The Formation of Consumer Choice Sets: a Longitudinal Investigation At the Product Class Level

ABSTRACT - In this paper, choice set formation was viewed as a categorization phenomenon in which the consumer simplifies the decision making process by assembling relevant brand/product alternatives into a category. The stability of category membership in a gift-giving context was then traced in a longitudinal study. It was found that items included in the category on two occasions tended to be elicited earlier and rated higher in exemplar-goodness and liking relative to items included only once. In addition, items deleted from the choice set were rated lower in exemplar-goodness and liking than those added to the set on the second occasion. Implications for studying the dynamics of consumers' choice set formation are discussed.

Citation:

David B. Klenosky and Arno J. Rethans (1988) ,"The Formation of Consumer Choice Sets: a Longitudinal Investigation At the Product Class Level", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 13-18.

Advances in Consumer Research Volume 15, 1988      Pages 13-18

THE FORMATION OF CONSUMER CHOICE SETS: A LONGITUDINAL INVESTIGATION AT THE PRODUCT CLASS LEVEL

David B. Klenosky, The Pennsylvania State University

Arno J. Rethans, California State University, Chico

[The authors thank Jerry Olson, Mita Sujan, and Jack Swasy for their helpful comments on an earlier draft.]

ABSTRACT -

In this paper, choice set formation was viewed as a categorization phenomenon in which the consumer simplifies the decision making process by assembling relevant brand/product alternatives into a category. The stability of category membership in a gift-giving context was then traced in a longitudinal study. It was found that items included in the category on two occasions tended to be elicited earlier and rated higher in exemplar-goodness and liking relative to items included only once. In addition, items deleted from the choice set were rated lower in exemplar-goodness and liking than those added to the set on the second occasion. Implications for studying the dynamics of consumers' choice set formation are discussed.

INTRODUCTION

Understanding how consumers form sets of choice alternatives is critically important for understanding consumer choice behavior. While researchers have spent considerable time assessing the impact of the size and nature of the choice set on choice processes and outcomes (Wright 1975; Hayes-Roth 1982; Huber, Payne, and Puto 1982; Park and Lutz 1982), they have only recently begun to address the question of how choice sets are formed.

Previous research has focused primarily on choice sets of competing brands within a specified product class and has examined these choice sets at only a single point in time (May 1979; Parkinson and Reilly 1979; Brisoux and Laroche 1980). The typical approach in these studies has been to ask consumers, at a particular point in time, to identify the brands they consider as acceptable purchase alternatives within a given product class. While this approach has been useful in developing our understanding of choice sets, it may only be relevant for situations in which the consumer has had enough experience with the choice situation in question and hence already has knowledge about the set of alternative brands that are acceptable. In such situations, the alternatives making up the consumer's choice set are likely to be relatively stable over time.

Consumers do face situations, however, in which they must make choices among alternatives from different product categories (Johnson 1984; 1986; Bettman and Sujan 1987). For example, a consumer often faces choices about what to do on a free weekend, about how to spend their income tax refund, or about what gift to buy for a friend's birthday. Since these "product class" choice situations tend to occur less frequently and are often more situation specific, the relevant alternatives are less obvious. As a result, the consumer's choice set may be less stable over time, that is, alternatives may be deleted from and added to the choice set as it is formed. To capture the process by which choice sets form under these conditions requires a longitudinal research design. Accordingly, a study was designed to examine the contents of consumer choice sets over time. In particular, the study investigated the stability of the alternatives considered in a product class choice situation by analyzing the contents of consumer choice sets at two points in time.

In the remainder of this paper we first propose a conceptual framework for viewing the formation of consumer choice sets. Next we derive study objectives from this proposed framework and we describe the results of an exploratory study designed to address some research propositions.

CONCEPTUAL FRAMEWORK

The formation of a choice set can be viewed as an act of categorization on the part of the consumer (Narayana and Markin 1975; Gutman 1982; Troye 1983). The purpose of this categorization act is to assemble the set of relevant brand/product alternatives for the purchase situation at hand and thereby facilitate the decision making process. This view recognizes that consumers form choice sets at a variety of levels, ranging from choice sets comprised of alternatives representing different product classes (e.g., buying a new stereo system versus taking a vacation), to different product forms (taking a ski vacation versus a caribbean vacation) to different brands (going to Vail, Colorado versus going to Stowe, Vermont). Regardless of the level at which the choice set forms, however, its function is to help the X consumer achieve the goal of making a purchase decision (cf. Barsalou 1983; 1985).

The present conceptualization views consumer goals as central to the process by which consumer choice -- sets form. When the consumer's goal is a familiar one, as it is in many brand choice situations, the choice category is likely to fairly well-established in memory and thus can be retrieved as a set. In these situations the consumer's choice set is likely to be relatively stable, though not totally invariant [It is important to note that even in highly familiar, recurring choice situations the contents of the choice category will likely vary from one purchase occasion to another due to the influence of situational and/or individual factors.], over time. In contrast, when the consumer's goal is a novel one, the consumer's choice category has yet to be established in memory and hence will have to be constructed for the first time. In these situations the consumer's choice set is likely to be relatively unstable over time--that is, the composition of the choice category is likely to change as the consumer learns more about the alternatives that might be relevant to the particular situation at hand. As the consumer gains experience with a particular choice situation, this category will tend to become more firmly established in memory, and hence more likely to be retrieved as opposed to constructed (cf. Barsalou 1983; Bettman and Zins 1977).

Viewing the formation process in this manner allows us to draw upon recent work in categorization theory and related consumer research for insights into the stability of consumer choice sets. This work has shown that the members of a category are not identical but vary in terms of their degree of category membership or "typicality." Thus, the members of a category (e.g., birds) vary in terms of their typicality ranging from those that are very typical (e.g., robin) to those that are very atypical (e.g., ostrich).

Researchers in both the categorization and consumer behavior areas have shown that the more typical members of a category tend to be listed earlier and more frequently in elicitation tasks than less typical members (Mervis, Catlin and Rosch 1976; Nedungadi and Hutchinson 1985). Stated differently, items that are frequently associated with the category over time tend to be the more typical members of the category. These results suggest that alternatives that are consistently included in the consumer's choice set should be elicited early and be rated high in typicality.

Early work in categorization theory viewed categories as inherently neutral entities. This view has recently been challenged in a variety of domains including consumer behavior. As Cohen (1982, p.98) argues, consumers "don't simply organize the environment out of idle curiosity...they group objects in categories to meet a common purpose or end. To achieve such a purpose, however, categories should not be neutral but inherently evaluative." Thus if consumer choice sets are to be functional (i.e., facilitate the decision making process), they will be comprised of alternatives with generally positive evaluations [We recognize that the consumer's choice set may also be comprised of alternatives with negative evaluations, as in the case of a homeowner who must choose among a number of alternative insurance policies each of which he evaluates negatively. However, even in such cases the consumer is likely to consider only those alternatives which he views as minimally negative.]. Further, the alternatives that are consistently included in the consumer's choice set over time should have more positive evaluations.

STUDY OBJECTIVES AND METHODOLOGY

The proposed conceptual framework has a number of implications for the study of choice sets. The framework would suggest that in a longitudinal study of consumer choice sets one might expect a subset of items to remain under consideration over time. That is, while alternatives may be dropped from and added to the choice set as it develops, certain alternatives are likely to be consistently included in the choice set over time, i.e., exhibit a pattern of stability. This stability is perceived to be a function of the degree of typicality and liking of the item. Conversely, one would expect other items to be unstable due to low degrees of typicality and lower positive evaluations. In combination, these expectations suggest the following research proposition:

Relative to the other members, stable members of a consumer choice set are listed earlier in elicitations, are rated higher in typicality, and are rated higher in liking.

To explore this research proposition a two-phase study within a gift-giving context was designed. The context was selected so as to be a familiar setting for the subject populationCundergraduate students. At each phase of the study, subjects completed a questionnaire consisting of two parts. In the first part, subjects read a choice scenario and generated a list of choice alternatives. The choice scenario described a person who needed to buy a birthday gift for their fatherCsomeone who seems to have "everything." Subjects were asked to place themselves in that situation and were instructed to list "the types of gifts that you would consider buying for the man who has everything." In the second part, subjects rated each of the self-generated alternatives on a series of four rating scales: two scales to measure liking and two scales to measure typicality. The liking variable was measured using two 7-point semantic differential scales (good/bad; like/dislike). For these ratings subjects were instructed to evaluate how they felt about the item as a gift to get for the man who has everything. Coefficient alpha was 0.96 for time period 1 (i.e., phase 1) and 0.94 for time period 2 (phase 2). Typicality was also measured with two 7-point semantic differential scales (excellent example/poor example; extremely typical/not typical at all). The first scale is frequently -used to operationalize degree of category membership, it measured how "good an example" the item was of its category (Rosch 1973). The second scale was expected to be related to the first, it measured how typical the item was of its category. The coefficient alphas for the two scales were disappointing, -0.21 and 0.15 respectively for time periods 1 and 2. At least in the context of the present study, it appeared that how good an example the item was of the category was not related to how typical the item was. Based on this finding, the decision was made to treat the two scales as separate in the analyses of the results. The two scales were termed "exemplar-goodness" and " typicality" respectively . Thus, three dependent measures were employed in the analyses: liking, exemplar-goodness and typicality.

In the first phase of the study, 83 subjects completed the study questionnaire. In the second phase, subjects completed the same questionnaire following a one-week time delay. The 61 subjects who completed both phases of the study produced the data analyzed here.

RESULTS

Number of Items Generated

As might be expected, subjects generated a wide variety of gift alternatives. (A rough classification of the items generated in each time period is provided in the Appendix). For each subject the items generated were coded as either matches or non-matches according to whether they appeared in the choice set in both time periods or in only one. The results of this procedure are presented in Table 1. In any given time period subjects' choice sets were made up of an average of 7-8 items, with a range of 2 to 10 items. Just over half (57%) of the items listed at T1 were listed again at T2; a similar percentage (54%) was found for the items listed at T2. Of the items that were listed in both time periods the majority were exact verbatim matches (e.g., the subject listed "a shirt" in both time periods); roughly 10% were specific to general matches (i.e., the subject listed "a shirt" in T1 and "clothing" in T2 1; and a fraction were general to specific matches (i.e., the subject listed "clothing" in T1 and "a shirt" in T2). A by-subject analysis of choice set stability reveals that only two subjects (3%) had no matching items, while five subjects (8%) recorded an exact matching of their first list. These results suggest that both stable and non-stable items were present in subjects' choice sets.

TABLE 1

MEAN NUMBER OF ITEMS LISTED BY TIME PERIOD

Characteristics of Stable and Non-stable Items

The research proposition stated that the stable members of subjects' choice sets would be elicited earlier, rated higher in typicality, and rated higher in liking than non-stable members. To explore this proposition the order in which the items were listed was analyzed first. As evidenced in Table 2, the stable members of the choice set (those that were listed twice) tended to be elicited earlier than members that were not stable (those that were listed only once). This pattern held true for both time periods.

To further examine the differences between the stable and non-stable members of consumers' choice sets, mean scores on the liking, exemplar-goodness, and typicality scales for the stable items were compared with those for the non-stable items (see Table 3). For each comparison, two scores were computed for each subject the mean rating for the subset of items listed once and the mean rating for the subset of items listed twice. The statistical significance of the difference between these two scores was computed using a simple t-test.

For the items listed in the first time period, the mean liking and exemplar-goodness scores for the items listed twice were higher than the scores for the items listed once (for liking: t(53) = 7.10, p < .0001; and for exemplar-goodness: t(53) = 5.99, p < .0001).

TABLE 2

PERCENT OF SUBJECTS LISTING EACH ITEM ONCE VERSUS TWICE

TABLE 3

MEAN SCORES FOR ITEMS LISTED ONCE VERSUS TWICE

Similar results were also found for the items listed in the second time period (for liking: t(55) = 3.26, p <.005; and for exemplar-goodness: t(55) = 2.90, p < .005). These results provide strong evidence for our research proposition. The typicality results, however, although in the hypothesized direction, did not differ significantly for either time period (T1: t(53) = 0.80, N.S.; T2: t(55) = 1 .34, p < . 10).

Liking, Exemplar-goodness and Typicality Characteristics

To examine the liking, exemplar-goodness, and typicality characteristics of the items generated, two additional analyses were performed. Mean scores on the study variables for the stable and non-stable items across the time periods were tested for differences. As shown in Table 4, the mean exemplar-goodness score for the matching items at T1 was slightly higher than at T2 (t(57) = 1.64, p < .05). A similar result was observed for the liking measure, although it only approached significance (t(58) = 1.30, p < .10). For the typicality measure, the mean score was slightly higher at T2, although not significantly higher (t(57) = -1.13, N.S.).

For the set of non-matching items, the items that were dropped from the choice set (those listed in T1 only) were lower in both liking and exemplar-goodness relative to the items that were added to it (those listed in T2 only) (for liking: t(53) = -3.13, p < .005; and for exemplar-goodness: t(52) = -2.03, p < .025). In contrast, for the typicality measure, the deleted items tended to be viewed as slightly more typical relative to the items that were added, although, once again, this difference was not statistically significant (t(52) = 0.97, N.S.). Thus, compared to the items considered in T1 but subsequently dropped, the items that were added to the choice set in T2 were generally higher in liking and exemplar-goodness and lower in typicality. It appears that subjects replaced the items they deleted with additional items that they considered to be better choice alternatives.

TABLE 4

MEAN SCORES FOR MATCHING AND NON-MATCHING ITEMS LISTED IN TIME 1 VERSUS TIME 2

DISCUSSION

The purpose of this investigation was to examine the dynamics of consumer choice set formation from a categorization perspective. The results of the study are encouraging. Items which remain in the choice set over time are liked more, seen as better examples of the category, and are slightly more typical than items which do not remain in the choice set. The stable items also tended to appear earlier in lists resulting from the elicitation procedures. Thus, some initial support for adopting a categorization perspective has been provided. At the same time, however, the study raises a number of issues which ought to be explored in future research.

One somewhat disturbing finding was that the two scales used to measure typicality were not related. The low coefficient alpha suggested that the two ratings, exemplar-goodness and typicality, should be treated as two separate measuresCone that captures how good an example an item is of its category and a second that captures how typical the item is of its category. One possibility is that subjects did not fully understand the rationale behind the exemplar-goodness measure and might have interpreted it as an attitude measure rather than as a measure of degree of category membership. A more intriguing possibility is that the observed independence of the two measures was only an artifact of the particular context involved in the study, that of "choosing a gift for a man who has everything." By construction, a good example of a gift for the man who has everything should not be something that is typical but, rather, something that is highly atypical, i.e., something novel and unique. In a different context, the two dimensions might actually have been highly related to each other. For example, when "deciding what to do with one's income tax refund," the two options of "taking a vacation" and "buying a stereo system" might both be seen as good examples that are also highly typical. This explanation suggests that, at least for certain categories, the typicality of an item does not necessarily correspond with how good an example the item is.

A more positive and intriguing finding is the strong relationship observed between liking and exemplar-goodness (the correlation between the two measures was 0.83 in T1 and 0.88 in T2). One plausible interpretation of this relationship is that the two measures both tap the same underlying dimension: how appropriate the alternative is for achieving the particular goal/objective in question. The choice alternatives that are liked and that are seen as good examples of their category are appropriate because they are likely to satisfy one's goal of making a suitable purchase decision. The results of the present study are consistent with this interpretation. The items that were included as members of subjects' choice sets were all seen as appropriate choice alternatives and hence all received generally positive evaluation ;. In addition, the stable items in subjects' choice sets, were rated high in liking and exemplar-goodness relative to the items that were not stable, suggesting that they may have been the more appropriate choice alternatives.

SUMMARY AND CONCLUSIONS

In this paper we have explored the dynamics of consumer choice set formation from a categorization perspective. Choice set formation was viewed as a categorization phenomenon in which the consumer simplifies his or her decision making process by assembling relevant brand/product alternatives into a category. The stability of category membership in a gift-giving context was then traced in a longitudinal study. It was found that the items that were included in the category on two occasions tended to be elicited earlier, rated higher in exemplar-goodness, and rated higher in liking than items that were included only once. In addition, it was found that items that were deleted from the choice set were rated lower in exemplar-goodness and liking than those that were added to the set on the second occasion.

These findings in combination with earlier reported findings in the literature suggest a number of issues for future research. Chief among these is the role of evaluative factors in the formation of consumer choice sets. Evidence from the present study suggests that these evaluative factors are the driving forces behind the formation of consumer choice sets. The origin of these factors and their relationship to other theoretical constructs are important topics for future research.

APPENDIX

REFERENCES

Barsalou, Lawrence W. (1983), "Ad Hoc Categories," Memory and Cognition, 11, 211-227.

Barsalou, Lawrence W. (1985), "Ideals, Central Tendency, and Frequency of Instantiation as Determinants of Graded Structure in Categories," Journal of Experimental Psychology: Learning, Memory and Cognition, 11 (October), 629-649.

Bettman, James R. and Mita Sujan (1987), "Effects of Framing on Evaluation of Comparable and Noncomparable Alternatives by Expert and Novice Consumers," forthcoming Journal of Consumer Research, 14 (September).

Bettman, James R. and Michael Zins (1977), "Constructive Processes in Consumer Choice," Journal of Consumer Research, 4 (September), 75-85.

Brisoux, Jacques E. and Michel Laroche (1980), "Evoked Set Formation and Composition: An Empirical Investigation Under a Routinized Response Behavior Situation," in Kent B. Monroe, ed., Advances in Consumer Research, Vol. 8, 357-61.

Cohen, Joel B. (1982), "The Role of Affect in Categorization: Toward a Reconsideration of the Concept of Attitude," in Andrew. A. Mitchell, ed., Advances in Consumer Research, Vol. 9, 94-100.

Gutman, Jonathan (1982), "A Means-End Chain Model Based on Consumer Categorization Processes," Journal of Marketing, 46 (Spring), 60-72.

Hayes-Roth, Barbara (1982), "Opportunism in Consumer Behavior," in Andrew. A. Mitchell, ed., Advances in Consumer Research, Vol. 9, 132-135.

Huber, Joel, Payne, John W., and Christopher Puto (1982), "Adding Asymmetrically Dominated Alternatives: Violations of Regularity and the Similarity Hypothesis," Journal of Consumer Research, 9 (June), 90-98.

Johnson, Michael D. (1984), "Consumer Choice Strategies for Comparing Non-comparable Alternatives," Journal of Consumer Research, 11 (December), 741-53.

Johnson, Michael D. (1986), "Modeling Choice Strategies for Non-comparable Alternatives," Marketing Science, 5 (Winter), 37-54.

May, Frederick (1979), "Evoked Set Formation and Composition: The Learning and Information Processing Hypothesis," in William L. Wilkie, ed., Advances in Consumer Research, Vol. 6, 222-226.

Mervis, Carolyn B., J. Catlin, and Eleanor Rosch (1976), "Relationship Among Goodness-of-Example, Category Norms, and Word Frequency," Bulletin of the Psychonomic Society, 7, 283-284.

Narayana, Chem L. and Rom J. Markin (1975), Consumer Behavior and Product Performance; An Alternative Conceptualization," Journal of Marketing, 39 (October), 1-6.

Nedungadi, Prakash and J W. Hutchinson (1985), "The Prototypicality of Brands: Relationships with Brand Awareness, Preference, and Usage," in E. Hirschman and M. Holbrook, eds., Advances in Consumer Research, Vol. 12, 498-503.

Park, C. Whan and Richard Lutz (1982), "Decision Plans and Consumer Choice Dynamics," Journal of Marketing Research, 19 (February), 108-115.

Parkinson Thomas L. and Michael Reilly (1979), "An Information Processing Approach to Evoked Set Formation," in William L. Wilkie, ed., Advances in Consumer Research, Vol. 6, 227-31.

Rosch, Eleanor (1973), "On the Internal Structure of Perceptual and Semantic Categories," in T.E. Moore, ed., Cognitive Development of the Acquisition of Language, New York: Academic Press.

Troye, Sigurd Villads (1983), "Evoked Set Formation as a Categorization Process," in Thomas C. Kinnear, ed., Advances in Consumer Research, Vol. 11, 180-86.

Wright, Peter (1975), "Consumer Choice Strategies: Simplifying versus Optimizing," Journal of Marketing Research, 12 (February), 60-67.

----------------------------------------