11 10 48 IN 42 7 6 5 37 42 1(1) -3 1 -3 5 7 -3 81 75 11 6 1 1m 84 10 5 -3 84 IN 8 2 -4 83 87 1 l -3 76 74 11 10 -3 13 13 4 7 4 4 aasasuassssaasaasaazza°sz-=asasa§asatass 8 w a a a c .L. ”8‘353“5’58863!$S$%8$$$2$383§ SSEGfiSGiBBB B“83$335825538333338$%3dd 8“33338‘3°=°332233!83&83§ %&8885°638$3& N 3119 CONFIRMATORY MODEL (SINGLE DIMENSION) a°xaxsu°=°~°gmg3a:xxsssazagsaat-~sé~s°eG ‘ APPENDIX A-8 INPUT R—MATRIX 17 18 19 1314 1516 10 11 12 9 gnmunummunmun992sgnnnunm7wu7mmuuaunnmmu9n uuunuuuunxaxauunuauvmwnunanxmmummmmnnuuaa uvnunsnwumuemnunnnnnunnsumsmwvununuuauuwu unnnnmuxnnnnaafinasxusanwnnmsnnummuunnsaww suunnvumum4uaam$anawsws$smnmaummuausauam o2ummsmuaaoammsw$nunusunmsnunvmvnmwnnmn2u aumnuoumoo4annnans«xaanmnswsummunmnanmn4u nnnuunmum2aeuau2 sussnnmm«snnnnunxnnvanmnx umuuuanmmu2amwsunmunnmuasmanmusnnm zummvm nnmnnanmnu2 mw2vu929nmnnauss4snxn3nn nnnmm snuumsun um .assnxuasmnnnwnw«nnnuua2nnmnuax 2wnuu256244anaaannmsau4au smuununuuu2 numan umunm2nnnn4annnmamnunmuas su44995nsnn unnm aumuuamuuo :unm manunwsnsag sgumuuuunagn 22o244u2222“N4 umvaanuunaws nnn224maunu2m2n nflufififlflxw“123456789mnnnMEMURHNflflBflfiunuNfl 31 APPENDIX A-9 INPUT R-MATRIX 20 21 22 23 24 25 26 27 28 29 I) 31 26 20 23 28 16 27 24 28 21 20 33 32 39 29 33 x 25 18 26 24 14 6 29 33 26 37 1) 25 26 32 28 39 29 14 36 34 40 31 47 36 35 40 35 26 27 24 34 3 19 29 38 52 28 25 30 28 24 23 35 36 17 23 31 26 33 33 25 24 16 18 24 37 16 19 20 27 42 39 27 25 25 15 28 38 22 27 I) 37 26 21 29 35 19 20 31 39 20 28 23 24 15 17 33 48 15 32 37 40 29 20 25 23 9 17 22 25 22 33 38 41 11 13 9 21 3 14 15 27 17 59 31 1 18 6 l8 3 17 17 17 9 16 2 19 2 19 14 25 11 27 27 33 22 I!) 8 28 3 11 6 19 4 28 28 27 16 22 7 26 4 8 6 14 0 22 19 24 13 24 6 26 5 8 -4 10 -5 15 17 24 8 20 -l 20 6 9 11 19 14 24 25 34 22 29 9 29 7 9 5 17 6 18 17 28 14 22 13 19 8 26 22 28 18 24 18 22 14 16 3 21 9 54 26 43 21 29 23 31 19 17 3 5 10 36 37 38 27 31 25 31 27 18 10 31 11 5 17 49 19 24 25 26 15 10 7 I) 12 28 32 38 51 30 29 42 33 31 13 39 13 19 15 29 20 52 38 31 16 27 4 28 14 16 7 23 10 40 47 27 16 23 2 25 15 20 19 26 28 34 35 66 45 45 16 32 16 15 19 20 23 19 22 53 65 45 30 40 17 12 17 24 23 28 28 46 41 74 N 33 18 15 18 16 19 10 21 32 38 28 61 43 19 54 46 45 44 33 32 30 19 23 9 21 20 1m 57 6 47 43 36 31 33 24 16 45 21 57 IN 56 48 39 31 33 36 27 23 33 22 65 56 1m 58 50 42 37 33 29 17 37 23 47 48 58 1G) 51 43 45 44 37 25 40 24 43 39 50 51 1m 66 41 28 42 16 37 25 36 31 42 43 66 1m 47 36 42 24 43 26 31 33 37 45 41 47 IN 71 54 32 52 27 33 36 33 44 28 36 71 1m 55 45 52 28 24 27 29 37 42 42 54 55 1m 3) 45 29 16 23 17 25 16 24 32 45 1) 1m 44 w 45 33 37 40 37 43 52 52 45 44 1M APPENDIX A-10 MULTIPLE GROUPS PROGRAM PCA (31-41) SPE 0-11) DIS (12-22) SAT (23-28) M (29-30) STANDARD SCORE COEFFICIENT ALPHAS 84. 95. 5. 84. 61. lfl APPENDD( A-ll nmmmmflmmmmmxm»snmummmcmumx anmwwunvmrmnmumwu 3 2 n 33 fl 5 3 M a 3 M 0 1 3 u U 4nuflflfimnnnflfififl6w$$3986uonwunu6fiuwwfiu 679 74 82 mumm93nnm5 JDMMU7NNQ$6wfinN“a5fl9n6m4uflnmn7ummnwu 56fiwu3mfiu6 1:”“3 fifi7fl9$fifl6 “fiflflfl”flfiflflflflflfl8flfl”flflfl 3333341157 :W ”afiaflflaflafia6fiflflfifiu6fl3nn79m2m8“ 96 191Nfimflnfl284184334442484042uumufl9n3flfififl7$flfldwufl Mnfiflfiwflfiflfl47652l mn4BNH103on9§nfimfifl9nnfin33$mwfifi Nflflfl§§fl93 «N51181ufl7unmw63Mflfifilmfluufi7flflflnflfl3fifl5 nmflu3wflw3 2&0”H82un6nflmumnwfimfiflnflwNNflfifimmflfluflfl$ wfiflfinfiwflufifil8 “WHHUSuuumfim“Enummwwfl03N335ufifl5fin 333wx3fiwfimm433.4914366H967M8MBHB3uflfifiumuufi2fifln 1Nflflfixnfl$333u974umnmflmxflmB3N1fi:2flfiflflflfl§w6$“0 m““6Mwfiuflflnnw”n7fiNfluflflfluDUnfinfiwn7fifiwfififlfluflflfifl ”Myflflflnnflfifi3flm UHuBHnBMfiM” ”8 ”flmfiflflnflfl4fifl4fi3fl 3%”“NBfimflfi936H74§munnm3u1“n7nnyflfiwfi8fifil6 513”“ ”yxm3fim”N31351618n86nnn60634fififiwfiu6NflflflOBfll 3: nnausxnnxww223456229mnuuuumnu2 9m2nnuuununmwmmmm APPENDIX A-12 FACTOR INTERCORREIATIONS AND LOADING MATRIX COMMUNALITY IN THE DIAGONAL 19 17 18 1516 13 14 6 10 11 12 17 12 21 9 6 1932272029 8 8 12 7 17 13 16 18 14 10 12 9 259nn25317117llgfifiuflfiw9u9flfl7flun ”flufifimnMflflflEBOfimmwmflnflflflflufiflfifl fl9ufl4n7 ”Mw633721uflflflfi4flmfl3fl2 flxflfififlfiflflwfiflufimmflunflfifimwfiflflfifl flflfiflfifififlflwawmmfiuufififi6mnnnfiflx flnfifififlfifi4 “”7M7BmflnflM32finflfiyfl “$Q6 Sflnfi$$fimflfifimflflflmn4flmfinfifl saga“: 333 4$fluw7w§5 7nflunnflwvannnynanfl$ dawwfiflwwwnnufiéxwwnwusnwwufiuum7wuflflnn 2&$2 flfi$n$n$nflaufififlfinxflnnmfinflmmmnfl%$m lafifinfiuufiflnflfiwswuflvnfluflnflnl9n3 finnnuu 5674437 «aauwufiwwaan4 xuununumu8n4 “Snufiflflm 17 22 18 7 1429 2915 14 25 15 18 19 26 15 29 19 35 22 11 48 43 36 15 60 42 16 1722 252225 7 1325 2718 8 1825 2224 1635 14 41 10 22 6 25 1417 o 16 15 13 19 6 .3 18 10 16 12 3 1620 6 171116 0 -4 0 68 79 82 77 68 ll 1 13 19 12 25 18 23 18 36 3 6 6 12 9 11 13 12 20 26 20 13 18 18 26 12 21 13 1524232428 11 -1 12 10 11 17 16 17 12 13 4afinn6aw$uflm3$7 4&995n6mnu4 nnmmflflfiu 1 nunmaafinwmaaauflflflnfiwanx59nwuufiunv9vaflxu 8 15 19 15 14 9 284méunu8 N4N2w 60 59 52 47 31 . Wufiuflflnflfiuna -1 -2 0 7 4 4 11 2 1 1 3 66 mm nnuu sununmmmmmm 12 13 14 15 16 17 18 19 nfiufixnwi4123456789mu 31 AnamMAn wmumnnmmmmmxmnammummmcmumx ammmwunvmtmummmwu 22 E u 3 fi‘fl fl 3 m:m.xnw3auas fi sungnnnynammnuuunumuwnmnnxnmnnuannmwengmmnmmmm gnaw«nannyummman“gnuanaanaswxunnaeumnamaamnumm munuwusnnmmannauaamnuannwxnaanmnamasanxnswumum mnunmzuunmquwmumnnwnnuamunanunnnnumuunuannmunm nmnmwssawwasmmuzmmannuumnnunuununnuansnammm an anxuguumnammuuumnmnsnwwuunmaenannwnonnuunssumm m6uuflufimn£$287649n33171426Nm691 BU3wunfiwn“08flfl nunnumumunnmmnumnnmnmmnnnaauununnnaausamanuamx uunuuuusaung mnsnuumnunmmasqumaxauuunnsunxnaan unusmnnnnnmnnnuuuuunm umnanmnwnnnannwnunnnumna nunwunxnnnunnuwn sunansnuznxn zamnnxaunuuswm “Eufiuflnum93nnunuuSMNBumnwumumfiawwfluwflunmflnwfiwx uwfifinxnflunn3u4osMGlznmflmouunwunaSwflQGMNfiwfluflaw l 33m”3mmw339mfiwumwnunflnflvfiumflM6$$£$w0N3$flflwflflfl mfifl3flfll 23NH61664lsnufl7nu7mmnmfi7 nfiax3fixflflflu77 ”2 uxfiwwflmnm”ummn8899uflxfiflwm unufififlfiflfixmfluM6flflwn6 nnnuxxnxmwulzs456739mmnnuumnmmmnnnusunuvmmmmmm REORDERED R-MATRIX 501 502 503 501 100 21 - 59 502 21 100 86 503 \ 59 86 100 504 67 37 86 505 73 30 67 APPENDDK A- 14 131 APPENDD( A-lS PATH ANALYSIS, ORDINARY LEAST SQUARES THE NUMBER OF EXOGENOUS VARIABLES IS 2 PATH COEFFICIENTS, DECIMALS OMITTED 501 502 503 504 505 506 507 508 §§§§§§§§ OOOgOOOO OOEOOOOO OBOOOOOO OSOOOOOO §OOOOOOO OOOOOOOO REPRODUCED CORRELATIONS, DECIMAIS OMITTED 501 502 503 504 505 506 507 508 501 100 76 19 9 24 33 60 47 502 76 100 14 12 19 43 50 39 503 19 14 100 2 -76 6 -5 -4 504 9 12 2 100 2 38 16 13 505 24 19 -76 2 100 8 34 27 506 33 43 6 38 8 100 46 36 507 60 50 -5 . 16 34 46 100 78 508 47 39 -4 13 27 36 78 100 501 502 503 504 505 506 507 508 501 0 0 0 10 0 32 10 9 502 0 0 -3 0 3 0 1 50 503 0 -3 0 -72 0 14 37 28 504 10 0 -72 0 79 0 -8 -14 505 0 3 0 79 0 17 5 5 506 32 0 14 0 17 0 19 3 507 10 1 37 -8 5 19 0 0 508 9 50 28 -14 5 3 0 0 THE SUM OF SQUARED DEVIATIONS IS 1.84 MODEL Al §§§§§ §§§§§ BOOOO aaags § 502 OOdOC 132 APPENDDK A-16 PATH ANALYSIS, ORDINARY LEAST SQUARES THE NUMBER OF EXOGENOUS VARIABLES IS 1 PATH COEFFICIENT, DECIMAIS OMITTED 503 504 505 0 0 0 0 0 0 0 0 0 71 0 0 0 80 0 REPRODUCED CORRELATIOm, DECIMALS OMITTED 503 504 505 59 67 53 86 66 53 100 86 68 86 100 80 68 80 100 OBSERVED MINUS PREDICI'ED CORRELATIONS 503 504 505 0 0 20 0 -29 -23 0 0 -1 0 0 0 -1 0 0 THE SUM OF SQUARED DEVIATIONS IS .18 MODEL A2 §§§§§§§§ §§§§§§§§ §§§§§§§§ 501 H ogoooooo BRE8B°°° 8 N APPENDIX A~17 PATH ANALYSIS, ORDINARY LEAST SQUARES THE NUMBER OF EXOGENOUS VARIABLES IS 2 PATH COEFFICIENTS, DECIMAIS OMITTED 503 504 505 506 507 508 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ~76 0 0 0 0 0 0 38 0 0 0 0 0 0 20 30 0 0 0 0 0 0 78 0 RELATIONS, DECIMAIS OMITTED 503 504 505 506 507 508 19 9 ~14 4 44 34 14 12 ~11 5 32 25 100 2 ~76 1 ~7 ~5 2 100 ~1 38 15 12 ~76 ~1 100 ~l 14 11 1 38 ~1 100 32 15 ~7 15 14 32 100 78 ~5 12 11 25 78 100 OBSERVED MINUS PREDICTED CORRELATIONS 503 504 505 506 507 508 0 10 39 61 26 22 ~3 0 33 38 18 64 0 ~72 0 19 38 29 ~72 0 82 0 ~7 ~14 0 82 0 25 25 21 19 0 25 0 33 14 38 -7 25 33 0 0 29 ~14 21 14 0 0 THE SUM OF SQUARED DEVIATIONS 133.13 MODEL B §§§§§§§§ §§§§§§§§ 8128132” §§§§§§§§ § 501 100 76 19 APPENDDK A~l8 PATH ANALYSIS, ORDINARY LEAST SQUARES THE NUMBER OF EXOGENOUS VARIABLES IS 2 PATH COEFFICIENTS, DECIMAIS OMITTED 502 503 0 0 0 0 0 0 12 0 0 ~83 39 0 0 0 0 0 504 OOKOOOOO 505 506 507 508 OgOOOOOO O$OOOOOO §OOOOO°O OOCOOOOO REPRODUCED CORRELATIONS, DECIMAIS OMITTED 502 503 76 19 100 14 14 100 12 2 19 ~76 43 6 30 ~15 23 ~12 amg~§~n° g 505 506 507 508 24 33 25 20 19 43 30 23 ~76 6 ~15 ~12 2 38 23 18 1(1) 8 29 23 8 100 61 48 29 61 100 78 23 48 78 100 OBSERVED MINUS PREDICTED CORRELATIONS 502 503 47 O 504 ~15 J) 505 506 507 508 0 32 45 36 3 0 Z) 65 0 14 47 36 79 0 ~15 ~20 0 17 10 9 l7 0 4 -9 10 4 0 0 9 ~9 0 0 THE SUM OF SQUARED DEVIATIONS IS 2.53 MODEL C §§§§§§§§ §§§§§§§§ §§§§§§§§ § 00666669 502 OOOOEOOO APPENDD( A~19 PATH ANALYSIS, ORDINARY LEAST SQUARES THE NUMBER OF EXOGENOUS VARIABLES IS 2 PATH COEFFICIENTS, DECIMALS OMITTED 503 504 505 506 507 508 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ~76 0 0 0 0 0 0 38 0 0 0 0 0 0 24 59 0 0 0 0 0 0 78 0 REPRODUCED CORRELATIONS, DECIMAIS OMITTED 502 502 888883°6~°° 503504505 19 9 ~14 4 ~1 ~1 14 12 ~11 5 0 0 100 2 ~76 1 ~18 ~14 2 100 ~1 38 22 17 ~76 ~1 100 ~1 24 19 l 38 ~l 100 ‘ 59 46 ~18 22 24 59 100 78 ~14 17 19 46 78 100 OBSERVED MINUS PREDICTED CORRELATIOINS 503 504 505 506 507 508 0 10 39 61 71 57 ~3 0 33 38 50 89 0 ~72 0 19 50 38 ~72 0 82 0 ~14 ~19 0 82 0 2.5 15 13 19 0 25 . 0 '6 ~7 50 ~14 15 6 0 0 38 ~19 13 ~7 0 0 THE SUM OF SQUARED DEVIATIONS 184.45 MODEL D 136 APPENDD( B~-1 Part 1. 1. Teachers are prepared and easy to take notes from. 5 4 3 2 1 Agree Disagree (5) 5 4 3 2 1 Highly Important Not Important (6) 2. Teachers enjoy teaching. 5 4 3 2 1 Agree Disagree (7 ) 5 4 3 2 1 Highly Important Not Important (8) 3. Teachers respect and understand students. 5 4 3 2 1 Agree Disagree (9) 5 4 3 2 1 Highly Important Not Important (10) 4. Teachers are enthusiastic. 5 4 3 2 1 Agree Disagree (11) 54321 5. Teachers stimulate thought and hard work. 5 4 3 2 1 Agree Disagree (13) s 4 3 2‘ 1 Highly Important Not Important (14) 137 APPENDDK B-2 Part 1 ( Cont.) 5. 10. 11. The classes are enjoyable. 5 4 3 2 1 Agree Disagree (15) 5 4 3 2 I Highly Important Not Important (16) ' The material is relevant to my life. 5 4 3 2 1 Agree Disagree (17) 5 4 3 2 1 Highly Important Not Important (18) Classesare not boring. S 4 3 2 1 Agree Disgree (19) s 4 3 2 1 Highly Important Not Important (20) Examinations are {air and cover expected material. 5 4 3 2 1 Agree Disagree (21) 5 4 3 2 1 Highly Important Not Important (22) Grades are fairly assigned. 5 4 3 2 1 Agree Disagee (23) 5 4 3 2 1 Highly Important Not Important (24) Give competence in the discipline. 5 4 3 2 1 Agree Disagree (25) 5 4 3 2 1 Highly Important Not Important (26) APPENDDI B~~3 Part 1 (Cont) 12. 14. Give competence in business. 5 4 3 2 1 Agree Disagree (27) 5 4 3 2 l Highlylmportant NotImportant (28) Improve understanding of my major. 5 4 3 2 1 Agree Disagree (29) S 4 3 2 1 Highly Important Not Important (30) See progress toward graduation. 5 4 3 2 1 Agree Disagree (31) 5 4 3 2 1 Highly Important Not Important (32) Lead to understanding of how people and institutions behave. 5 4 3 2 1 Agree Disagree (3) 5 4 3 2 1 Highly Important Not Important (34) 5 4 3 2 1 139 APPENDIX B~~4 Part I]. Now please indicate the extent to which you believe that each statement will be true of the class youarenowin. Thenhowimportant to your evaluation oftheclass. Example: WillbetaughtbytheKingofRuratania. 5 4 3 2 1 Very Likely Very Unlikely 5 4 3 2 1 Highly Important Not Important Part II. 1. Teacher(s) will be prepared and easy to take notes from. S 4 3 2 1 Very Likely Very Unlikely (36) 5 4 3 2 1 Highly Important Not Important (37) 2. Teacher(s) will enjoy teaching 5 4 3 2 1 . Very Likely Very Unlikely (38) 5 4 3 2 1 Highlylmportant NotImportant (39) 3. Teacher(s) will respect and understand students. 5 4 3 2 1 Very Likely Very Unlikely (40) S 4 3 2 1 High Important Not Important (41) 4. Teacher(s) will be enthusiastic. 5 4 3 2 1 Very Likely Very Unlikely (42) 5 4 3 2 1 Highly Important Not Important (43) 140 APPENDIX B~~5 Part II (Cont) 5. 10. Teacher(s) will stimulate thought and hard work. 5 ' 4 3 2 1 Very Likely Very Unlikely (44) 5 4 3 2 1 Highly Important Not Important (45) The class will be enjoyable. 5 4 3 2 1 Very Likely Very Unlikely (46) S 4 3 2 1 Highly Important Not Important (47) The material will be relevant to my life. 5 4 3 2 1 . Very Likely Very Unlikely (48) 5 4 3 2 1 Highly Important Not Important (49) Theclasswillnotbeboring. 5 4 3 2 1 Very Likely Very Unlikely (50) 5 4 3 2 1 Highly Important Not Important (51) Examinations will be fair and cover expected material. 5 4 3 2 1 Very Likely Very Unlikely (52) s 4 3 2 1 Highly Important Not Important (53) Grades will be assigned fairly. 5 4 ' 3 2 1 Very Likely Very Unlikely (54) 5 4 3 2 1 Highly Important Not Important (55) 141 APPENDD( B~~6 Part II (Cont) 11. 14. The class will make me more competent in marketing. 5 4 3 2 1 Very Likely Very Unlikely (56) 5 4 3 2 1 Highly Important Not Important (S7) The class will make me more competent in business. 5 4 3 2 1 Very Likely Very Unlikely (58) 5 4 3 2 1 Highly Important Not Important (59) The clms will improve my understanding of my major. 5 4 3 2 1 ' Very Likely Very Unlikely (60) 5 4 3 2 1 Highly Important Not Important (61) The class will help my progress toward graduation. 5 4 3 2 1 Very Likely Very Unlikely (62) 5 4 3 2 1 Highly Important Not Important (63) The class will lead to an understanding ofhow people and institutions behave. 5 4 3 2 1 Very Likely Very Unlikely (64) 5 4 3 2 1 Highly Important Not Important (6S) 142 APPENDIX B~7 EVALUATION PROJECT 11-1 STUDENT NUMBER Please fill in your student number. It is for data collection purposes only and will not be matched to your responses nor be available to your instructors. 12mm: Circle the appropriate number. 1. For each of the following statements, indicate the degree to which you agree/disagree that the statement is m of this class. We: Was taught by the King of Ruratania. 5 4 3 l 1 Agree Disagree Part I. 1. Teacher(s) was prepared and easy to take notes from. 5 4 3 2 1 Very likely Very Unlikely (1) 2. Teacher(s) enjoyed teaching. 5 4 3 2 1 Very Likely Very Unlikely (2) 3. Teacher(s) respected and understood students. 5 4 3 2 1 Very Likely Very Unlikely (3) 4. Teacher(s) was enthusiastic. 5 4 3 2 1 Very Likely Very Unlikely (4) 10. 11. 143 APPENDIX B~8 Teacher(s) stimulated thought and hard work. 5 4 3 2 1 Very likely Very Unlikely (5) The clss was enjoyable. S 4 3 2 1 Very likely Very Unlikely (6) The material was relevant to my life. 5 4 3 2 1 Very Likely Very Unlikely (7) The class was not boring. 5 4 3 2 1 Very Likely Very Unlikely (8) Examinatiom were fair and covered expected material. 5 4 3 2 1 Very Likely Very Unlikely (9) Grades assigned fairly. 5 4 3 2 1 Very Likely Very Unlikely (10) The class made me more competent in marketing. 5 4 3 2 1 Very Likely Very Unlikely (11) 14. 144 APPENDDI B~9 The class made me more competent in business. 5 4 3 2 1 Very likely Very Unlikely (12) The class improved my understanding of my major. 5 4 3 2 1 Very Likely Very Unlikely (13) The class helped my progress toward graduation. S 4 3 2 1 Very Likely Very Unlikely (14) The class lead to an understanding of how people and institutions behave. 5 4. 3 2 1 Very Likely Very Unlikely (15) Part 11. Please circle the number which best corresponds to your belief: 1. To what extent was this class what you expected it to be like? 5 4 3 2 1 Very Similar Very Dissimilar (16) To what extent was this class similar to what you had heard about it? 5 4 3 2 1 Very Similar Very Dissimilar (17). To what extent do you feel satisfied with this class. 5 4 3 2 1 Very Satisfied Very Dissatisfied (18) How satisfied are you with this class compared to other courses you have had? 5 4 3 2 1 Very Satisfied Very Dissatisfied (19) 10. 11. 145 APPENDD( B~10 To what extent do you feel satisfied that you what you received was worth what you paid for the elm? S 4 3 2 1 Very Satisfied Very Dissatisfied (20) How similar was the level of effort required in this course to the level you expected? 5 4 3 2 1 Very Similar Very Dissimilar (21) Relative to your own effort, how satisfied are you with your grade in this course? 5 4 3 2 1 Very Satisfied Very Dissatisfied (22) To what extent are you satisfied with what you learned in this class? 5 4 3 2 1 Very Satisfied Very Dissatisfied (23) Place yourself on the ladder: Probably the bes class I could expect. (24) Probably the worst class I could expect. Would you take another course from this instructor? 5 4 3 2 1 Definitely Yes Definitely No (25) Would you recommend this course to a friend? 5 4 3 2 1 Definitely Yes Definitely No (26) Part III Please indicate the number corresponding to the correct response. 1. 9‘9"?!” Major: 1) Marketing/ISM 2) Finance 3) Management/MLM 4) Accounting 5) My major is in the next question. (27) Major: 1) Packaging 2) General Business 3) Advertising/Journalism/Telecomm. 4) Retailing 5) Other (28) Sex: 1) Female 2) Male (29) GPA: 1) 1~2.0 2) 2.1-2.5 3) 26-30 4) 3.1-3.5 5) 3.6-4.1 (30) Number of previous business courses: 1) 0 2) 1~2 3) 3-4 4) 5 or more (31) This class was Mpaid for by: 1) parents 2) self 3) scholarship 4) loan 5) other (32) REFERENCES 147 BIBLIOGRAPHY Anderson, R.E. "Consumer Dissatisfaction: The Effect of Disconfirmed Expectancy on Perceived Product Performance,“ MW Vol. X (February, 1973), 38-44. Anderson, R.E., Choi, KS” & Boyles, G. V. 'Disconfirmed Student Expectations and Instructor 7 Evaluatiom: An Emoiriool Study.‘ MW (1973). 51% Anderson, R.E., & Hair, J.F., Jr. ”Consumerism, Consumer Expectations, and Perceived Product Performance," ' d . Emu, M. Venkatesan, editor (1972), 67—79. Andreasen, AR. “A Taxonomy of Consumer Satisfaction/Dissatisfaction Measures,“ W193 - - . . 2. 3 2.1.0 t. MSI, H. Keith Hunt, editor (1977),11-35. Atkinson, J. W., Bastian, JR, Earl, R.W., & Litwin, G.E. “The Achievement Motive, Goal Setting, and Probability Preferences” WM. 1960 60 27-36 Atkinson, J.W. 'Motivational Determinants of Risk-taking Behavior.“ W 1957, 64, 359-372. Alien. 1. & Fishbein. M. Wat. Enslewood Cliffs, NJ: Prentice Hall, 1980. Basozzi. RJ’. WW “may. 1980 (New York). Bandura, A. W. Englewood Cliffs, NJ .: Prentice-Hall, 1977. Bearden, W.0., Durand, R. M., Mason, LB. & Teel, J.E., Jr., "Dimensions of Consumer Satisfaction/Dissatisfaction with Services: The Cme of Electric Utilities," W W (Fall. 1978). V016. No 4. 278-290 Bearden, W. 0. & Teel, J. E. “Selected Determinants of Consumer Satisfaction and Complaint Reports," WWW February 1983. 21-28 Bentler, PM. & Bennett, D. G. ”Significance Tests and Goodness-of-Fit In the Analysis of Covarience Structure.” WWII. 88. 1980 588-606 Bertsch, T. 8: Peek, L. ”Determination of Measurement Sales for Revising or Developing Teacher Evaluation Instruments.“ MW Spring 1987- Bettman, JR. ”A Flmctional Analysis of the Role of Overall Evaluation of Alternatives in Choice Processes.“ WW. 1981. JMitchell. at (Ann Arbor. MI) ' ’. Reading, Mass: Addison-Wesley, University of North Carolina Press Blaloch. HM. Wages Aldine 1971 (Chimgo). 148 Bonnett, D.G. & Bentler, P.M. "Goodness-of-fit Procedures for the Evaluation and Selection of Losrlinear Models." Walla. 1983. 93. 149-166- Burt, R.f. "Confirmatory Factor-analytic Structures and the Theory Construction Process," W Wasatch. 1973. 2. 131-190 Burt, RS. 'Interpretational Confounding of Unobserved Variables In Structural Equation Models," Wm Aug 1976. Vol 5. 3-52- Cadotte, E., Woodruff, R. & Jenkin, R. "Norms and Expectation Predictions: How Different are the Measures?" WW 1982. Day and Hunt. eds (Knoxville. TN) Churchill, GA. MW Dryden 1983 (New York). Churchill, G.A. & Suprevant, C. "An Investigation Into the Determinants of Customer Satisfaction," Wm November 1982. 491-504 Connolly, J. "Some Conceptual and Methodolical Issues in Expectancy Models of Work Performance Motivation" W Oct. 1976. 37-47. Crandall, V.C. "Sex Differences in Expectancy of Intellectual and Academic Reinforcement." In C.P. Smith (Ed) W New Yorlc Russell Sage Foundation. 1969. Crandall, V. C., & McGhee, P. E. "Expectancy of Reinforcement and Academic Competence." m affirm 1968. 36. 635-648 Darden, WR., & Rao, C.P. "Satisfaction with Repairs under Warranty and Perceived Importanceo of Warranties for Appliances," . 1 Proceedings of the 2nd annual CS/D&CB conference, April 20-22, 1977, Ralph L. Day, editor (1977), 167~l70. Day, R.L. "Consumer Satisfactions/Dissatisfactions with Services and Intangible Products," Paper presented to Marketing Research Seminar, Institut d’Administration des Entreprises, University d’Aix Marseille, France (1975), 89-107. Day R.L. "Extending the Concept of Consumer Satisfaction," WWW Vol. IV Proceedings of the 1976 annual ACR conference, William D. Perreault, Jr., editor (1977), m Day, R.L. "How Satisfactory 13 Research on Consumer Satisfaction?" WM]; Vol. VII, Proceedings of the 1979 annual ACR conference, Jerry C. Olson, editor (1980), 5857. Dillon, WR. & Goldstein, M. W John Wiley, 1984, New York. Engel, J., Blackwell, R. 8: Miniand, P. W Dryden, 1986 (Chicago). Feather, N.T. W Lawrence Erlbaum Associates, 1982 (Hillsdale, NJ .). Feather, N.T. "Subjective Probability and Decision Under Uncertainty." WM 1959, 66, 150-164.(a) Feather, N.T. "Success Probability and Choice Behavior." MW 1959, 58, 257~266. (b) Feather, N.T. "Cognitive Dissonance, Sendtivity and Evaluation." MW 2m 1963. 66. 157-163.(a) 149 Feather, N.T. The Effect of Differential Failure on Expectation of Success, Reported Anxiety, and pronsc Unoertainty-" Mmalnlfierscnalim 1963. 31. 289-312 (b) Feather, N. T. "Mowrer’s Revised Two-factor Theory and the Motive-expectancy- Value Model." WW 1963. 70. 500-515 (C) Feather, N.T. "Performance at a Difficult Task in Relation to Initial Expectation of Success, Test Anxiety, and Need Achievement." MW 1965, 33, 200-217. (a) Feather, N. T. "Effects of Prior Success and Failure on Expectations of Success and Subsequent Performance" MW 1966. 3. 287-298 Feather, N.T., & Simon, J .G. "Fear of Success and Causal Attribution for Outcome." W muslin. 1973. 41. 525-542. Fishbein, M., 82 Ajzen, I. "Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research." Reading. Mass.: Addison-Wesley, 1975. Fornell, C. "Issues in the application of covarience structure analysis: A covarient," W W 1983. 9. 443-448. Gerlsing, B. & Hunter, J. "The return to multiple groups: Analysis and critique of confirmatory factor analysis with LISREL." unpublished paper, 1980, (East Lansing, MI). Gilly, M. C., Cron W. L. & Barry, LE. 'The Expectatiom- Performance Comparison Process: An Investigation of Expectation Types." WSW 1982. Day and Hunt, eds. (Knoxville, TN). Handbook of Learning & Cognitive Processes, Vol. 3. Edited: W.K. Estes Lawrence Erlbaum Associate (Hillsdale, NJ .), 1976. Heshhauser, H. "Achievement motivation and its constructs: A cognitive model," MM Emotism. 1977. 1. 283-329- Howard, J. & Sheth, J.N. W. New York: Wiley, 1969. Hunt, H.K. "CS/D—Overview and Future Research Directions," Werner]; ‘ , . of the lst annual CS/D&CB conference, Hunter, J.E. & Cohen, S.H. "Pachage" , - tt - I, 1969, 29, 697-700. Hunter, J .E and Gerbing, D. 'Unidimensional Measurement, Second Order Factor Analysis, and Causal Models." WW Vol 4.1982. 267-320 Hunter, 13., Gerbing, D., Cohen, S. & Nicol, T. M Baylor University 1980 (Waco, Texas). Joresborg, KG. &Sorbam, D. ' I W 1981, International Educational Services (Chicago). Kelley, H.H. "Attn'bution Theory' m Social Psychology." In D. Levine (Ed), Wm Mmjyafign (V0115). lincoln: University of Nebrmka Press, 1967. 150 Kennedy, J. & Thirbill, P. "Consumer Satisfaction as a Function of Search, Experience, Individual Dillerences and Circumstances of Use." Wu. 1982. Day and Hunt, eds. (Knoxville, TN). Kenny. DA. macaw Wiley. 1979 (New York) Kotler, P. W Simon and Schuster 1983 (Englewood Cliffs, NJ .). LaTour, SA. & Peat, N. C. "An Experimental Investigation of Responses to Service Quality Inequity," Working Paper #15, Center for Consumer Research, College of Business Administration, University of Florida January, 1979. LaTour, S.A. & Peat, N. C. "Conceptual and Methodological Issues In Consumer Satisfaction Research," W Vol. V1. Proceed- iner of the 1979 annual ACR conference. William L. Wilkie, editor (1979), 431-437. LaTour S A. & Peat N. C. "Determinants of Consumer Satisfaction: A Field Experiment" Proceeding; t l 'Mosi ii°J?-_!!;tlmll .1110 o :ellennl')‘ 0024-..i1Am'rt'Ll September 1~5, 1979, New York City, Ivan Ross, editor (1979), 83-85. LaTour, S.A. & Peat, N.C. "Effects of Prior Experience and Quality of the Best Alternative on Comumer Satisfaction, Complaints, and Purchase Termination Responses to Quality Decline," Working Paper Series #5, Center for Consumer Research College of Business Administration, University of Florida 1978. Levitt, T.H. "Marketing Intangiable Products and Product Intangibles," Wag, Jan/Feb. 1980. liechty, M. G. & Churchill, G.A., Jr. "Conceptual Imights into Consumer Satisfaction with Services," Wm AMA. Neil Beekwith. et at. editors (1979). 509-515 Lutz, RJ. "Changing Brand Attitudes Through Modification of Cognitive Structure." W W March 1978. 4959- Lutz, R.J. "An Experimental Investigation of Causal Relations Among Cognitions, Affect, and Behavioral Intention," W Vol. 3, March 1977, 197-207. Lutz, R.J. "The Role of Attitude Theory in Marketing," WM 1979. Lutz, RJ. "The Role of Attitude Theory in Marketing," in Kassorjian and Robermon, eds. Scott, Lovesman and Company, 1981, (Glenview, Ill). Lutz, RJ. & Bettman, JR. "Multiattribute Models In Marketing: A Bicentennial Review," in MI W Woodside, Sheth, and Bennett, eds. North Holland, 1977 (New York). Maddox, R .N. "Consumers" Satisfaction with Supermarkets: A Factor Analytic Study," We: a - of the 2nd annual CS/D&CB conference, April N22, 1977, Ralph L.Day, editor (1977), 163-166. Maddox, R. N. "The Structure of Comumers’ Satisfaction: Cross-Product Comparisons," W WSW Vol 10. No 1 (Winter. 1982). 37-53 151 Miller, GR., & Rokeach, M. "Individual Differences and Tolerance for Inconsistency." In R. Abelson, E. Aronson, W. McGuire, T. Newcomb, M. Rosenberg & P. Tannenbaum (Eds. ), W W Chicago: Rand McNally. 1969 Miller, IA. "Consumer Satisfaction/Dissatisfaction: Findings-Some Comments," W ._ ._ _ Proceeding of the 3rd annual CS/D&CB conference, October 57,1978, Ralph 1.. Day and H. Keith Hunt, editors (1979), 52-55. Miller, LA. "Exploring Some Alternative Measures of Consumer Satisfaction,, Mgflgetm’ g; mg 1225 1976 Educators" Conference Proceeding, AMA, Kenneth L. Bernhardt, editor (1976), 661-664. Miller, LA. "Studying Satisfaction, Modifying Models, Eliciting Expectations, Posing Problems, and Making Meaningful Measurements," . . ._ - ‘ Wm, Proceeding of the lst annual CS/D8ICB conference, MSI, April 11~13, 1976, H. Keith Hunt, editor, (1977), 72-91. Mitchell, TR. "EXpectancy Models of Job Satisfaction, Occupational Preference and Efforts: A Theoretical, Methodological and Empirical Appraisal," WW Vol. 81, No. 12, (1974), 1053-1077. Moore, EM. 8: Shuptrine, F. K. "Diseonfirmation Effects on Consumer Satisfact~' mo and Decision Making Processes" W 1984. 299-304 Nebrasks Symposium on Motivation, 1980. "Cognitive Processes," John H. Flowers, ed. University of Nebraska Press, 1981. lincoln, Nebraska. Oliver, R.L. "A Theoretical Reinterpretation of Expectation and Disconfirmation. Effects on Postexposure Product Evaluations: Experience In the Field," e a i "fa WW5 Proceeding of the 2nd annual CS/D&CB conference, April 20-22, 1977, Ralph L. Day, editor (1977) 2~9. Oliver, R.L. "An Opponent Process Model of Consumer Satisfaction/Dissatisfact- ion Dynamics," ik’.‘11‘l*°.,e.!. 3!-‘.J‘r"1;.”§!'41L.,AA!.MI'll.2W!" w September 1~5, 1979, New York City, Ivan Ross, editor (1979), 81-82. Oliver, R.L. "Conceptualization and Measurement of Disconfirmation Perceptiom in the Prediction of Consumer Satisfaction." W W Proceeding of the 4th annual CS/D&CB conference, October 3-5, 1979, H. Keith Hunt and Ralph L.Day, editors (1980), 2-6. Oliver, R.L. "Measurement and Evaluation of Satisfaction Processes in Retail Settings,' " 193mg! of Retailing. Vol. 57, No. 3 (Fall, 1981), 25-48. Oliver, R.L. "Product Satisfaction as a Function of Prior Expectation and Subsequent Disconfirmation: of the 3rd annual CS/D&CB conference, October 5-7, 1978, Ralph L. Day and H. Keith Hunt, editors (1979), 66-71. Oliver, R.L. "Theoretical Bases of Comumer Satisfaction Research: Review, Critique, and Future Direction." WMWMW Lamb. Jr andPatrickM Dunne, editors (1980) 206-210. Oliver, R.L. 8: Bearden W.0. "The Role of Involvement in Satisfaction Processes," WW 1983.Basoni.R.andlyband.A.eds(AnnArbor. MI). 250-255. 152 Oliver, R.L & Gerald, L. Effect of Satisfaction and Its Antecedents on Consumer Preference and Intention, W Vol. VIII, Proceeding of the 1980 annual ACR conference, Kent Monroe, editor (1981), 88-93. Olson, J. C. & Dover, PA. "Disconfirmation of Consumer Expectatiom Through Product Trial," mm mm Vol 64 (April. 1979). 179-189 Olson, J. C. & Dover, PA. "Effects of Expectation Creation and Disconfirmation on Belief Elements of Cognitive Structure." WM Proceedings of the 1975 annual ACR conference, Beverlee B. Anderson, editor (1976), 168-175. Prakash, V. "Validity and Reliability of the Confirmation of Expectations Paradigm as a Determinate of Comumer Satisfaction." MmalaLtheAQdeImLMyketiasSsieng. Fall. 1984. 63-76- Prakxh, V. 82 Loumbury, J .W. "The Role of Expectations in the Determination of Consumer Satisfaction." WWW Summer 1984.1-77 Quelch, J .A. & Ash, S.B. "Consumer Satisfaction with Professional Services," W Proceeding of AMA’ s Special Conference on Services Marketing, Orlando, Florida 1981, James H. Donnelly and William R. George, editors, 82-85. Raynor, J .0. "Future Orientation and Motivation of Immediate Activity: An Elaboration of the Theory of Achievement Motivation. WM 1969, 76, 606-610. Rosenberg, MJ. "An Analysis of Affective~Cognitive Consistency." In AW m edited by C1. Howland arxl MJ. Rosenberg. New Haven, Conn: Yale University Press, 1960, 15-64. Rotter, J.B. "The Role of the Psychological Situation In Determining the Direction of Euman Behavior." In M R Jones (Ed) W Lincoln University of Nebraska Press. 1955. Sherif, C.W., Sherif, M. & Nebergall, R. W Yale University Press, 1961 (New Haven). Simon, J. G., & Feather, N. T. "Causal Attributions for Success and Failure at University Examinations." Mmalnffimncnallmhclm 1973. 64 46-56- Surprenant, C. "Product Satisfaction as a Function of. Expectations and Perfonmance," Com: ~ - of the 2nd annual CS/D&CB conference, April 20-22 1977, Ralph 1.. Day, editor (1977), 36-37. Swan, J.B. "Consumer Satisfaction Research and Theory: Current Status and Future Directions," WM 1982. DayandHunt. eds (KnoxvilleTN) Swan, J.B. "Consumer Satisfaction with a Retail Store Related .to the Fulfillment of Expectations on an Initial Shopping Trip," , ._ = »- Proceeding of the 2nd annual CS/D&CB conference, April 20-22, 1977, Ralph 1.. Day, editor (1977), 10-17. Swan, J.B., Longman,’ D.S., & Combs, LJ. "Brand and Store Loyalty Related to Consumer Satisfaction ' . ' _ w - n'etot 1974 annual conference, Barnett A. Greenberg, editor (1975), 113-115. 153 Swan, J.B. & Martin, WS. "Testing Comparison Level and Predictive Expectations Models of Satisfaction," WM Vol. VIII, Proceeding of the 1980 annual ACR conference, Kent Monroe, editor (1981), 77-82. Swan, J .E & Mercer, AA. "Consumer Satisfaction as a Function of Equity and Disconfirmation," ~ ' c '.t,- Proceedingofthe6thannual (S/D&CB conference, September 30-October 8, 1981, H. Keith Hunt and Ralph L. Day, editors (1982), 2-8. Swan, J.B. & Trawick, I.F. "Disconfirmation of Expectations and Satisfaction with a Retail Service," .Mmalntnslailina VoL 57, No. 3 (Fall, 1981), 49-67. Swan, J.B. & Trawick, I.F. "Patronage Satisfaction and Fulfillment of Desired, Predictive Expectations: Theory and Applications," ' 9 . ,~- - i _ F. Lusch and William R. Darden, editors (1981), 59-65. Swan, J. E. & Trawick, I ..F "Satisfaction, Disconfirmation, & Comparison of Alternatives," 9211992139. f o l Proceeding of the 6th annual CS/D&CB conference, September 30-October 3, 1981, H. Keith Hunt and Ralph 1.. Day, editors (1982), 17-24. Swan, J.B. & Trawick, I..F "Satisfaction Related to Predictive vs. Desired Expectations," Rm -- ‘ - = - -. Proceeding of the 4th annual CS/D8ICB conference, October 3-5, 1979, H. Keith Hunt and Ralph L. Day, editors (1980), 7-12. Swan, J.B. & Trawick, LF. "Satisfaction with and Support of a Continuously Utilized Service: A Public Radio Station." WW 1982. H Keith Hunt and Ralph L Day, editors. (Knoxville, TN). Swan, J .E. & Trawick, I.F. "Triggering Cues and the Evaluation of Products as Satisfactory or Dissatisfactory."1fl?;fldiimmt££9nfersnz_tmedlnss AMA. Neil Beckwith. e.t_al. editors (1979), 231-234 Swan, J.E., Trawick, [.F., & Carroll, M. G. "Satisfaction Related to Predictive, Desired Expectations: AFieldStudy," - , ProceedingoftheSth annual CS/D&CB conference, November 6-8, 1,980 Ralph L. Day and H. Keith Hunt, editors (1982), 15-22. Swan, J.B., Trawich, [.F., & Jarden, WE. "Testing a Path Analysis Model of Preshopping Image, Postshopping Satisfaction and Patronage Intentiom," ; - - . . - Marketing, Darden, Monroe, and Dillon, editors, AMA, 1983 (Chicago). - Tato, A., Burke, T.J., & Ball T. "Construct Validity of Student Evaluations Via Causal Modeling Methodology," ' -. a . Darden, Monroe, and Dillon, editors, AMA 1983 (Chicago). Tolman, E.C. "Principles of Performance." Weill, 1955, 62, 315-326. Tolman, E.C. "Purposive Behavior in Animals and Men." New York: Century, 1932. Vroom, V.H. "Work and Motivation." New York: Wiley, 1964. Weaver, D., & Brickman, P. "Expectancy, Feedback, and Disconfirmation as Independent Factors In Outcome Satisfaction." WWW Vol 30. No 3 (March. 1974). 420-428. 154 Weiner, B. W New York: Holt, Rinehart, & Winston, 1980. Weiner, B. - u. Chicago: Markhen 1972. Westbrook, RA. "A Rating Scale for Measuring Product/Service Satisfaction," W Vol. 44 (Fall, 1980), 68-72. Westbrook, RA. "A Study of Consumer Dissatisfaction Before Purchase," mm; m Vol. IV, Proceeding of the 1976 annual ACR conference, William D. Perreault, Jr. ., editor (1977), 142-148. Westbrook, RA. "Consumer Satisfaction and the Phenomenology of Emotions During Automobile Ownership Experiences." WWW Proceeding of the 7th annual CS/D&CB conference, October 17-19,1982, Ralph 1.. Day and H. Keith Hunt, editors (1983). Westbrook, RA. "Correlates oft sPost Purchase Satisfaction with Major Household Appliances," = . = - Proceeding of the 2nd annual CS/D8ICB conference, April 20-22, 1977, Ralph L. Day, editor (1977), 85-90. Westbrook, RA. "Interpersonal Affective Influences Upon Consumer Satisfaction with Products," W11. (June. 1980). 49-54. Westbrook, RA. "Prepurchase Information Search and Post Purchase Product Satisfaction," m WWW Proceedings of the 4th annual CS/D&CB conference, October 3-5, 1979, H. Keith Hunt and Ralph L. Day, editors (1980), 68-73. Westbrook, RA., Newman, J.W., 8: Taylor, JR. "Satisfaction/Dissatisfaction in the Purchase Decision Process." MmalsfMarltetias (October. 1978). 54-60 Westbrook, RA., & Oliver, R.L. "Developing Better Measures of Consumer Satisfaction: Some Preliminary Results," W Vol. VIII, Proceeding of the 1980 annual ACR conference, Kent Monroe, editor (1981), 94-99. Westbrook, RA., 8!. Reilly, M. D. "Value Percept Disparity: An Alternative to the Disconfirmation of Expectations Theory of Consumer Satisfaction." WW ed Bagozzi. R. and Tyband, A. 1983. Wheatly, JJ. ,Walton, R. G., & Chin, IS. "The Influences of Prior Product Experience, Price, and Brand on Quality Perception." WW 1978 (Ann Arbor) Wilkie, W.L., & Pessemier, EA. "Issues In Marketing's Use of Multiattribute Models," M19: Wan. Nov 1973. 428-441 Wilson, T.C. "Student-faculty Evaluation Forms in Marketing: A Review," W Eduiaticn. Spring 1982. 744- Woodruff, R.B., Cadotte, ER, &Jenkins, R.L. "ChartingaPath for CS/D Research," W WlmDayandHuntcdmncxvifleTN) 155 Woodruff, R.B., Cadotte, ER., & Jenkim, R.L. "Modeling Consumer Satisfaction Processes Using Eltperience-Based Norms." W11. August. 1983. 296-304 Wright, P., and Rip, PJ. "Product Class Advertising Effects on First-Time Buyers Decision Strategies," W Vol 7. September 1980-