FACTORS DETERMINING CONSIDERATION SET COMPOSITION IN NOVEL PURCHASE SITUATIONS By AMITAV CHAKRAVARTI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE O F DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2002
ii ACKNOWLEDGEMENTS This dissertation leaves me heavily indebted to many people. First, I would like to thank my advisor, Christopher Janiszewski, for guiding me through the entire dissertation proces s. Experimental research, like any other research, can often be very frustrating, but his steady mentoring ensured that the process remained enjoyable and fun. Interactions with other members of my dissertation committee were also invaluable in shaping thi s dissertation. In particular, I benefited immensely from the extensive comments and feedback from Barton Weitz, Richard Lutz, and Alan Cooke, and from the discussions I had with James Algina. I am also indebted to participants in the interviews held at th e 2001 American Marketing Association conference held at Washington, D. C. and seminars at the University of Colorado, New York University, INSEAD, Northwestern University, Indiana University, University of Maryland, Rutgers University, University of Miami , University of Toronto, and Hong Kong University of Science and Technology. I would like to specifically thank Dipankar Chakravarti, Brian Sternthal, and Bob Wyer for their insightful comments. I would also like to take this opportunity to thank Jinhong X ie, Joe Alba, Alan Sawyer, Steve Shugan and Joel Cohen for helping me out at different stages of my doctoral studies. Of special importance in my Ph.D. experience were my friends in the doctoral program, especially Tom Meyvis, Hayden Noel, Els De Wilde, Ma rcus da Cunha, YuBo Chen, Tim Silk, and Eduardo Andrade. My thanks also go to Bijoy Sarma Roy, Aparna Sarma Roy, Pathikrit Sengupta, and Carol Mendoza for their
iii constant support. Last, but not least, I am extremely thankful to my family -my parents and my sister -for their enthusiasm and unflagging encouragement throughout the entire duration of my doctoral studies.
iv TABLE OF CONTENTS page ACKNOWLEDGEMENTS ................................ ................................ .......... ii ABSTRACT ................................ ................................ ................................ .. vii CHAPTERS 1 INTRODUCTION ................................ ................................ ............ 1 2 THEORETICAL BACKGROUND ................................ .................. 4 Stimulus Based Consideration Sets ................................ .................. 4 Simplifying a Decision ................................ ................................ ..... 4 Optimize a Choice ................................ ................................ ............. 8 Organization of Studies ................................ ................................ ..... 9 3 STUDY 1: THE ROLE OF ALIGNABILITY ................................ . 11 Design and Stimuli ................................ ................................ ............ 12 Proce dure ................................ ................................ .......................... 13 Results ................................ ................................ ............................... 15 Discussion ................................ ................................ ......................... 16 Study 1A ................................ ................................ ........................... 17 Study 1B ................................ ................................ ............................ 18 Study 1C ................................ ................................ ............................ 18 4 STUDY 2: THE ROLE OF SIMILARITY ................................ ....... 21 Design, Stimuli, and Procedure ................................ ........................ 21 Results ................................ ................................ ............................... 23 Discussion ................................ ................................ ......................... 23 Study 2A ................................ ................................ ........................... 24
v 5 STUDY 3: THE EFFECT OF TIME PRESSURE ........................... 29 Procedure ................................ ................................ .......................... 29 Results ................................ ................................ ............................... 30 Discussion ................................ ................................ ......................... 30 6 STUD Y 4: THE EFFECT OF NEGATIVELY CORRELATED BENEFITS ................................ ................................ ................................ ............ 32 Design and Manipulation ................................ ................................ .. 32 Stimuli, Procedure, and Dependent Measures ................................ .. 33 Results ................................ ................................ ............................... 33 Discussion ................................ ................................ ......................... 34 Study 4A ................................ ................................ ........................... 34 7 STUDY 5: THE EFFECT OF TIME PRESSURE AT CONSIDERATION VERSUS FINAL CHOICE STAGES ................................ .............. 37 Design and Manipulation ................................ ................................ .. 37 Stimuli and Procedure ................................ ................................ ....... 38 Results ................................ ................................ ............................... 38 Discussion ................................ ................................ ......................... 39 8 GENERAL DISCUSSION ................................ ............................... 40 9 LIMITATION S AND FUTURE RESEARCH ................................ . 44 REFERENCES ................................ ................................ ............................. 47 APPENDICES A SAMPLE OF STIMULI USED IN STUDY 1 ................................ . 51 B ATTRIBUTES AND BRANDS USED IN STUDY 1 ..................... 53 C DEPENDENT VARIABLE USED IN STUDY 1 ............................ 54 D DETAILED RESULTS OF STUDY 1 1B ................................ ....... 55 E SAMPLE OF STIMULI USED IN STUDY 1C ............................... 56 F STIMULI USED IN STUDY 2 ................................ ........................ 57 G FORMULA FOR JACCARD RATIO OF SIMILARITY ................ 58 H DETAILED RESULTS FROM STUDY 2 ................................ ....... 59
vi I STIMULI USED IN STUDY 2A ................................ ..................... 60 J EXAMPLE OF UND ERLYING BENEFITS IN STUDY 4 ............ 62 K STUDY 4 DESIGN ................................ ................................ ........... 63 L SAMPLE OF STIMULI USED IN STUDY 4A ............................... 64 M FEATURES AND BENEFITS USED IN STUDY 4A AND 5 ....... 65 BIOGRAPHICAL SKECTH ................................ ................................ ........ 66
vii Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor in Philosophy FACTORS DETERMINING CONSIDERATION SET C OMPOSITION IN NOVEL PURCHASE SI TUATIONS By Amitav Chakravarti August 2002 Chairman: Christopher A. Janiszewski Major Department: Marketing Despite the presence of a long research tradition investigating consideration set formation, this research negl ects a very important issue -consideration set content. Consideration set content refers to the composition of the consideration set with respect to the similarities or dissimilarities of the constituent brands or alternatives. Although this issue is relev ant to all types of purchase situations, it acquires added importance in novel purchase situations since it is in these situations that consideration set composition tends to be most malleable. This dissertation investigates several important factors that influence the composition of consideration sets in novel purchase situations. In situations where consumers have to create consideration sets from a novel set of alternatives, the contents of a consideration set depend on a combination of two motives. Firs t, consumers prefer to create a consideration set of easy to compare alternatives. It is easier to compare
viii alternatives that have alignable attributes or alternatives that are perceptually similar; hence consideration sets tend to consist of homogeneous a lternatives. Second, consumers prefer to create consideration sets that have a high likelihood of containing their optimal alternative. When the set of available alternatives requires the consumer to make tradeoffs between benefits (i.e., to be compensator y), the consumer often delays making a decision about which benefits are preferable and the consideration set tends to be more heterogeneous. Through ten experiments I document factors that influence the relative importance of one or the other motive in co nsideration set formation and discuss implications for brand managers .
1 CHAPTER 1 INTRODUCTION Consideration set formation is a fundamental stage of pre choice decision making (Alba, Hutchinson, and Lynch 1991; Howard and Sheth 1969; Nedungadi 1990; R atneshwar and Shocker 1991). Consideration sets can be used to understand consumer choice strategies, brand loyalty, and changes in market share that are independent of brand evaluation (Kardes et al. 1993; Nedungadi 1990). In fact, it is the inherent inst ability of the composition of a consumerâ€™s consideration set that creates an opportunity for brand managers to increase a brandâ€™s market share (Desai and Hoyer 2000; Hutchinson, Raman, and Mantrala 1994). Although this opportunity exists in all markets and with all brands, the composition of a consideration set is likely to be most malleable when the consumer encounters novel product categories, novel buying situations, and novel consumption contexts (Desai and Hoyer 2000). The majority of research on consi deration set composition has focused on how situational factors influence memory based consideration set formation (Desai and Hoyer 2000; Hutchinson et al. 1994; Ratneshwar and Shocker 1991). In general, we know that advertising and consumption experience combine to reinforce associations between products/brands and benefits (Mitra 1995; Ratneshwar and Shocker 1991). These associations, in turn, allow a consumer to retrieve the names of particular brands when encountering a particular consumption context (D esai and Hoyer 2000; Posavac, Sanbonmatsu, and Fazio 1997; Ratneshwar and Shocker 1991). Thus, the consumption
2 context creates usage goals and memory mediates whether a brand is retrieved and considered appropriate for achieving these goals (Desai and Hoy er 2000; Nedungadi 1990; Posavac et al. 1997; Ratneshwar and Shocker 1991). Consideration set composition may be influenced by goals in addition to those created by the consumption context. For example, consumers may prefer to create consideration sets tha t consist of items that are easy to compare or that increase the likelihood an optimal alternative is retained for further consideration (Huber and Kline 1991; Wright 1975). Macro level goals, like trying to reduce the complexity of a decision, have been w ell documented in decision making (e.g., Bettman 1979; Johnson and Payne 1985; Ratneshwar and Shocker 1991), but have usually been used to explain limits on the size of a consideration set as opposed to determining the composition of a consideration set (e .g., Jarvis and Wilcox 1973; Mitra and Lynch 1995; Roberts and Lattin 1991). It is possible that these macro level goals also make consumers sensitive to the characteristics of the alternatives themselves. This sensitivity to the characteristics of the alt ernatives in the consumerâ€™s awareness set may be particularly salient when the choice is stimulus based and when the consumer has limited experience with the product category or consumption occasion. This dissertation is organized as follows. I begin with a review of whether consumers tend to create consideration sets of homogeneous or heterogeneous alternatives and offer insight into why each strategy might be useful. I argue that consumers prefer to create consideration sets of alternatives that are easy to compare and that both attribute alignability and alternative similarity make alternatives easier to compare. I then use six studies to document (1) that the desire to have easily compared
3 alternatives encourages consumers to create homogeneous considera tion sets, and (2) that competing factors (e.g., alternative familiarity, preferences, category structure, attribute importance) are not responsible for the consumerâ€™s desire to create homogeneous consideration sets. Using two additional studies, I argue t hat concerns about erroneously leaving out an optimal alternative and anticipated time pressure at consideration vis a vis choice encourage people to create more heterogeneous consideration sets. I end with a discussion of marketing implications and avenue s for future research.
4 CHAPTER 2 THEORETICAL BACKGROUND Stimulus Based Consideration Sets Stimulus based consideration sets are often constructed when the consumer confronts a large number of alternatives (e.g., an in store display) and needs to reduc e these alternatives to a manageable set before engaging in a more effortful evaluation process. Situations in which stimulus based consideration sets are likely to be generated include product categories with many alternatives (e.g., toys), infrequent pur chases (e.g., furniture), novel purchase situations (e.g., buying a souvenir), and ambiguous purchase goals (e.g., buying food for a large party). In these situations, consumers may have general goals that dictate the formation of a consideration set. I pr opose that two of these goals may be a desire to simplify the decision and a desire to ensure that the optimal alternative is available at the time of choice (Huber and Kline 1991; Wright 1975). Simplifying a Decision Consideration set formation has often been described as a relatively effortless process aimed at simplifying the more burdensome final choice task (e.g., Bettman 1979; Huber and Kline 1991; Johnson and Payne 1985). The desire to simplify the final choice can encourage the consumer to (1) reduc e the number of alternatives to a more manageable size (Roberts and Lattin 1991), and (2) retain alternatives that are easy to compare (Medin, Goldstone, and Markman 1995). In turn, the ease of comparing alternatives has been hypothesized to depend on the alignability of the
5 attributes used to describe the alternatives and the similarity of the alternatives being compared (i.e., feature overlap). Medin et al. (1995) observe that comparable alternatives (i.e., alternatives having alignable attributes) are p erceived as more similar and as easier to compare than non comparable alternatives. Likewise, Johnson (1984, 1986) suggests that comparisons among available choices are much easier to make when they possess features that can be compared (i.e., when alterna tives have overlapping features). Attribute Alignability There is evidence that people naturally assign more weight to alignable attributes and that they experience difficulty comparing alternatives with non alignable attributes. Markman and Medin (1995) present consumers with pairs of alternatives consisting of comparable (i.e., brands described on the same attributes) and non comparable (i.e., brands described on different attributes) brands. They find that people pay more attention to alignable attribut es and that people are more likely to mention alignable attributes as a justification for a decision. Tversky and Sattath (1979) provide additional evidence of the difficulty people experience comparing alternatives with non alignable attributes. They aske d subjects to choose a famous person they would like to meet from a set comprised of movie stars and politicians (e.g., Gandhi, Churchill and Chaplin). Tversky and Sattath (1979) find that subjects eliminate the non comparable alternative (e.g., Chaplin) f rom the set so that their final choice can be between comparable alternatives (e.g., Gandhi and Churchill). Johnson (1989) further observes that non comparable alternatives can be compared on a higher order attributes (i.e., the common currency argument), but that the creation of higher order attributes requires considerable effort.
6 Alternative Similarity In cases where there is no attribute alignability, consumers must use an alternative strategy for comparing alternatives. The consumer must convert the at tributes into their corresponding benefits and then compare the benefit information across alternatives. The generation of this benefit information is likely to be guided by the effort minimizing heuristics often witnessed in consumer behavior studies (e.g ., Bettman 1979), as only a minority of the consuming population (e.g., experts) possess enough knowledge about a class of products to be able to naturally focus on the underlying benefits (Maheswaran and Sternthal 1990). One possible low effort strategy that could promote the comparability of alternatives relies on overlapping features to guide the benefit generation process. When alternatives are described using feature information, some alternatives will have one set of features and other alternatives w ill have another set of features. When subsets of alternatives have common features, these features are overlapping. Overlapping clusters of features facilitate the identification of underlying benefits (i.e., it is easier for the consumers to see the comm onality in the products) and, thus, are weighted more heavily in decision making (Johnson 1984, 1986, 1989). In effect, to the extent subsets of alternatives become more similar, it becomes easier for the consumer to establish an underlying benefit structu re, and the alternatives become more comparable. Evidence To date, no studies have investigated whether the ease of alternative comparison is an important motivation during consideration set formation. Yet, if attribute alignability and alternative simila rity do promote alternative comparability, then one
7 consequence of this motive should be that consumers naturally create homogeneous consideration sets. Evidence that consumers naturally create homogeneous consideration sets is mixed. Lattin and Roberts (1 992) investigate consideration set formation and choice using 26 brands in the ready to eat cereal market. Lattin and Roberts (1992) find that perceptually similar brands tend to co occur in consumer consideration sets. Moreover, engaging in marketing acti ons that make a brand more similar to a well defined cluster of brands increases the likelihood that the brand will enter consumersâ€™ consideration sets and will experience an increase in market share. Lattin and Roberts (1992) conclude that cluster members hip and consideration set membership are not independent of each other. In contrast to Lattin and Roberts (1992), Hauser and Wernerfelt (1989) find no evidence that an increase in the perceptual similarity of brands will lead to an increase in the likelih ood the brands will co occur in a consideration set. Hauser and Wernerfelt (1989) ask consumers to create consideration sets of up to four brands from a set of four brands of plastic wrap and a set of four brands of dishwashing detergent. They find that in clusion of one brand in the consideration set is not contingent on the inclusion of another brand. Unfortunately, they did not measure the similarity among the four brands in each product category, so we cannot know if any pair of brands was significantly more similar than any other pair of brands. Troye (1984) found mixed evidence about the nature of consideration set composition. Troye (1984) had people create consideration sets of apartments from a set of 12 apartments that varied in similarity. He found that similarity ratings of consideration set members tended to be significantly higher than those of randomly
8 selected pairs of alternatives. In contrast, evidence for a tendency to create heterogeneous consideration sets came from a third experimental co ndition. In this condition, people created consideration sets from a list of 12 apartments that clustered into three groups. Consideration set composition did not differ significantly from randomly chosen subsets of apartments in terms of pair wise similar ity ratings. Further, more than half the subjects created their consideration sets by selecting brands from multiple clusters (i.e., an attempt to maximize dissimilarity). Optimize a Choice A second goal that may be active during consideration set formati on is the desire to optimize a choice. Huber and Kline (1991) describe consideration set formation as a tradeoff between imposing cutoffs that simplify a decision and taking a chance that an optimal alternative will be discarded. This concern about optimiz ation is likely to become more salient as the risk of neglecting an optimal alternative increases. Bettman (1979) argues risk increases when (a) the consumer lacks experience, (b) the available brand information is ambiguous, and (c) the consumer is not co nfident about the brand information. These are characteristics of purchases in novel product categories, novel buying situations, and novel consumption contexts. Risk also increases when the consumer is uncertain about his/her preferences (Kardes et al. 19 93; Simonson 1990). Uncertainty about preferences can occur for a variety of reasons. First, the consumer can be uncertain about the combination of desired features (Kardes et al. 1993). First time buyers or buyers of new products and innovations will not have well established preference structures. Second, consumers can be uncertain about the weights to assign to desired benefits when making an overall evaluation. In
9 some product classes, alternatives have negatively correlated benefit structures (Huber an d Kline 1991). This means that the consumer can buy an alternative that performs well on one benefit but not the other or vice versa. The negatively correlated benefit structure creates uncertainty about the reliability of the choice. Third, the consumer c an have a difficult time anticipating the benefit weights at the time of choice. The further into the future a choice will be made, the more uncertain the consumer becomes about the stability of currently known benefit weights (Simonson 1990). This anticip ated uncertainty has been closely linked to product heterogeneity in multi product choice sets (Simonson 1990; Simonson and Winer 1992). Organization of Studies The studies are grouped into two sets. The first set of studies shows that a consumerâ€™s desire to simplify a choice encourages the formation of a homogeneous consideration set. Study 1 uses a manipulation of attribute alignability to manipulate the degree to which alternatives are easily comparable and shows that people naturally create consideratio n sets of more comparable alternatives. Studies 1a, 1b, and 1c rule out other processes that could have encouraged the formation of a homogenous consideration set. Study 2 uses a manipulation of alternative similarity to manipulate comparability and shows that people naturally create consideration sets of more similar alternatives. Study 3 shows that people are willing to consider more heterogeneous alternatives, but only when they are encouraged to create higher order benefits that facilitate the compariso n. The second set of studies shows that a consumer will activate an optimization strategy when necessary. Study 4 shows that people create more heterogeneous consideration sets when confronted with an alternative set having negatively correlated benefits. Study 5 shows
10 that the consideration set becomes less heterogeneous if the consumer is encouraged to determine the relative importance of the competing benefits.
11 CHAPTER 3 STUDY 1: THE ROLE OF ALIGNABILITY The goal of study 1 was to assess whether or n ot people are more likely to include easily comparable alternatives in a consideration set. I manipulated comparability by altering the degree of attribute alignability between pairs of brands in a large set of alternatives. The choice of attribute alignab ility as a means of manipulating comparability was based on findings in the decision making literature. Investigations into decision making document that people attend to three types of information when comparing alternatives -commonalties, alignable diffe rences and non alignable differences (Medin et al. 1995). When two alternatives (a) share an attribute and also have identical values on that attribute, it is referred to as a commonality, (b) share an attribute but have different values on that attribute, it is referred to as an alignable difference, and (c) are such that one alternative has an attribute/feature that the other one does not have, it is referred to as a non alignable difference. Findings from prior research in this area indicate that in deci sion tasks involving comparisons, alignable differences are weighed more heavily than either non alignable differences or commonalties (Markman and Medin 1995). Thus, brands that have more alignable attributes are perceived as more comparable than brands t hat have more non alignable attributes (Markman and Medin 1995). Said differently, the research question that study 1 tried to address was as follows: when faced with product alternatives that vary in their alignability to each other, were consumers more likely to
12 include the alignable alternatives in their consideration sets than the non alignable ones? If comparability played a significant role in consideration set formation then we would find that alignable alternatives were more likely to enter the co nsideration set in comparison to their non alignable counterparts. Design and Stimuli Three replicates (e.g., video games, cars, and restaurants) were used to investigate the hypothesis that people are more likely to include more comparable alternatives in their consideration set. Each replicate comprised of 16 alternatives and each alternative was described on six attributes. The comparability of these sixteen alternatives was manipulated by varying the number of alignable attributes. Highly comparable alt ernatives had six alignable attributes, moderately comparable alternatives had three alignable and three non alignable attributes, and non comparable alternatives had six non alignable attributes. The nature of the stimuli can be illustrated with the video game replicate (see Appendix A). Sixteen video games were made available to the subjects, of which there were eight sports video games and eight combat video games. The eight sports video games included four baseball games and four football games and the eight combat video games included four aerial combat games and four individual combat games. The key manipulation was the number of alignable attributes that the different types of games shared. Games of a particular type were highly alignable alternatives . For example, any two baseball games were described on the same six attributes. Games within a subcategory but from two different types were moderately alignable alternatives. For example, a baseball game and a football game were described on three common attributes
13 and three unique attributes. Finally games from two different subcategories were non alignable. For example, a sports game and a combat game were described by two unique sets of six attributes. Based on the manipulations discussed above, attri bute descriptions were created for the sixteen alternatives in each replicate. The description for each alternative consisted of an introductory sentence, a concluding sentence, and six attribute description sentences. The attribute description sentences w ere based on actual descriptions available on product packages and web sites. A sample of these descriptions for the video game replicate is provided in Appendix A. The attributes and brand names used to describe the options for the other two replicates (i .e., cars and restaurants) are shown in Appendix B. Procedure There were four steps in the procedure. Step 1 At the beginning of the computer based questionnaire, subjects were told that they were about to participate in a two step consumer decision maki ng task. They were further told that they would have to look at the alternatives available in a category and form a â€œshortlistâ€ of the alternatives that they would like to consider further. Then in the second step they would have to choose one alternative from the shortlist they created earlier. Subjects were specifically told to base their decisions strictly on the information that was provided to them and discount any previous experiences they may have had with the brands or the categories.
14 Step 2 Subj ects were then exposed to two screens. On the first screen, subjects saw the names of sixteen brands displayed as links. Clicking on the link for a particular brand displayed the eight sentence description. Below these links were instructions asking the su bjects to form a short list of alternatives for further consideration. At the bottom of the screen was a text box where the subjects had to enter the names of the alternatives they wished to consider further. There was no restriction on either the amount o f time the subjects could spend creating the consideration set or the number of items they could choose to include in their consideration set. There were only two constraints in the consideration set formation task: (a) subjects were required to list more than one brand in their consideration set, and (b) subjects had to click on each link at least once. The latter constraint was to make sure that no alternative was unintentionally overlooked. Additionally, this constraint also reduced variance in the "awar eness set" across individuals. After the subjects had formed their consideration set they clicked on a button to continue with the rest of the study. On the screen that followed, subjects provided a brief explanation regarding how they went about creating their consideration set. They then clicked on a button to continue with the rest of the study. Step 3 On the third screen, subjects were asked to make their final choice from their consideration set. After making their final choice, subjects were required to provide brief explanations about how they made their final choice. Steps 2 and 3 were completed for each replicate prior to considering the next step. The presentation order of the replicates was random.
15 Step 4 In the final step, as a manipulation che ck, subjects judged the comparability of the various subcategories within each replicate. For example, for the video game replicate they rated how comparable they thought (a) the baseball games were to the football games, (b) the baseball games were to the aerial combat games, (c) the baseball games were to the individual combat games, (d) the football games were to the aerial combat games, (e) the football games were to the individual combat games, and (f) the aerial combat games were to the individual com bat games. Subjects also assessed the similarity of the different subcategories. These measures could verify that alternatives with more alignable attributes were perceived as more similar and that consideration sets composed of these alternatives were mor e homogeneous. Results Eighty four subjects from an introductory marketing course subject pool were given extra credit to participate in the study. The comparability and similarity ratings confirmed that the alignability manipulation was effective. Simila r subcategories ( M = 4.36) were rated easier to compare than dissimilar subcategories ( M = 3.07; F (1, 83) = 14.77, p < .05). Also, the similarity ratings of moderate similarity subcategories ( M = 4.33) were significantly higher than the similarity ratings of dissimilar subcategories ( M = 3.08, F (1, 83) = 15.10, p < .05). To test the hypothesis that people include more comparable alternatives in their consideration set, I created a ratio of the number of highly comparable, moderately comparable, and non com parable pairs in each personâ€™s consideration set to the number of possible pairs of each type (i.e., the base rate). Among the 16 alternatives in each
16 stimulus set, there were 24 highly comparable, 32 moderately comparable and 64 non comparable possible pa irs. Please see Appendix C for a detailed description of the dependent variable based on a hypothetical example of consideration sets constructed by three subjects. The data supported the hypothesis that people naturally create more homogeneous considerat ion sets. The proportion of highly comparable pairs ( r = .13) was significantly higher than the proportion of moderately comparable pairs ( r = .08) included in the consideration sets ( Z = 9.10, p < .05). Likewise, the proportion of moderately comparable pa irs ( r = .08) was significantly higher than the proportion of non comparable pairs ( r = .05) included in the consideration sets ( Z = 8.29, p < .05). The statistics for each of the three replicates used in the study is reported in Appendix D. Discussion The results of study 1 are consistent with the hypothesis that people naturally create considerations sets of homogeneous alternatives. Subjects were more likely to consider alternatives that were comparable (i.e., had alignable attributes) than alternatives that were non comparable (i.e., had non alignable attributes). In other words, alignable pairs of alternatives were more likely to co occur in a consideration set than non alignable pairs of alternatives. I propose that consumers prefer to create considera tion sets of alternatives having alignable attributes because these alternatives are easier to compare at the time of final choice. Despite support for the hypothesis that people create consideration sets of alternatives that are easier to compare, there a re three potential procedural and conceptual problems with study 1. First, it could be argued that the computer presentation format
17 encouraged consumers to rely on their memory to form consideration sets. It is easier to remember alternatives described on common attributes, hence the observed concentrations of easily comparable alternatives may really be evidence that easily remembered, or easily accessible alternatives enter consideration sets. Second, the use of well known brand names for some of the repl icates may have led to consideration set inclusion decisions based on strong likes or dislikes towards particular brands. To the extent these well known brand names were clustered in a subcategory or game type, they could have encouraged the selection of c omparable alternatives. Third, the consideration set homogeneity may be a consequence of preference for certain types of products. If subjects had strong likes or dislikes for certain subcategories or product â€œtypesâ€ within each replicate, then elimination of entire subcategories and product types would naturally lead to retention of a more homogeneous set of alternatives. For example, if a subject's dislike for violence led to the natural exclusion of the combat video games from the outset, then the altern atives remaining for consideration would automatically be relatively more homogeneous. To address these limitations three additional studies were conducted which are discussed next. Study 1A To address the concern that the results of study 1 were sensitive to memory processes, the study was rerun using sixteen stimulus cards. These stimulus cards could be easily laid out on the table and simultaneously aligned and compared. This procedure significantly reduces the potential for memory based effects noted ea rlier. Twenty five subjects participated in the study. Barring the use of stimuli cards, the procedure was identical to that of study 1. Similar to study 1, the proportion of highly comparable pairs
18 ( r = .12) was significantly higher than the proportion of moderately comparable pairs ( r = .05) included in the consideration sets ( Z = 6.79, p < .05). Likewise, the proportion of moderately comparable pairs ( r = .05) was significantly higher than the proportion of non comparable pairs ( r = .03) included in the consideration sets ( Z = 3.68, p < .05). Please refer to Appendix D for results specific to each replicate. Study 1B To address the concern that a concentration of well known brand names in sub categories or game types could have encouraged the selection of comparable alternatives for further consideration, I reran study 1 using alphanumeric brand names. Barring the use of alphanumeric names (e.g., A1, A2 etc.) the procedure was identical to that used in study 1. Thirty seven subjects participated in the stu dy. The results were very similar to what we observed earlier. The proportion of highly comparable pairs ( r = .09) included in the consideration sets was significantly higher than the proportion of moderately comparable pairs ( r = .05; Z = 5.11, p < .05). Likewise, the proportion of moderately comparable pairs ( r = .05) included in the consideration sets was significantly higher than the proportion of non comparable pairs ( r = .04; Z = 3.06, p < .05). For detailed results pertaining to each replicate, pleas e refer to Appendix D. Study 1C Study 1c addresses the concern that the consumerâ€™s preference for homogeneous consideration sets could be a consequence of strong likes or dislikes for certain subcategories of products. To address this concern, I manipulate d the attribute alignability between alternatives using three between subject conditions that had equivalent stimulus labels and attributes (see Appendix E). If the stimulus labels and
19 attributes remained constant, then any influence of attribute alignabil ity on consideration set homogeneity could not be attributed to preference for a certain type of product. Moreover, random allocation of subjects to different experimental conditions ensures that likes and dislikes towards particular subcategories are not differentially distributed across the three conditions. Thus any changes in the consideration set composition that we may observe can be solely attributed to changes in alignability, and not affect towards certain subcategories of products. The stimuli wer e ten alternatives, five from one subcategory (e.g., baseball) and five from a second subcategory (e.g., football), for four replicates (video games, cars, restaurants, and vacations), three of which were used in study 1. In the high cross category compara bility condition, alternatives in each subcategory had four alignable attributes (e.g., baseball and football games were both described on trading/draft, stadium graphics, game analysis, and physical impact) and three unique attributes (e.g., baseball game s were additionally described on batting stance, pitching deliveries, and catches, whereas football games were additionally described on play books, offensive formations, and defensive formations). In the moderate cross category comparability condition, th e alternatives in the two sub categories had two alignable attributes and four unique attributes. In the non comparable condition, alternatives in each subcategory had no alignable attributes and five unique attributes. The procedure was identical to study 1 except 65 subjects were asked to form consideration sets of four alternatives. Analysis of the manipulation checks verified that it was easier to compare alternatives from different subcategories in the high cross category comparability condition ( M = 6.43) than in the moderate cross category comparability condition ( M =
20 5.41; F (1, 62) = 3.98, p < .05). It was also easier to compare alternatives in the moderate cross category comparability condition ( M = 5.41) than in the non comparable condition ( M = 3 .51; F (1, 62) = 7.36, p < .05). The similarity ratings also followed a similar pattern. Alternatives from different subcategories were seen as more similar in the high comparability condition ( M = 6.05) than in the moderate comparability condition ( M = 4.6 8, F (1, 69) = 4.75, p < 0.01). Likewise, alternatives from different subcategories were seen as more similar in the moderate comparability condition ( M = 4.68) than in the low comparability condition ( M = 3.51, F (1, 69) = 5.55, p < 0.01). Tests of the hypo theses showed the alignability manipulation had a significant influence on the degree to which alternatives from both sub categories were included in the consideration set ( F (2, 62) = 91.87, p < .05). The number of cross category pairs (e.g., baseball foot ball pairs) that entered the consideration sets in the high cross category comparability condition ( M = 3.83) was significantly more than in the moderate cross category comparability condition ( M = 2.95; F (1, 62) = 11.42, p < .05). Likewise, the number of cross category pairs that entered the consideration sets in the moderate cross category comparability condition ( M = 2.95) was significantly greater than in the non comparable condition ( M = .89; F (1, 62) = 61.47, p < .05). These results show that as the c omparability of two subcategories (e.g., football and baseball) of games increased, people were more likely to include games from both subcategories in their consideration set.
21 CHAPTER 4 STUDY 2: THE ROLE OF SIMILARITY In the first set of studies, I pr ovide evidence that people naturally create homogeneous consideration sets. I argued that the desire to have alternatives that are easy to compare encourages the creation of these homogeneous consideration sets. Yet, it is important to recognize that there may be other factors beyond attribute alignability that promote comparability. For example, people may find it easier to compare similar alternatives independent of their alignability (Medin et al. 1995). When faced with a set of objects subjects often us e the similarity between the objects as an organizing principle in order to reduce information load and facilitate further information processing (Tversky 1980). Thus, in study 2, I manipulated the similarity of alternatives while holding attribute alignab ility constant. In this way, I could provide evidence that similarity also promotes the comparability of alternatives. Design, Stimuli, and Procedure The hypothesis was tested using a within subject design with five replicates (apartments for rent, automob ile radar detectors, vacuum cleaners, microwave ovens, and air conditioners). Subjects saw twenty alternatives that were described on a common set of six binary features (see Appendix F). Each alternative had three features that were present and three feat ures that were absent. Similarity was varied by manipulating the degree of feature overlap among pairs of the alternatives. Following Tversky (1980), for the purpose of this study, I define similarity between any two alternatives, say A and B,
22 to be a func tion of three things: (a) the number of features common to both A and B, (b) the number of features unique to A, and (c) the number of features unique to B. Specifically, the degree of feature overlap was determined by the Jaccard ratio of similarity (see Appendix G for details). The Jaccard ratios for high, moderate and low similarity pairs were 0.5, 0.2 and 0.0 respectively. Our hypothesis was that similar options, like options A and B (see Appendix F), were more likely to enter the consideration set, tha n dissimilar options, like options A and D (see Appendix F). Our prediction is based on the fact that the dissimilar options, though perfectly alignable, are still difficult to compare. The difficulty in comparison stems from the fact that in order to be a ble to compare options with very little feature overlap (like A and D), consumers need to translate features into their corresponding benefits. This is an effortful process, and it is much easier to compare a set of alternatives that have a fair degree of feature overlap. Note that the stimulus sets contained 90 high similarity pairs, 90 moderate similarity pairs, and 10 low similarity pairs, so the dependent measure was a ratio of the number of each type of pair in the consumerâ€™s consideration set relative to the possible number of each type of pair. The procedure was identical to study 1 except for small changes in the manipulation check. As a manipulation check subjects were asked to rate how similar they thought (a) the members of the consideration set w ere to each other, and (b) the members of the consideration set were to the options that were not short listed. The manipulation check would verify that options included in the consideration set were indeed perceived as significantly more similar to each o ther than the options that were excluded.
23 Results Seventy seven subjects from an introductory marketing course subject pool were given extra credit to participate in the study. The results are consistent with the hypothesis that people naturally create con sideration sets consisting of similar alternatives. A greater proportion of high similarity pairs ( r = .06) entered the consideration sets than moderately similarity pairs ( r = .01; Z = 32.77, p < .05). The proportion of moderate similarity pairs ( r = .01) occurring in the consideration sets was also significantly greater than the proportion of the low similarity alternatives ( r = .002, Z = 10.86, p < .05). Detailed results for each replicate are provided in Appendix H. The manipulation check comprising a p aired samples t test, confirmed that the perceived similarity of the alternatives within the consideration sets ( M = 6.43) was significantly higher than the perceived similarity between alternatives within the consideration sets and alternatives outside th e consideration sets ( M = 4.92; t (76) = 8.16, p < .05). Discussion Study 2 provides evidence that people prefer to create consideration sets of similar alternatives, even though all of the alternatives were described on the same attributes (i.e., attribute s were alignable). We believe that this desire for similarity can be attributed to consumers wanting to avoid the need to create a common currency in order to compare alternatives. When options have very few overlapping features, the consumer has to transl ate attributes into benefits in order to compare alternatives. This is an effortful process. It is much easier to compare a set of alternatives that have a fair amount of feature overlap.
24 There is one final alternative explanation that the current design d oes not address. The pattern of results observed in study 2 could also have been generated if subjects used a strategy of reverse elimination by aspects. In other words, a strong preference for any two features, and subsequent elimination of alternatives t hat did not possess those two features, could also have lead to retention of highly similar alternatives. Since we did not have specific measures of attribute importance in the main experiment, it was not possible for us to rule out this alternative explan ation. Study 2a was conducted to investigate this alternative explanation. Study 2A Study 2a was designed to demonstrate that the homogeneous consideration sets observed in study 2 were a function of the consumerâ€™s desire to have easily comparable alternat ives as opposed to the consumerâ€™s use of a reverse elimination by aspects strategy. Our strategy for providing this evidence was similar to the strategy used in study 1c, i.e., manipulate the degree of feature overlap between alternatives in a between subj ects design. In study 1c, we saw that increasing the alignability between two sets of dissimilar items (e.g., baseball and football video games), led to more heterogeneous consideration sets. In the current study, we manipulated the similarity between two sets of dissimilar alternatives while keeping the similarity within each set of alternatives constant. Design, Stimuli, and Procedure The stimuli were 30 alternatives, 15 from one cluster and 15 from a second cluster, described on 12 features (see Appendi x I) for four replicates (automobile radar detectors, vacuum cleaners, microwave ovens, and air conditioners) used in study 2. The
25 average Jaccard ratio of similarity between alternatives within a cluster was held constant at 0.48 across all conditions. Th e two identifiable clusters occurred because the alternatives 1 15 were more likely to have features F1 F6, while alternatives 16 30 were more likely to have features F7 F12. The key manipulation was the Jaccard ratio of similarity between alternatives acr oss these two clusters. The average Jaccard ratio of similarity between alternatives across clusters was manipulated to be 0.29 in the high similarity condition, 0.11 in the moderate similarity condition, and 0.0 in the low similarity condition. For exampl e, in the high similarity condition there were six overlapping features across the two clusters (e.g., alternatives 1 15 were more likely to have features F1 F9 and alternatives 16 30 were more likely to have features F4 F12), in the moderate similarity co ndition there were two overlapping features across the two clusters (e.g., alternatives 1 15 were more likely to have features F1 F7 and alternatives 16 30 were more likely to have features F6 F12), and in the low similarity condition there were no overlap ping features across the two clusters. The experimental procedure was similar to study 1c. The primary difference was people were asked to rank features in the order of their importance at the beginning of the experimental session. The feature rank informa tion was stored and then used to determine the feature labels used in the stimulus matrix. The ranking information was used to assign the feature labels in a manner such that (a) both clusters of alternatives were roughly equally preferred, and (b) the ove rlapping features in the moderate and high similarity conditions were the least important features. More specifically, the clusterâ€™s features were assigned so that the features were likely to be irrelevant if the consumer used a reverse elimination by aspe cts strategy. In the high similarity condition, the six overlapping
26 features (F4 through F9) were the six least important features. In the moderate similarity condition, the two overlapping features (F6 and F7) were the two least important features. In the low similarity condition, the features were distributed across the matrix so that the two clusters of alternatives were roughly equally preferred (e.g., features ranked 1st, 3rd, 5th, 7th, 9th and 11th were assigned to the first cluster, while the feature s ranked 2nd, 4th, 6th, 8th, 10th and 12th were assigned to the second cluster). The only other difference from study 1c was people were asked to create consideration sets of size eight. For the manipulation check, we were interested in verifying that the feature overlap manipulation had the desired effect on the perceived similarity and comparability of the alternatives in the different clusters. Subjects were shown the original stimuli matrix and asked how similar and comparable they found alternatives 1 15 to alternatives 16 30. Results Sixty seven subjects from an introductory marketing course subject pool were given extra credit to participate in the study. The manipulation check verified that alternatives in different clusters were more similar and ea sier to compare in the high similarity condition than both in the moderate similarity condition and in the low similarity condition. Alternatives from different clusters were more similar in the high similarity condition ( M = 5.04) than in the moderate sim ilarity condition ( M = 3.56; F (1, 64) = 23.62, p < .05). Alternatives from different clusters were more similar in the moderate similarity condition ( M = 3.56) than in the low similarity condition ( M = 2.79; F (1, 64) = 5.71, p < .05). The comparability rat ings also displayed an identical pattern of
27 results ( M high = 5.53, M moderate = 4.54, M low = 3.42; F (1, 64) = 6.60, p < .05; F (1, 64) = 7.51, p < .05). The results were generally consistent with our predictions. The overlap manipulation had a significant ma in effect ( F (2, 64) = 32.19, p < .05) on the number of dissimilar pairs that were considered. Planned contrast tests revealed that the number of across cluster pairs that entered the consideration set in the high cross category comparability condition ( M = 13.93) was significantly higher than those in the moderate similarity condition ( M = 6.47; F (1, 64) = 33.64, p < .05). However, the number of across clusters pairs that entered the consideration set in the moderate similarity condition ( M = 6.47) was not significantly greater than in the low similarity condition ( M = 4.51; F (1, 64) = 2.95, p > .10). Discussion The results of study 2a show that, for any two objects, the probability of their co occurrence in the consideration set appears to depend on the ex tent of feature overlap between the two alternatives. When faced with alternatives of varying similarity (feature overlap), consumers form consideration sets that are relatively homogeneous in order to facilitate comparison in the final choice task. The ex tent to which dissimilar alternatives are included in the consideration depends on the extent of feature overlap between these dissimilar alternatives. The data from study 2a also rule out the alternative hypothesis that consideration set homogeneity is a function of the consumerâ€™s use of a reverse elimination by aspects selection strategy. The studies presented thus far show that people have a preference for creating consideration sets consisting of alternatives that are easy to compare. We argue that the
28 ease of comparison can occur because there is attribute alignability across alternatives or because alternatives are similar. Earlier, we also argued that consumers can compare dissimilar alternatives, but only when there is a common currency (i.e., people translate attributes into benefits). Thus, if consumers are encouraged to put in the extra effort and time required to create these common currencies, it is quite likely that consideration sets will consist of more heterogeneous alternatives. In other wor ds, there should be a negative relationship between the amount of time spent creating a consideration set and the homogeneity of the consideration set. This issue is investigated next.
29 CHAPTER 5 STUDY 3: THE EFFECT OF TIME PRESSURE In study 3, we manipul ated the amount of time available to create a consideration set using the stimulus set from study 1 and study 1c. Following standard procedures used in psychological research (e.g., see Maule, Hockey, and Bdzola 2000), in the low time pressure condition, s ubjects were encouraged to take "as much time as possible" to form their consideration sets. In the moderate time pressure condition, subjects were told (a) the average time that people took to complete the consideration set, and (b) that they had the same amount of time as the average person. In the high time pressure condition, subjects were told (a) the average time that people took to complete the consideration set, and (b) that they had about half the amount of time as the average person. The average t imes used in the latter two conditions were based on the average time subjects used to complete their consideration sets in study 1. Procedure The experimental procedures were similar to those of study1, barring (a) the instructional manipulations mentione d above, (b) a request to create consideration sets of five alternatives, and (c) the use of countdown clocks in the moderate and high time pressure conditions in order to reinforce the time pressure manipulation. Most (81%) of the 58 subjects assigned to the moderate and high time pressure conditions completed their consideration sets within the specified time limits. For those who exceeded the time limit, a small icon popped up on the screen informing them that the time limit had
30 expired. Further, they we re told that I would nevertheless like them to complete their consideration set. Since the primary manipulation involved varying the "felt" time pressure, the manipulation check in the final task consisted of a single question that asked the subjects to in dicate the degree to which they felt hurried while creating the consideration set. Results Eighty eight subjects from an introductory marketing course subject pool were given extra credit to participate in the study. The experimental manipulation had the desired effect with regard to the felt time pressure across the three conditions. Subjects felt more hurried in the high time pressure condition (M = 6.61) in comparison to the moderate time pressure condition (M = 5.00, F (1, 84) = 2.88, p < 0.05). Subject s also felt more hurried in the moderate time pressure condition (M = 5.00) than in the low time pressure condition (M = 3.17, F (1, 84) = 2.48, p < 0.05). There was a significant effect of the time pressure manipulation on the number of high similarity alt ernatives that were considered by the subjects ( F (2, 79) = 108.27, p < .05). The number of high similarity pairs in the consideration set was significantly greater in the high time pressure condition ( M = 5.44) than in the moderate time pressure condition ( M = 4.51; F (1, 79) = 36.48, p < .05). Likewise, the number of high similarity pairs in the consideration set was significantly greater in the moderate time pressure condition ( M = 4.51) than in the low time pressure condition ( M = 3.12; F (1, 79) = 79.21, p < .05). Discussion The results verify the negative relationship between the amount of time spent in creating a consideration set and its underlying homogeneity. As consumers spend more
31 time and effort in creating their consideration set they are more like ly to create the common currency needed to compare dissimilar items. For example, consumers who spend more time in creating a consideration set of video games are more likely to come up with a common currency measure like "excitement" in order to compare a baseball video game that has a "120 batting stances" to an aerial combat game that has "98 flying missions.â€ In fact, there is evidence that our subjects were more likely to create common currencies when they had more time to create the consideration set. Analysis of the open ended responses for video games revealed that the creation of common currency measures was indeed more common in the low time pressure condition. Significantly more common currency attributes were mentioned in the low time pressure co ndition ( r = .69) than in the moderate time pressure condition ( r = .31; Z = 4.86, p < .05). Significantly more common currency attributes were mentioned in the moderate time pressure condition ( r = .31) than in the high time pressure condition ( r = 0; Z = 3.61, p < .05). One of the major remaining issues is the assumption that consumers have an ease of comparison motive. Although it is difficult to directly test for this motive, we can certainly think of situations in which ease of alternative comparison is less likely to be an active motive. For example, when consumers are made sensitive to the possibility of committing a type II error (i.e., erroneously leave out a good alternative), they may become less motivated to have easily compared alternatives. T o the extent we can align the motive to avoid a type II error with the selection of a more heterogeneous consideration set, we should be able to show how altering a general motive alters the composition of a consideration set. We investigate this possibili ty in the next study.
32 CHAPTER 6 STUDY 4: THE EFFECT OF NEGATIVELY CORRELATED BENEFITS In study 4, we created a situation in which consumers were concerned about leaving a potentially optimal alternative out of their consideration set. The situation invo lved competing clusters of alternatives that had a negatively correlated benefit structure. To illustrate this idea, reconsider the video game stimuli used in study 1. In study 1, all 16 alternatives were described on attributes that had two equally salien t underlying benefits -realism and challenge (see Appendix J). Both types of video games (i.e., sports and combat video games) were equally endowed with features related to realism and challenge. This meant that subjects could create consideration sets of easily comparable alternatives without being afraid that they might leave an optimal game out of their consideration set. In contrast, consider an alternative set where all sports games are high on challenge but low on realism and all combat games are low on challenge but high on realism. Under these conditions, consumers are likely to become sensitive to the fact that a homogeneous consideration set could result in a consideration set that lacked alternatives having one of these two important benefit dimen sions. As a result, consumers become more likely to retain heterogeneous items in their consideration set. Design and Manipulation A 3 (benefits tradeoff) level by 2 (replicates) level mixed design was used to test the hypothesis. Appendix K illustrates the design using the video game example. In the no tradeoff condition, all alternatives contained features consistent with both benefits.
33 However, in the benefits tradeoff conditions, one subcategory of alternatives was high on one benefit and low on the o ther, whereas the reverse was true for the other subcategory. For example, in the video game replicate, we varied the attribute values such that the sports video games were rich in terms of challenge but poor in regard to realism, while the combat video ga mes were rich in realism but poor in terms of challenge (Version1). The third condition simply reversed this pattern of benefits tradeoff (Version2). Stimuli, Procedure, and Dependent Measures The stimuli were identical to those used in study 1 for the vi deo games and vacations replicates, barring the attribute value manipulations in the benefits tradeoff conditions. The procedure used was also identical to that of study 1, barring (a) the random assignment of subjects to the three different experimental c onditions (study 1 had a within subject design), (b) instructions to create a consideration set of four alternatives, and (c) a manipulation check that consisted of a single question asking subjects to indicate the extent to which they were aware of a trad eoff between the two benefit dimensions. The key dependent measure involved comparing the number of high similarity pairs (out of the possible six pairs) in the consideration set, across the three experimental conditions. Results Fifty one subjects from an introductory marketing course subject pool were given extra credit to participate in the study. There was no difference between the two negatively correlated conditions, for either the dependent measure (M Version1 = 3.75, M Version2 = 3.83, F (1, 46) = 0.2 7, p > 0.10) or the manipulation check (M Version1 = 6.14, M Version2 = 6.33, F (1, 48) = 0.30, p > 0.10), so data are collapsed from these two conditions.
34 The manipulation check confirmed that subjects perceived a greater tradeoff between the two benefit di mensions in the benefits tradeoff conditions ( M = 6.23; F (1, 49) = 5.97, p < .05) than in the no tradeoff condition ( M = 5.12). Further, the number of low similarity pairs in the benefits tradeoff conditions ( M = 3.79) was significantly higher than in the no tradeoffs condition ( M = 1.96; F (1, 49) = 67.44, p < .05). Consumers considered a more diverse set of alternatives when the two clusters of alternatives had negatively correlated benefits. Discussion When faced with a benefits tradeoff among alternative s, consumers have two tasks that need to be accomplished: (a) compare scale values, and (b) resolve the benefit weights problem. The former involves comparing the attribute values of the various alternatives and selecting those that have high attribute val ues. The latter refers to the tradeoff between the two benefit dimensions and involves a resolution of this tradeoff. The results show that when faced with a tradeoff, subjects postpone resolving the weights issue until the final choice stage. It is not su rprising that they choose to do so because the latter task is more effortful to accomplish and consumers typically partition their choice task such that the more effortful comparisons are left for the final choice stage (e.g., see Ratneshwar, Pechmann and Shocker 1996). Consumers prefer to invest relatively less effort during the consideration stage. Thus when faced with a benefits tradeoff, heterogeneous items are more likely to enter the consideration set. Study 4A There is a potential limitation to the r esults observed in study 4. It is possible that the non alignability of the alternatives in the two sub categories encouraged people to
35 create common currencies (e.g., benefits) and this in turn increased their sensitivity to the negatively correlated bene fit structure. In other words, people may have avoided resolving the benefit trade off problem because the stimuli were constructed to highlight this problem. To rule out this possibility, we reran study 4 using the feature overlap stimuli used in study 2. Recall that the stimuli from study 2 had alignable features, some overlapping and some non overlapping. We wanted to assess if people remain sensitive to negatively correlated benefits even when the stimuli are alignable and only differ with respect to th eir similarity (i.e., the extent of feature overlap). We retested our negatively correlated benefits hypothesis using a 2 (benefits tradeoffs) by 3 (replicates) level mixed design. We can illustrate the key experimental manipulation using the sedan example in Appendix L. The stimuli consisted of eight alternatives described on six features with â€œYes/Noâ€ labels indicating the presence or absence of a feature (see top matrix). Each alternative had three features that were present and three features that were absent. Alternatives one through four were highly similar to each other, as were alternatives five through eight (0.5 Jaccard ratio of similarity). The average across cluster Jaccard ratio of similarity was 0.2. Of the six features chosen for each replica te, three represented one benefit dimension while the remaining three represented a second benefit dimension. For example, for the replicate sedans, three features corresponded to economy (e.g., fuel efficiency, warranty and reliability) and three correspo nded to comfort (e.g., good ride, legroom and rapid acceleration). In the condition without any tradeoffs, the economy and comfort features were randomly distributed such that they occurred in every alternate column (see second matrix in Appendix L). This distribution of features ensured that both
36 clusters of options were equally endowed with comfort and economy related features. In the condition with tradeoffs, the features were distributed such that the economy features occurred in the first three columns and the comfort features occurred in the last three columns (see third matrix in Appendix L). In effect, this led to one cluster (alternatives one four) being high on the comfort dimension, and the other cluster (alternatives five eight) being high on the economy dimension. Thus, this condition had a negatively correlated benefit structure. The replicates consisted of sedans, MBA schools and apartments for rent. The features and the underlying benefits for each replicate is provided in Appendix M. Forty se ven subjects participated in the study. Subjects again created consideration sets of four alternatives. The results were consistent with our hypothesis. First, the manipulation check shows that subjects perceived a greater degree of tradeoff between the t wo benefit dimensions in the trade off condition ( M = 6.73) than in the no tradeoff condition ( M = 5.08; F (1, 45) = 5.23, p < .05). Further, the number of low similarity pairs was higher in the tradeoff condition ( M = 3.36) than in the no tradeoff conditio n ( M = 2.42; F (1, 45) = 20.73, p < .05). This means that approximately 60% of the pairs in the consideration set tend to be low similarity pairs and the remainder high similarity pairs in the tradeoff condition, whereas 40% of the pairs tend to be low simi larity pairs with the remainder high similarity pairs in the no tradeoff condition. Thus, alternative sets with benefit tradeoffs appear to reduce the motivation to create homogeneous consideration sets. At this juncture it is important to note that not al l benefit tradeoffs will lead to this pattern of results. A key requisite to replicating this pattern of results is that the two benefits need to be equally salient or important to the consumer.
37 CHAPTER 7 STUDY 5: THE EFFECT OF TIME PRESSURE AT CONSIDERAT ION VERSUS FINAL CHOICE STAGES In study 4, we showed that people create heterogeneous consideration sets when they become concerned about leaving a potentially optimal alternative out of their consideration set. When benefits are negatively correlated ac ross alternatives, people have to make decisions about the relative importance of each benefit. This is an effortful task, so they create heterogeneous consideration sets and leave the more effortful comparisons to the final choice stage. If our reasoning is correct, then the relative heterogeneity of a consideration set should depend on the consumerâ€™s anticipation of the time available to form the consideration set and the time available to make the choice. For example, when the consumer anticipates havin g very little time for consideration, but a lot of time to make the final choice, the consumer should create a more heterogeneous consideration set. However, when a consumer anticipates having very little time during the final choice stage, but a lot of ti me for consideration, the consumer should create a more homogeneous consideration set. Design and Manipulation The hypothesis was tested using a 2 (anticipated time pressure) by 7 (replicates) level mixed design. Anticipated time pressure was manipulated using instructions as well as countdown clocks embedded in the computerized procedure. In one condition subjects
38 were told that although they would have very little time (e.g., two minutes) for creating their consideration set, they could have as much time as they wanted to make their final choice. In another condition subjects were told that although they could have as much time as they wanted for creating their consideration set, they would have very little time (e.g., two minutes) to make their final cho ice. Stimuli and Procedure Of the seven replicates, two were the tradeoff stimuli used in study 4 (e.g., video games and vacations) and five were tradeoff stimuli similar to the ones used in study 4a (e.g., sedans, beach vacations, MBA schools, laptops, an d apartments). The two replicates from study 4 used alignable feature descriptions and five replicates similar to those used in study 4a used binary feature descriptions. The procedure was also identical to the experimental procedures used in study 4. Subj ects selected four alternatives to include in their consideration set. Results Twenty six subjects from an introductory marketing course subject pool were given extra credit to participate in the study. When consumers experienced high time pressure during consideration set formation and anticipated low time pressure during choice, there was an average of 2.13 similar pairs in their consideration set. This is similar to the results of the tradeoff conditions in studies 4 and 4a. When consumers experienced lo w time pressure during consideration set formation and anticipated high time pressure during choice, the number of similar pairs in their consideration sets significantly increased to an average of 3.80 pairs ( F (1, 24) = 57.33, p < .05).
39 Discussion Manipu lating whether the consumer experienced time pressure at consideration set formation or anticipated there would be time pressure at choice had a significant influence on how consumers reacted to a benefit tradeoff in the set of alternatives. When consumers anticipated they would have a lot of time at choice, they created heterogeneous consideration sets and delayed the difficult task of resolving which benefit was more important. When consumers anticipated they would have little time at choice, they made de cisions about which benefit was more significant prior to forming the consideration set and, thus created more homogeneous consideration sets. Thus a situational variable like the anticipated time pressure at consideration versus choice could play an impor tant role in determining the composition of a consideration set. Moreover, these results confirm the process explanation that we offered earlier regarding how consumers partition their choice task. Recall, that when faced with a benefits tradeoff subjects have two tasks to accomplish: resolve the "weights" to be attached to the two benefits on which a tradeoff exists, and compare "scale values.â€ The pattern of results observed here seem to arise from the fact that subjects choose to perform the more effortf ul task (i.e., resolving the "weights" issue) when the time constraint is the least. Thus the nature of task partition drives the composition of the consideration set.
40 CHAPTER 8 GENERAL DISCUSSION Studies 1 through 2a provide strong evidence that c onsideration set formation not only involves reducing the number of alternatives to a more manageable size, but also involves retaining alternatives that are easy to compare during the final choice stage. This motivation to have easily compared alternative s encourages people to retain alternatives that are alignable and to retain alternatives that are similar. As a consequence, people have a natural tendency to create homogeneous consideration sets. This may be one reason consumers often consider brands tha t bunch together in perceptual space (e.g., Lattin and Roberts 1992, Lehmann and Pan 1994). Studies 3 through 5 show there are also situations that motivate people to create heterogeneous consideration sets. First, when people are able to create common cu rrencies (e.g., benefits) that allow them to compare dissimilar alternatives, they become more likely to include dissimilar alternatives in their consideration sets. In effect, alternatives become similar at a higher level of abstraction. Second, when peop le want to avoid type II errors (i.e., errors of exclusion), they often include dissimilar alternatives in their consideration sets. Consumers do not want to exclude an alternative that could be judged the most preferred at the time of choice, even though that choice is often minutes away. This motive becomes particularly salient when different clusters of alternatives pose a benefit tradeoff (i.e., one set of alternatives delivers benefit one but not benefit two, whereas a second set of alternatives deliv ers benefit two but not benefit one).
41 However, this effect can be moderated by anticipated time pressure at the consideration stage or choice stage. Study 5 showed that consumers anticipating ample time at the choice stage delayed resolving their conflict about which benefit they preferred and created heterogeneous consideration sets, whereas consumers anticipating limited time at the choice stage resolved their conflict about which benefit they preferred and created homogeneous consideration sets. Some o f the studies reported in this dissertation (especially the studies highlighting the role of alignability) help resolve some of the conflicting findings reported earlier. Recall that Hauser and Wernerfelt (1989) found the assumption of probabilistic indepe ndence to be a reasonable one in their data set, while Roberts and Lattin (1991) rejected probabilistic independence. In other words, Hauser and Wernerfelt (1989) reported evidence supporting heterogeneous consideration sets, while Roberts and Lattin (1991 ) found evidence for relatively more homogeneous consideration sets. It is possible that this conflicting pattern of results could be explained by the differences in alignability across the two data sets. Hauser and Wernerfelt (1989) used data on the plast ic wraps market, while Roberts and Lattin (1991) used data from the ready to eat cereals market. While the plastic wraps market had a relatively simple two segment structure comprising "national" and "store" brands, the ready to eat cereal category had fou r clearly identified clusters: â€œhealthful,â€ â€œartificial,â€ â€œinteresting,â€ and â€œnon adult.â€ Thus it is quite likely that the degree of alignability between alternatives was much higher in the plastic wraps market in comparison to the highly segmented ready t o eat cereal market. This is consistent with the findings in study 1 1c, where consideration set heterogeneity was found to covary with the degree of alignability between alternatives.
42 These studies also offer several important implications for marketing p ractitioners. From the retailer's point of view, adding alignable attributes to existing product descriptions may encourage heterogeneity in consumer considerations sets, and as a consequence, ensure relatively fast turnarounds for different types of inven tory and stock. For example, a category like wines is characterized by non alignable descriptions. It is becoming more common to find retailers (e.g., the internet based â€œwine.comâ€) and information sources (e.g., Wine Spectator ) that provide â€œtasting chart sâ€ describing even the most dissimilar wines on a common set of attributes (e.g., acidity, tannin). These descriptions increase benefit alignability and have the potential to encourage inclusion of otherwise diverse alternatives in the consumerâ€™s considera tion set. From a market competition point of view, these studies underscore the role of alignability in promoting or dampening competition. Manufacturers may actively reduce the alignability between their products and competing alternatives and at the sam e time create alignability between their own branded variants. For example, manufactures use irrelevant attributes (e.g., â€œMountain grownâ€ for Folgers Coffee) to reduce alignability and discourage consumers from considering competing alternatives. Manufact urers of appliances, computers, and toys create product lines that encourage comparison of alignable alternatives within the product line instead of between competing lines. This strategy may be a particularly effective when novice consumers are gathering product attribute information at the time of purchase. Finally, this paper provides a framework for anticipating the nature of the composition of the consideration set. For example, categories characterized by negatively correlated benefit structures (e.g ., computers, apartments for rent) may benefit from a
43 more heterogeneous display of alternatives. However, if the consumer anticipates being under relatively high time pressure at the final choice stage, then it is likely that (s)he will consider a more ho mogeneous set of alternatives. It is important to anticipate the nature of the consideration set because previous studies (e.g., Schlosser and Kanfer 1999) have shown that matching the buying context (e.g., the displayed alternatives) to consumer expectati ons has a favorable impact on consumer purchase intentions, and repeat purchases.
44 CHAPTER 9 LIMITATIONS AND FUTURE RESEARCH Limitations While this dissertation takes an important first step towards understanding the determinants of consideration set co mposition, there are several issues that remain unexamined. First, there are several characteristics of the experiments that limit the generalizability of the findings. For example, all the studies were restricted to those situations in which consideration set formation is primarily stimulus based. However, it is quite well known that consideration sets can be both stimulus based, as well as memory based. In reality, consumers are most likely to use a combination of both these methods to form consideration sets, thus limiting the generalizability of the findings discussed in these studies since all the studies were based on stimulus based scenarios. Another limiting characteristic of these studies was that all information on all the alternatives was made ava ilable to the subjects simultaneously. Contrary to such a scenario, it is fairly common practice for consumers to look at alternatives in a sequential fashion and use only a partial amount of all the available information to create a consideration set. In short, there is usually a difference in the amount of information that is available during the consideration and choice phases, a fact that is not addressed by the ten studies presented in this dissertation. Second, the studies presented here do not clari fy the relationship between similarity and comparability. In most of the studies the similarity ratings mimicked the
45 comparability ratings, i.e., alternatives that were similar to each other were also found to be difficult to compare. This leads to an impr ession that similarity between objects and their ease of comparison are necessarily highly correlated. However, similarity and comparability remain conceptually distinct constructs and it is perhaps not too difficult to imagine situations in which similari ty and comparability do not covary. It is important to clarify the inter relationship between these two concepts since it is likely to have a direct bearing on the nature of the consideration set content. Future Research Two extensions of this research ha ve the potential to be useful. First, it would be useful to understand the influence of the alignability and similarity of the alternative set on the size of a consumerâ€™s consideration set. As alternatives become more alignable or similar, they become easi er to compare. As alternatives become easier to compare, consumers should be willing to compare more alternatives. Thus, we might expect that there is more brand switching behavior, an indicator of larger consideration sets, in highly alignable product cat egories. We might also find larger product lines, an attempt to block competing products from being considered, in product categories having more alignable product attributes. Second, it would be useful to understand the potential malleability of consider ation sets. Desai and Hoyer (2000) find that consideration sets become more stable as consumers gain more experience (see also Johnson and Lehmann 1997). Our research shows that consideration set composition is a function of the characteristics of the set of alternatives. We have little insight into how individual differences related to product experience will interact with characteristics of the
46 alternative set during consideration set formation. For example, it may be that experienced consumers have well established preferences that lead to a small, focused consideration set (selecting a movie). It may also be that experienced consumers can easily create a common currency that allows them to compare a heterogeneous set of alternatives (e.g., selecting a d inner entree, buying wine). The factors that encourage a consumer to rely on pre existing preferences or alternative set characteristics during consideration set formation is worthy of further study.
47 REFERENCES Alba, Joseph W., J. Wesley Hutchinson, an d John G. Lynch (1991), "Memory and Decision Making," in Handbook of Consumer Theory and Research , ed. Harold H. Kassarjian and Thomas S. Robertson, Englewood Cliffs, NJ: Prentice Hall, 1 49. Bettman, James R. (1979), An Information Processing Theory of C onsumer Choice , Reading, MA: Addison Wesley. Desai, Kalpesh K. and Wayne D. Hoyer (2000), "Descriptive Characteristics of Memory Based Consideration Sets: Influence of Usage Occasion Frequency and Usage Location Familiarity," Journal of Consumer Research , 27 (December), 309 323. Hauser, John R. and Birger Wernerfelt (1989), â€œThe Competitive Implications of the Relevant Set/Response Analysis,â€ Journal of Marketing Research , 26 (March), 393 408. Howard, John A. and Jagdish N. Sheth (1969), The Theory of Bu yer Behavio r, New York: Wiley. Huber, Joel and Noreen M. Kline (1991), â€œAdapting Cutoffs to the Choice Environment: The Effects of Attribute Correlation and Reliability,â€ Journal of Consumer Research , 18 (December), 346 357. Hutchinson, Wesley J., Kalya n Raman, and Murali Mantrala (1994), "Finding Choice Alternatives in Memory: Probability Models of Brand Name Recall," Journal of Marketing Research , 31 (November), 441 461. Jarvis, Lance P. and James B. Wilcox (1973), "Evoked Set Size: Some Theoretical F oundations and Empirical Evidence," in Increasing Marketing Productivity and Conceptual and Methodological Foundations of Marketing, Combined Proceedings Spring and Fall Conference , 35, ed. Thomas V. Greer, Chicago, IL: AMA, 236 240. Johnson, Eric J. and John W. Payne (1985), "Effort and Accuracy in Choice," Management Science , 31 (April), 395 414.
48 Johnson, Michael D. (1984), â€œConsumer Choice Strategies for Comparing Non comparable Alternatives,â€ Journal of Consumer Research , 11 (December), 741 753. ____ ________ (1986), â€œModeling Choice Strategies for Non comparable Alternatives,â€ Marketing Science , 5 (Winter), 37 54. ____________ (1989), â€œThe Differential Processing of Product Category and Non comparable Choice Alternatives,â€ Journal of Consumer Researc h , 16 (December), 300 309. Johnson, Michael D. and Donald R. Lehmann (1997), â€œConsumer Experience and Consideration Sets for Brands and Product Categories,â€ Advances in Consumer Researc h, 24, ed. Merrie Brucks and Deborah MacInnis, Provo, Utah, Associatio n for Consumer Research, 295 300. Kardes, Frank R., Gurumurthy Kalyanaran, Murali Chandrashekaran and Ronald J. Dornoff (1993), â€œBrand Retrieval, Consideration Set Composition, Consumer Choice, and the Pioneering Advantage,â€ Journal of Consumer Researc h, 20 (June), 62 73. Lattin, James M., and John Roberts (1992), Testing for Probabilistic Independence in Consideration of Ready to Eat Cereals , Research Paper No. 1208, Graduate School of Business, Stanford, CA. Lehmann, Donald R. and Yigang Pan (1994), " Context Effects, New Brand Entry, and Consideration Sets," Journal of Marketing Research, 31 (August), 364 374. Maheswaren, Durairaj and Brian Sternthal (1990), "The Effects of Knowledge, Motivation and Type of Message on Ad Processing and Product Judgmen ts," Journal of Consumer Research , 17 (June), 66 73. Markman, Arthur B. and Douglas L. Medin (1995), â€œSimilarity and Alignment in Choice,â€ Organizational Behavior & Human Decision Processes , 63 (August), 117 130. Maule, A. John, Robert G. Hockey, and L. Bdzola (2000), "Effects of Time Pressure on Decision Making under Uncertainty: Changes in Affective States and Information Processing Strategy," Acta Psychologica , 104 (June), 283 301. Medin, Douglas L., Robert L. Goldstone, and Arthur B. Markman (1995), â€œComparison and Choice: Relations between Similarity Processes and Decision Processes,â€ Psychonomic Bulletin & Review , 2 (March), 1 19.
49 Mitra, Anusree (1995), â€œAdvertising and the Stability of Consideration Sets over Multiple Purchase Occasions,â€ Internat ional Journal of Research in Marketing , 12, 81 94. Mitra, Anusree and John G. Lynch (1995), â€œTowards a Reconciliation of Market Power and Information Theories of Advertising Effects on Price Elasticity,â€ Journal of Consumer Research , 21 (March), 644 659. Nedungadi, Prakash (1990), â€œRecall and Consumer Consideration Sets Influencing Choice without Altering Brand Evaluations,â€ Journal of Consumer Researc h, 17 (December), 263 276. Posovac, Steven S., David M. Sanbonmatsu and Russell H. Fazio (1997), "Consid ering the Best Choice: Effects of the Salience and Accessibility of Alternatives on Attitude decision Consistency," Journal of Personality & Social Psychology , 72 (February), 253 261. Ratneshwar, S., Cornelia Pechmann and Allan D. Shocker (1996), "Goal De rived Categories and the Antecedents of Across Category Consideration," Journal of Consumer Research , 23 (December), 240 250. Ratneshwar, S. and Allan D. Shocker (1991), â€œSubstitution in Use and the Role of Usage Context in Product Category Structures," J ournal of Marketing Research , 28 (August), 281 295. Roberts, John and James M. Lattin (1991), â€œDevelopment and Testing of a Model of Consideration Set Composition,â€ Journal of Marketing Research , 28 (November), 429 440. Schlosser, Ann and Alaina Kanfer ( 1999), â€œCurrent Advertising on the Internet: The Benefits and Usage of Mixed media Advertising Strategies,â€ in Advertising and The World Wide Web , eds. David W. Schumann and Esther Thorson, Mahwah, NJ: Lawrence Erlbaum Associates, 41 60. Simonson, Itamar (1990), â€œThe Effect of Purchase Quantity and Timing on Variety Seeking Behavior,â€ Journal of Marketing Research , 27 (May), 150 162. Simonson, Itamar and Russell S. Winer (1992), â€œThe Influence of Purchase Quantity and Display Format on Consumer Preference for Variety,â€ Journal of Consumer Research , 19 (June), 133 138. Troye, Sirgurd Villads (1984), â€œEvoked Set Formation as a Categorization Process,â€ Advances in Consumer Researc h, 11, ed. Thomas C. Kinnear, Ann Arbor, Michigan: Association for Consumer Res earch, 180 186.
50 Tversky, Amos (1980), "Features of Similarity," in Similarity and Choice , eds. Ernst D. Lantermann and Huber Feger, Bern: Hans Huber Publishers, 12 53. Tversky, Amos and Shmuel Sattath (1979), â€œPreference Trees,â€ Psychological Review , 86 (November), 542 573. Wright, Peter (1975), â€œConsumer Choice Strategies: Simplifying vs. Optimizing,â€ Journal of Marketing Research, 12 (February), 60 67.
51 APPENDIX A SAMPLE OF STIMULI USED IN STUDY 1 (Category) VIDEO GAMES (16) (Subcateg ory) SPORTS GAMES (8) COMBAT GAMES (8) (Type) BASEBALL (4) FOOTBALL (4) AERIAL COMBAT (4) INDIVIDUAL COMBAT (4) BASEBALL FOOTBALL AERIAL COMBAT INDIVIDUAL COMBAT 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 <= Batting Stances 2 <= Pitching Styles 3 <= Catches 4 <= Trading of Players 5 <= Stadium Graphics 6 <= Game Analysis/Commentary 7 <= Signature Play books 8 <= Formations 9 <= Player Specific Details 10 Number of Aircraft => 11 Number of Missions => 12 Cockpit Details => 13 Training => 14 Number of Levels => 15 Physical â€œImpactâ€ => 16 We apons => 17 Moves => A T T R I B U T E S 18 Game Guide => Brands: Ballpark Battles, Baseball Match Up, High Heat Baseball, Pro Baseball Brands: Super Bowl Sunday, Coaches Clicker, Monday Night Football, Kick Off Brands: World War II Ace, Bombs Away, Dogfigh t, Hellcat Fighter Brands: Street Fighter, Kung Fu Master, Ninja Cop, Fists & Fury
52 BALLPARK BATTLES BALLPARK BATTLES Baseball is Americaâ€™s past time and now you can bring the excitement of baseball home with you! Enjoy the versatility of over one hundred and thirty uni que batting stances. Mow down the batters with an arsenal of over fifty pitching deliveries. And when you need that extra effort, the future throw technology allows you to make incredibly super smooth catches and throws. Trade the batters, pitchers and fie lders to your liking but donâ€™t exceed the salary cap. Play in thirty MLB stadiums with active dugouts and bullpens. Enjoy a play by play analysis from the former Arizona Diamondbacksâ€™ manager Bob Zeilweger. Try to bring home the pennant for the home fans! SUPERBOWL SUNDAY SUPERBOWL SUNDAY Welcome to the future of football gaming â€“ more exciting than ever before! Control the action with one of twenty five predefined offensive plays or one of the fifteen defensive alignments. Use signature play books from real NFL coaches to fine tune your game strategy. Throw in player specific celebrations and emotions and you have the perfect ringside view. Prior to game day you look for the best â€“ drafting and swapping league talent. The stadium crowd will roar as you watch from the best seat in the house. Get a blow by blow account of the action from the top analysts in the NFL. If your efforts pay off then you reach the most important Sunday of your life! WORLD WAR II ACE You are the pilot in this flight simulator and battle game! WW II Ace has an all new combat simulator with over thirty authentic aircraft. Test your nerves with more than two hundred randomly generated missions. Enhanced cockpit details based on authentic WW II footage give you the real feel of being in the hot seat. You can go to flight school and hone your skills before joining the battle. Variable levels of flight and mission difficulty make our game ideal for all types of players. Realistic battle sounds and graphics bring the scenes alive. Just line up the enemy a nd fire away to reign supreme over the WW II skies! NINJA COP NINJA COP Guide a lone vigilante in his search for his own brand of justice on mean city streets! If you are not ready to face the mean streets yet, then gain confidence using one of the practice modes. C hoose from eight different levels of play to vary the difficulty and challenge you face. Hear the realistic sound of steel blades swishing through the air as the Ninja yields his weapons. Take your pick â€“ the blue steel swords, double or triple shurikens from the ninjaâ€™s lethal arsenal of thirty weapons. Master the Kuji in, a set of thirty different combinations of blows, punches and stealth moves that made the ninja famous. The game is accompanied by a guide that will tell you about weapons and enemy str ategies. Enjoy the unparalleled might of the Ninja!
53 APPENDIX B ATTRIBUTES AND BRANDS USED IN STUDY 1 CARS â€œSERIOUSâ€ CARS (FUNC TIONAL) â€œFUNâ€ CARS (LESS FUN CTIONAL) SEDANS HATCHBACKS CONVERTIBLES SUVS 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Turn Radii Braking Convertible Top Off road Capabilities Interior Design Legroom Top Speed All Wheel Drive Safety Noise Level Acceleration Room/Cargo Space Fuel Economy Fuel Economy Power/Engine Power/Engine Transmission Transmission Stability Control Stability Control F E A T U R E Depreciation Depreciation Ride Ride Brands: Toyota Corolla, Honda Civic, Ford Taurus, Volvo S40 Brands: Toyota Echo, Honda Insight, Ford Focus, Hyundai Accent Brands: Pontiac Firebird, Ford Mustang, Toyota Camry Solara, Volvo C70 Brands: Dodge Durango, Ford Explorer, Isuzu Rodeo, Mitsubishi Montero RESTAURANTS WESTERN CUISINE EASTERN CUISINE AMERICAN WEST EUROPEAN MIDDLE EAST FAR EAST 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Grills Seafood/Fowl Deep Fry Stir Fr y Cheese Oven Usage Bread Base Rice Base Fast/Efficient Beverage Meat/Lamb Fish Sauces Sauces Spicy Spicy Herb Based Herb Based Rich Rich F E A T U R E Simple/Mild Simple/Mild Distinctive Flavor Distinctive Flavor Brands: Hudson River Club, An American Plac e, Union Square Cafe, Gotham Bar & Grill Brands: Le Bernardin, Scalini Fedeli, Cafe de Brussels, The Swiss Chalet Brands: The Star of Persia, Istanbul Grill, Salam Cafe & Restaurant, The Nile Brands: Saigon Grill, Planet Thailand, Nippon Seafood & Sushi, C hina Palace
54 APPENDIX C DEPENDENT VARIABLE USED IN STUDY 1 Assume each video game is represented by one of the four letters, B, F, A or I. The table below comprises a hypothetical example of consideration sets constructed by three subjects. Type of Pairs (Number) Dependent Variable (Type of Pairs Total Available) Subject Consideration Set High Moderate Low High Moderate Low 1 B, B, B 3 0 0 3/24 0/32 0/64 2 B, F, F 1 2 0 1/24 2/32 0/64 3 B, F, I, A 0 2 4 0/24 2/32 4/64 TOTAL r high r moderate r low Dependent Variable = 3 Tests of Proportions ( r high vs. r moderate ; r high vs. r low ; r moderate vs. r low ). (Category) VIDEO GAMES (16) (Type) Sports Games (8) Combat Games (8) (Sub-type) Baseball (4) Football (4) Aerial Combat (4) Individual Combat (4) B, B, B, B F, F, F, F A, A, A, A I, I, I, I
55 APPENDIX D DETAILED RESULTS OF STUDY 1 1B Type of Alignable Pairs in Consideration Set Replicates High Moderate Low S tudy 1 Cars 11.70% a, b 9.05% b 4.99% Video Games 14.58% a, b 9.17% b 5.11% Restaurants 11.84% a, b 5.47% 6.02% Total 12.72% a, b 7.92% b 5.37% Study 1a Cars 9.50% a, b* 6.75% b 3.69% Video Games 13.50% a, b 5.63% b 1.88% R estaurants 11.70% a, b 4.25% 5.13% Total 11.56% a, b 5.54% b 3.56% Study 1b Cars 10.23% a, b 6.91% b 4.49% Video Games 8.87% a, b 5.14% b 3.02% Restaurants 10.03% a, b 5.56% 5.51% Total 9.35% a, b 5.90% b 4.18% a implies significa ntly different from % of moderately alignable pairs, at 95% level of confidence b implies significantly different from % of low alignable pairs, at 95% level of confidence b* implies significantly different from % of low alignable pairs, at 90% level of co nfidence
56 APPENDIX E SAMPLE OF STIMULI USED IN STUDY 1C VIDEO GAMES: HIGH COMPARABILITY CONDITION BASEBALL GAMES FOOTBALL GAMES 1 2 3 4 5 6 7 8 9 10 1 Batting Stances 2 Pitching Deliveries 3 Catches 4 Trading/Draft 5 Stadium Graphics 6 Game Analysis 7 Physical "Impact" 8 Play Books 9 Offensive Formations A T T R I B U T E S 10 Defensive Formations VIDEO GAMES: MODERATE COMPARABILITY CONDIT ION BASEBALL GAMES FOOTBALL GAMES 1 2 3 4 5 6 7 8 9 10 1 Batting Stances 2 Pitching Deliveries 3 Catches 4 Trading/Draft 5 Stadium Graphics 6 Game Analysis 7 Physical "Impact" 8 Play Books 9 Offensive Formations A T T R I B U T E S 10 Defensive Formations VIDEO GAMES: NON COMPARABLE CONDITION BASEBALL GAMES FOOTBALL GAMES 1 2 3 4 5 6 7 8 9 10 1 Batting Stances 2 Pitching Deliveries 3 Catches 4 Trading/Draft 5 Stadium Graphics 6 Game Analysis 7 Physical "Impact" 8 Play Books 9 Offensive Formations A T T R I B U T E 10 Defensive Formations
57 APPENDIX F STIMULI USED IN STUDY 2 REPLICATE(S) Attribute1 Attribute2 Attribute3 Attribute4 Attribute5 Attribute6 Radar Detectors 360 Detection City/Highway Mode Digital Voice Alerts Signal Strength Safety Alert Electronic Compass Vacuum Cleaners Special Attachments Detachable Power Nozzle Full Dirt Bag Indicator Dirt Sensor Micro filter Capability Revolving Brush Air Conditioners Energy Saver Mode Built in Dehumidifier Anti bacterial Filters Timer Remote Control â€œQuietâ€ Mode Microwave Ovens Turntable Infrared Sensor Stainless Steel Interior Numeric Keypads Short cut Keys Popcorn Key Alternative A Yes No No Yes Yes No Alternative B No Yes No Yes Yes No Alternative C Yes Yes No No No Yes Alternative D No Yes Yes No No Yes Al ternative E No No Yes No Yes Yes Alternative F No Yes Yes No Yes No Alternative G Yes No Yes No Yes No Alternative H Yes No Yes No No Yes Alternative I No No Yes Yes Yes No Alternative J No No No Yes Yes Yes Alternative K Yes Yes Yes No No No Altern ative L No Yes No No Yes Yes Alternative M Yes Yes No Yes No No Alternative N No Yes Yes Yes No No Alternative O No No Yes Yes No Yes Alternative P Yes No Yes Yes No No Alternative Q Yes No No Yes No Yes Alternative R No Yes No Yes No Yes Alternativ e S Yes No No No Yes Yes Alternative T Yes Yes No No Yes No Note. â€” Jaccard ratio for alternatives A and B = 0.5 (high), A and C = 0.2 (moderate), and A and D = 0.0 (low).
58 APPENDIX G FORMULA FOR JACCARD RATIO OF SIMILARITY The Jaccard ratio of simila rity between any two objects A and B is given by: R AB = C AB /(C AB + U A + U B ) Where, R AB = The Jaccard ratio of similarity between A and B, C AB = the number of features common to both A and B, U A = the number of features contained in A, but not in B (i. e., unique features of A), and U B = the number of features contained in B, but not in A (i.e., unique features of B).
59 APPENDIX H DETAILED RESULTS FROM STUDY 2 Types of Similarity Pairs Included in the Consideration Set Replicates High Medium Low A ir Conditioners 6.34% a, b 1.56% b 0.67% Apartments for Rent 6.48% a, b 1.49% b 0.13% Microwave Ovens 5.95% a, b 1.26% b 0.13% Radar Detectors 5.40% a, b 1.20% b 0.13% Vacuum Cleaners 5.44% a, b 1.03% b 0.13% Total 5.92% a, b 1.30% b 0.24% a implies significantly different from % of moderately similar pairs, at 95% level of confidence b implies significantly different from % of low similar pairs, at 95% level of confidence
60 APPENDIX I STIMULI USED IN STUDY 2A High Similarity Condition Moderate Si milarity Condition Alternative F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 1 0 1 1 1 1 0 0 1 1 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 2 1 1 0 0 1 1 1 1 0 0 0 0 1 1 0 0 1 1 1 0 0 0 0 0 3 0 1 1 0 1 1 0 1 1 0 0 0 0 1 1 0 1 1 0 0 0 0 0 0 4 1 0 0 1 1 1 1 0 0 0 0 0 1 0 0 1 1 1 1 0 0 0 0 0 5 1 1 1 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 1 0 0 0 0 0 6 0 0 1 1 1 1 0 0 1 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 7 1 1 1 1 0 0 1 1 1 0 0 0 1 1 1 1 0 0 1 0 0 0 0 0 8 0 1 0 1 1 1 0 1 0 0 0 0 0 1 0 1 1 1 0 0 0 0 0 0 9 1 1 0 1 0 1 1 1 0 0 0 0 1 1 0 1 0 1 1 0 0 0 0 0 10 0 1 1 1 0 1 0 1 1 0 0 0 0 1 1 1 0 1 0 0 0 0 0 0 11 1 0 1 1 0 1 1 0 1 0 0 0 1 0 1 1 0 1 1 0 0 0 0 0 12 1 0 1 1 1 0 1 0 1 0 0 0 1 0 1 1 1 0 1 0 0 0 0 0 13 1 1 1 0 1 0 1 1 1 0 0 0 1 1 1 0 1 0 1 0 0 0 0 0 14 1 1 0 1 1 0 1 1 0 0 0 0 1 1 0 1 1 0 1 0 0 0 0 0 15 1 0 1 0 1 1 1 0 1 0 0 0 1 0 1 0 1 1 1 0 0 0 0 0 16 0 0 0 1 1 0 0 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 0 17 0 0 0 0 1 1 1 1 0 0 1 1 0 0 0 0 0 1 1 1 0 0 1 1 18 0 0 0 0 1 1 0 1 1 0 1 1 0 0 0 0 0 1 0 1 1 0 1 1 19 0 0 0 1 1 1 1 0 0 1 1 1 0 0 0 0 0 1 1 0 0 1 1 1 20 0 0 0 0 0 1 1 1 1 0 0 1 0 0 0 0 0 1 1 1 1 0 0 1 21 0 0 0 1 1 1 0 0 1 1 1 1 0 0 0 0 0 1 0 0 1 1 1 1 22 0 0 0 1 0 0 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 0 0 23 0 0 0 1 1 1 0 1 0 1 1 1 0 0 0 0 0 1 0 1 0 1 1 1 24 0 0 0 1 0 1 1 1 0 1 0 1 0 0 0 0 0 1 1 1 0 1 0 1 25 0 0 0 1 0 1 0 1 1 1 0 1 0 0 0 0 0 1 0 1 1 1 0 1 26 0 0 0 1 0 1 1 0 1 1 0 1 0 0 0 0 0 1 1 0 1 1 0 1 27 0 0 0 1 1 0 1 0 1 1 1 0 0 0 0 0 0 0 1 0 1 1 1 0 28 0 0 0 0 1 0 1 1 1 0 1 0 0 0 0 0 0 0 1 1 1 0 1 0 29 0 0 0 1 1 0 1 1 0 1 1 0 0 0 0 0 0 0 1 1 0 1 1 0 30 0 0 0 0 1 1 1 0 1 0 1 1 0 0 0 0 0 1 1 0 1 0 1 1 0 => â€œNoâ€, 1 => â€œYesâ€ Imp ortance Rank(s) 1 3 5 7 9 11 12 10 8 2 4 6 1 3 5 7 9 11 12 2 4 6 8 10
61 Low Similarity Condition Alternative F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 1 0 1 1 1 1 0 0 0 0 0 0 0 2 1 1 0 0 1 1 0 0 0 0 0 0 3 0 1 1 0 1 1 0 0 0 0 0 0 4 1 0 0 1 1 1 0 0 0 0 0 0 5 1 1 1 0 0 1 0 0 0 0 0 0 6 0 0 1 1 1 1 0 0 0 0 0 0 7 1 1 1 1 0 0 0 0 0 0 0 0 8 0 1 0 1 1 1 0 0 0 0 0 0 9 1 1 0 1 0 1 0 0 0 0 0 0 10 0 1 1 1 0 1 0 0 0 0 0 0 11 1 0 1 1 0 1 0 0 0 0 0 0 12 1 0 1 1 1 0 0 0 0 0 0 0 13 1 1 1 0 1 0 0 0 0 0 0 0 14 1 1 0 1 1 0 0 0 0 0 0 0 15 1 0 1 0 1 1 0 0 0 0 0 0 16 0 0 0 0 0 0 0 1 1 1 1 0 17 0 0 0 0 0 0 1 1 0 0 1 1 18 0 0 0 0 0 0 0 1 1 0 1 1 19 0 0 0 0 0 0 1 0 0 1 1 1 20 0 0 0 0 0 0 1 1 1 0 0 1 21 0 0 0 0 0 0 0 0 1 1 1 1 22 0 0 0 0 0 0 1 1 1 1 0 0 23 0 0 0 0 0 0 0 1 0 1 1 1 24 0 0 0 0 0 0 1 1 0 1 0 1 25 0 0 0 0 0 0 0 1 1 1 0 1 26 0 0 0 0 0 0 1 0 1 1 0 1 27 0 0 0 0 0 0 1 0 1 1 1 0 28 0 0 0 0 0 0 1 1 1 0 1 0 29 0 0 0 0 0 0 1 1 0 1 1 0 30 0 0 0 0 0 0 1 0 1 0 1 1 Importance Rank(s) 1 3 5 7 9 11 2 4 6 8 10 12
62 APPENDIX J EXAMPLE OF UNDERLYING BENEFITS IN STUDY 4 Sports Video Games Combat Video Games Baseball Games Football Games Aerial Combat Individual Combat Challenge # of pitching styles # of batting stances # of defensive alignments # of offensive formations # of aircraft to choose from # of missions available # of lethal weapons # of fighting moves Realism Stadium details and graphics Game analysis and commentary Actual Monday Night venues Draft/trading of players Details of cockpit â€˜Physicalâ€™ impact of bullets Sound effects and visuals â€˜Impactâ€™ of blows
63 APPENDIX K STUDY 4 DESIGN VIDEO GAMES SPORTS GAMES COMBAT GAMES BASEBALL FOOTBALL AERIAL COMBAT INDIVIDUAL COMBAT CONDITION 1: NO TRAD EOFF Challenge HIGH HIGH HIGH HIGH Realism HIGH HIGH HIGH HIGH CONDITION 2: TRADEOF F BETWEEN BENEFITS ( VERSION 1) Challenge HIGH HIGH LOW LOW Realism LOW LOW HIGH HIGH CON DITION 3: TRADEOFF B ETWEEN BENEFITS (VER SION 2) Challenge LOW LOW HIGH HIGH Realism HIGH HIGH LOW LOW
64 APPENDIX L SAMPLE OF STIMULI USED IN STUDY 4A The Base Stimuli Matrix: Attribute 1 Attribute 2 Attribute 3 Attribute 4 Attribute 5 Attribute 6 Alternative 1 Yes No No Yes Yes No Alternative 2 No Yes No Yes Yes No Alternative 3 No No Yes Yes Yes No Alternative 4 No No No Yes Yes Yes Alternative 5 Yes Yes No No No Yes Alternative 6 No Yes Yes No No Yes Alternative 7 Yes No Yes No No Yes Alternative 8 Yes Yes Yes No No No Condition 1: No Tradeoff E1 C1 E2 C2 E3 C3 Alternative 1 Yes No No Yes Yes No (2E, 1C) Alternative 2 No Yes No Yes Yes No (1E, 2C) Alternative 3 No No Yes Yes Yes No (2E, 1C) Alternative 4 No No No Yes Yes Yes (1E, 2C) Alternative 5 Yes Yes No No No Yes (1E, 2C) Alternative 6 No Yes Yes No No Yes (1E, 2C) Alternative 7 Yes No Yes No No Yes (2E, 1C) Alternative 8 Yes Yes Yes No No No (2E, 1C) Condition 2: Tradeoff E1 E2 E3 C1 C2 C3 Alternative 1 Yes No No Yes Yes No (1E, 2C) Alternative 2 No Yes No Yes Yes No (1E, 2C) Alternative 3 No No Yes Yes Yes No (1E, 2C) Alternative 4 No No No Yes Yes Yes (0E, 3C) Alternative 5 Yes Yes No No No Yes (2E, 1C) Alt ernative 6 No Yes Yes No No Yes (2E, 1C) Alternative 7 Yes No Yes No No Yes (2E, 1C) Alternative 8 Yes Yes Yes No No No (3E, 0C) Note. â€” E = economy and C = comfort. High Simi larity High Similarity Low Sim. 2 economy, 2 comfort cars 2 economy, 2 comfort cars Cars high on comfort Cars high on economy
65 APPENDIX M FEATURES AND BENEFITS USED IN STUDY 4A AND 5 Replicates Benefit Dimension 1 Benefit Dimension 2 Sedans Economy/Reliability Feature 1: Fuel efficiency Feature 2: Warranty Feature 3: Reliability Comfort Feature 4: Good ride Feature 5: Ample legroom Feature 6: Rapid acceleration MBA Schools Placements Feature 1: Top 25 ranking Feature 2: Proximity to big city Feature 3: Good placement record Ease of Entry Feature 4: Low acceptance rates Feature 5: High GMAT requirement Feature 6: High work experience requirement Apartments for Rent Location Feature 1: Close to school Feature 2: Quiet neighborhood Feature 3: Nice view Apartment Feature 4: Spacious Feature 5: Pool and gym Feature 6: Free cable Beach Vacations Quality of Beach Feature 1: Fine, white sand beaches Feature 2: Predictable weather Feature 3: Clear, b lue waters Activities Feature 4: Diving and snorkeling Feature 5: Water sports Feature 6: Bars and nightclubs Laptop Computers Computing Power Feature 1: Fast processor Feature 2: Large memory Feature 3: Large disk space Ergonomics Feature 4: Large sized screen Feature 5: Full size keyboard Feature 6: Lightweight
66 BIOGRAPHICAL SKETCH Amitav Chakravarti was born in Digboi, India, on May 14, 1973. After his initial schooling at the Carmel Convent School, Digboi, he went on to complete his high school edu cation at the Scindia School, Gwalior, India. He then attended the Elphinstone College (University of Bombay) in Bombay, India, where he earned a Bachelor of Arts (Honors) degree in Economics. Subsequently he earned a Master of Arts (Part I) degree in Econ omics, also at the University of Bombay. He then completed a Masters in Business Administration degree at the Indian Institute of Foreign Trade, New Delhi, India. His subsequent internship at Unilever (India) Ltd., and work experience in the area of market research and advertising research at J. Walter Thompson Associates (India), Ltd., were instrumental in shaping his interest in pursuing doctoral studies in the area of marketing. During his four years in the Ph.D. program at the Department of Marketing, h e has been collaborating with Christopher Janiszewski, Barton Weitz, Jinhong Xie, and Alan Cooke from the University of Florida, on several of research papers. In 2000, Amitav taught a masters course in International Marketing at the University of Florida. He has now accepted a position as an Assistant Professor of Marketing at the Leonard N. Stern School of Business, New York University.