Mall Entertainment and Shopping Behaviors: a Graphical Modeling Approach
ABSTRACT - The shopping landscape is filled with malls, each competing for their share of the consumers wallet. A significant method of trying to differentiate the mall product and increase market share has been an attempt by mall developers and management to increase the entertainment component of the mall. The current research was designed to examine the relationship between the multiple ways that malls may create entertainment value for the consumer and certain shopping behaviors.
Citation:
Iksuk Kim, Tim Christiansen, Richard Feinberg, and Hyunjip Choi (2005) ,"Mall Entertainment and Shopping Behaviors: a Graphical Modeling Approach", in NA - Advances in Consumer Research Volume 32, eds. Geeta Menon and Akshay R. Rao, Duluth, MN : Association for Consumer Research, Pages: 487-492.
The shopping landscape is filled with malls, each competing for their share of the consumers wallet. A significant method of trying to differentiate the mall product and increase market share has been an attempt by mall developers and management to increase the entertainment component of the mall. The current research was designed to examine the relationship between the multiple ways that malls may create entertainment value for the consumer and certain shopping behaviors. The findings of the research clearly identified a relationship between certain specific entertainment values in a shopping mall environment and shoppers mall visits but not all entertainment values are directly associated. INTRODUCTION Why do consumers go to shopping malls? There is no doubt that malls provide necessary and desired products for consumers in a modern economy. However, a mall may also provide a pleasurable diversion from everyday activities and chores (Bloch, Ridgway, and Dawson, 1994; Eastlick, Lotz, and Shim, 1998; Hirschman and Holbrook, 1982). It is this latter category of consumer benefits that has been the focus of attention in current mall management and development. Malls are now being built with large entertainment centers including rides, skating rinks, amusement parks, tens of movie theaters, museums, and virtual reality centers. As competition between malls increases, enhancing the entertainment value for the consumer appears to be becoming an important way of differentiating the mall "product." Marketers have frequently suggested that the entertainment value of a shopping mall is an important area for study (Christman, 1988; Eastlick, Lotz, and Shim, 1998; Hoban, 1997; Wakefield and Baker 1998). However, empirical research of the relationship between entertainment and consumer shopping behaviors and/or attitudes has been suggestive but not conclusive. Sherman and Smith (1987) found that positive moods were related to self-reported purchase and spending. Wakefield and Baker (1998) found that consumers who rated the mall environment to be exciting tended to stay longer, have higher repatronage intentions, and were less likely to go outshopping. In an extensive literature review Langrehr (1991) presented an impressive array of studies which showed the influence of environmental stimuli on perceptions and behaviors in non-shopping environments. Langrehr speculated that these stimuli acting in a natural retail/mall arena would have the positive impact of facilitating shopping and spending behavior. It is possible that part of the reason for the weak relationship found between mall entertainment and shopping behaviors and/or attitudes is that mall entertainment has been loosely defined and is not a unidimensional construct. No one has tried to look at how individual entertainment aspects are related to each other and whether some mall entertainment aspects may be mediated by others. Thus not all entertainment aspects would make a mall attractive directly. Only entertainment aspects directly associated with shopping behaviors and/or attitudes would make malls attractive. This study begins by developing an instrument to measure a wide range of mall entertainment attributes. These attributes will then be analyzed through the Coefficient of Belonging analysis procedure, which will then be used to graphically model the relationship between the mall entertainment factors and certain shopping behaviors. DEVELOPMENT OF SURVEY INSTRUMENT The extant literature was examined for previous scales used to measure the entertainment value of a mall. No scale or set of scales was found, however a study by Wakefield and Baker (1998) did use a number of scales to assess the level of excitement at the mall. These items measured three aspects of the physical environment of the mall; the variety of retail offerings at the mall; and how involved the individual was with the activity of shopping. It was believed that the items measuring the physical environment of a mall were an excellent starting point, but did not capture the full range of what makes a mall entertaining. It was also believed that there is a difference between the concept of exciting and entertaining. Excitement is commonly associated with a stimulation of the senses associated with high levels of pleasure and arousal (cf. Wakefield and Baker 1998). Entertainment is "something that amuses, pleases or diverts, especially a performance or a show" (Random House Dictionary 1993). There is an undeniable overlap between the two concepts, however, they do not appear to be interchangeable. An entertaining mall may make shopping more pleasurable or a welcome diversion from daily life without reaching what may be termed a high level. It is possible that exciting malls may be malls with intensive levels of entertainment. For example, your local regional mall may provide a pleasant shopping experience that provides some level of entertainment, while the Mall of America with its indoor amusement park and walk-through aquarium may be an exciting place to shop. The next step was to conduct two focus groups to elicit what makes a mall entertaining to consumers. The focus groups consisted largely of individuals known to the researchers, but were acknowledged shoppers of many different malls across the country. A number of ideas arose from these discussions including the physical environment of the mall, activities which took place at the mall (e.g., special exhibits and events), and the people who worked and shopped at the mall. A set of fifty-one items were generated, based upon the focus groups and the literature search. These items were then examined for clarity, conciseness, and relevance (DeVellis, 1991) by a set of three academic researchers not involved in the research and six graduate students. The reviewers were all fully briefed on the purpose of the scale, previous research in this area, and the ideas that had come from the focus groups. These individuals suggested a number of items for elimination from the scale. The primary researchers then used these suggestions to reduce the scale to thirty-eight items. A survey instrument was developed that would be administered through the mail. This survey included the thirty-eight entertainment items measured by seven point (Strongly Agree to Strongly Disagree) Likert scales, the number of visits to the stimulus mall within the last thirty days, the number of items purchased at the mall during the last trip, and a number of demographic questions,. The survey was then given to a convenience sample of consumers to see if they understood the directions and if it was easy to read and respond to and they reported no problems with it. DATA COLLECTION The next step was to develop a sample for the survey. The researchers had previously purchased a commercially available computer program which listed malls and their mall management across the country. Thirty malls which were in excess of 500,000 square feet of gross leaseable area were randomly selected to be the stimulus mall for the consumers answering the surveys. Malls of this size are typically classified as regional or super-regional malls and are the ones most likely to have the ability and resources to make themselves more entertaining (Berman and Evans 2001). Once the malls were selected, a sample of one hundred households within a fifteen-mile radius of the location of each mall was purchased from a commercial research service. This provided a total of 3,000 households for the sample. Each household was mailed a survey, a cover letter, and a postcard. The letter asked for the respondent to complete the survey and provided information about an incentive that was being given for their cooperation. The respondent was asked to fill out the survey and the postcard and mail both of them back to the researcher. The postcard entered the respondent in a drawing for two prizes of $50 each and because the survey came back under a separate cover it assured anonymity of the respondent. Of the 3,000 surveys mailed, only twenty-two came back for bad addresses. A total of 485 completed surveys were returned for a response rate of 16.3%. The individual response rate by mall varied from 4% to 25%. DATA ANALYSIS The sample was overwhelming female (314 females, 140 males, 31 no response), which is not surprising given that the survey was sent to a household and females still tend to do the majority of shopping of all types for the family. The age of the respondents was split about evenly between under and over 45. About half of the respondents reported a family income of under $50,000 and half were over this amount. The number of visit to the stimulus mall within the last thirty days ranged from 0 to 25, with a mean of 3.6 visits. The number of items purchased at the stimulus mall during the last trip ranged from 1 to 6, with a mean of 4.6 items. The first step was to look at the distribution of scores on all items. The mean value for each scale item ranged from 3.2 to 5 on a seven-point scale. The standard deviations ranged from 1.3 to 2.1. The total score across the thirty-eight items created an entertainment index of a mallBthe higher the score the greater the entertainment value in that mall. Before the main analysis, the thirty-eight mall entertainment items were classified into non-overlapping groups by using the Coefficient of Belonging (B-coefficient) (Holzinger and Harmon, 1941). When factor analysis is employed simply as a statistical tool to group of variables, a procedure is readily available under the assumption that the variables of a group identifying a factor have higher inter-correlations than with the other variables of the total set. Such an index is designated as the B-coefficient, measuring the degree of cohesiveness of variables in a group. Thus, the B-coefficient focuses on the positive relations of variables in a group rather than maximizing total explained variance in the principal component method. The B-coefficient is defined as 100 times the ratio of the average of the intercorrelations among the variables of a group to their average correlation with all remaining variables (Harmon, 1976). Thus, a value of B=100 means that the average correlations of the variables in a group is the same as the average correlation of the remaining variables. The detailed grouping procedures based on B-coefficient are as follow: 1. Select two variables which have the highest value of B-coefficient as an initial group and denote the B-coefficient at this step as B(Gi)j, where Gi is a set of these two variables for group i and j denotes the grouping step. 2. Add another variable to the group Gi which maximizes the B-coefficient of the existing variables in the group Gi. Denote the B-coefficient at this step as B(Gi)j+1. 3. Evaluate the difference between the two values of the previous steps A B(Gi)j+1=B(Gi)j-B(Gi)j+1. 4. Let j=j+1 and do step 2 and step 3 as long as A B(Gi)j+1 is decreased. Or stop the grouping process for group i and designate the variables in Gi as a group of variables that belong together. Let i=i+1 and go to step 1 for another group. 5. Do 1-5 until all variables are grouped. For example, among the thirty-three questions, item 22 and 25 were initially selected because of their high B-coefficient value (B(G1)1=208.90, G1={22, 25}). Then item 23 was added as the next variable which maximizes the sum of B-coefficient with item 22 and 25 (B(G1)2=197.64, -B(G1)2=11.26, G1={22, 25, 23}). Item 24 was the next variable. However, item 24 (B(G1)3=183.64, -B(G1)3=13.99) was withdrawn because -B(G1)3 was bigger than that of item 23, -B(G1)2. Item 24 was put back into the set of remaining variables to be considered for the next grouping procedure. Each of the remaining 34 items was considered for grouping with items 22, 23, and 25 but none of them improved the B-coefficient for the group. Then the two items wih the next highest B-coefficient were selected for the nucleus of the next group and the procedure for grouping was repeated. Since the B-coefficient is the ratio of two averages, it is sensitive to total number of variables in a subset. Also, as the number of variables in B increased, the average of the intercorrelations of the variables in a subset (the numerator of B) tends to decrease and the average of the remaining variables (the denominator of B) tends to increase, since the variables are added on the basis of highest correlation with those already in the argument B. Thus, in general, the B-coefficient within a subset is decreased as the more variables are included. In step 3 and 4 of the above procedures, instead of using a single cutoff standard (e.g. 30 point drop in B) applying for all groups, we used a relative standard based on - B(Gi)j+1 in each group. In other words, a 30 point drop of B may be enough in one group to deleting the newest item, while a 30 point drop in another group may be not enough to justify dropping the item. As result of this process all 38 items were grouped into nine groups and each group has at least three, but at most five entertainment items (see TABLE 1). The title of each group was decided based on the contents of questions in a group. For example, Group 1 has been titled #Layout as the group meaning since item 22, item 23 and item 25 are questions regarding the layout of the mall. In order to estimate the internal consistency of each group, a Cronbachs alpha was computed for the items in each group. TABLE 1 shows that all coefficients except Group C and Group I were well above Nunally and Bernsteins (1994) criteria in an applied study, which is at least .80. Although, two groups were below the criteria, Group C was included for further analysis because of its high B-coefficient, while Group I was excluded. In the main analysis, the graphical-modeling (GM) approach was used to determine the relationships among eight individual entertainment aspects, mall visit frequency and number of purchased items. GM is a form of multivariate analysis that uses graphs to represent a model. A graph, g=(n,e), is defined as a structure of a finite set n of nodes and a finite set of e edges between these nodes (Edwards, 2000). Edges are a straight line drawn between the nodes ("Groups" in this research) to represent the relationship. In terms of using a graph to represent a clear cut of relationships between variables, GM is similar to Structural Equation Model (SEM). In general, GM provides more understanding of multiple equivalent models and following statistical properties than SEM. SEM is used to formulate a theory-driven model based on temporal ordering of variables, and then evaluate it. However, GM requires no variable order and searches through all possible classes of models that fit best. RESULTS OF B-COEFFICIENT ANALYSIS WITH INDIVIDUAL ITEMS RELATIONSHIP OF ENTERTAINMENT GROUP VARIABLES, MALL VISITS, AND PURCHASES The relationships among the constructs in graphical-modeling can be undirected (i.e., bidirectional) yet it has the ability to conduct a directed one like SEM. Also, SEM expresses the dependent variables as a function (usually linear) of those variables which determine it causally, together with an error term. However, GM represents conditional independence relations without an error term. Thus, recursive SEMs without correlated errors are special case of directed GM. In a large number variable set without the variable ordering knowledge, although it is theoretically possible to use SEM but practically impossible because of the complexity of the possible model (e.g. if this research uses SEM, 40,320 (8!) models should be evaluated). In this research, there was no previous research that provided guidance on which entertainment aspects may directly affect shopping behaviors or may mediate other entertainment aspects. It was thus determined to use an undirected graph in which all the edges are undirected and all nodes have equal probability to be considered as the first node. The model was tested using MIM, graphical-modeling software Version3.1 (Edwards, 2000). The total score of each group was used to represent each entertainment aspect. Thus, the value of each node was the sum of items in a group (i.e. Layout=I22+I25+I23). Edges were added based on one-step forward selection stepwise method. In other words, at each step, the edge with the smallest p-value, as long as the whole model was significant at p=.01, was added to the current model. Finally the partial correlations between the linked nodes in the obtained model were computed to find out the independent relationship from the given remaining nodes. RESULTS The obtained model provides a clear illustration of the ties between the groups (see GRAPH). The resulting model yielded the following fit statistics against the previous stepwise model: Deviance=41.33, df=25, p<0.01, Adjusted R2=0.95 (1B[G2(Mo)/dfo] / [G2(Mi)/dfi], where G2(Mo) is deviance of resulting model, G2(Mi) is deviance of complete independent model which has no edge between the nodes). In general, all included entertainment groups have at least three edges to other groups in the graph; however some groups are more densely tied to others. This result suggests that groups such as "Getting Out" "Music & Food" and "Layout" are perceived by the shoppers as common elements of entertainment which they can enjoy in the shopping mall environment. The graph also provides a visual linkage between the types of entertainment available through mall shopping and the two shopping behaviors of visit frequency and purchasing. The graph suggests that visit frequency is related, either directly or indirectly, to a number of the mall entertainment values while the number of items purchased is not. The two entertainment groups that have a direct effect on visits are the "Layout" of the mall and simply "Getting Out" for the consumer. In other words, only "Layout" and "Getting Out" groups are conditionally dependent of mall visit frequency. One implication of the model is that to predict the mall visit, "Layout" and "Getting Out" groups are sufficient. These two groups seem to be related to both utilitarian use of the mall and hedonic (pleasurable) use of the mall (Babin, Darden, and Griffin 1994). A mall with an easy to shop layout would allow the shopper to complete their shopping tasks efficiently and effectively. The "Getting Out" group certainly seems to suggest that the mall is viewed as a place to go to socialize, to be with friends and family, to simply get away from it all. The concept that shoppers may experience both utilitarian and hedonic value from a shopping trip has frequently been suggested in the literature (e.g., Babin, Darden, and Griffin 1994; Bellenger and Korgaonkar 1980). This research provides additional support for this notion. Although "Layout" and "Getting Out" are the only groups that directly influence mall visiting in this research, the other entertainment groups may have an indirect influence on mall visits. However, since the edges between the variable groups are undirected (or bi-directional), it is difficult to say whether these groups had an indirect influence on mall visits through the two variables with direct ties to the number of visits. It is noticeable that there is no significant dependency between any of entertainment aspects and purchasing behavior. This result indicates that the shoppers number of mall trips and possibly the amount of time spent at the mall are influenced by the entertainment aspects of a shopping mall while the actual purchasing or number of purchase item is influenced by other factors if there is any. A fitted partial correlation analysis between the groups shows that all correlations are positive (see TABLE 2). The fitted partial correlation was calculated after fitting the resulting undirected model into the data. Thus, only linked edges between the nodes have the value of partial correlation. Since the partial correlation is generated based on the variance overlaps of the corresponding variables, it has the same meaning as the standardized parameter estimates in a linear model. The great value of the partial correlation is that it expresses the effect of one variable on another without regard to how differently the variables are scaled. In general, if shoppers have more values in A and F entertainment groups, they tend to visit the mall more often. Thus, we can conclude that the higher entertainment values related to "Layout" and "Getting Out" directly results in more mall visits. FITTED PARTIAL CORRELATION MATRIX FOR LINKED EDGES DISCUSSION AND CONCLUSION The myth of the female shopper who wanted to nothing more than to go shopping for a day as a means of entertaining herself has been replaced by the myth of the female shopper who finds shopping a chore and wants to complete the task as quickly as possible. Neither of these views is entirely correct or incorrect. Todays shopper is time pressured, but can still find enjoyment through the act of shopping. Mall developers and managers have been working hard to try to make the shopping trip as enjoyable as possible and provide our shopper with a reason to stay in the mall longer and to come more often. Malls are largely trying to accomplish this goal through making the shopping trip more entertaining, but they have been lacking in basic research as to what creates an entertaining shopping trip. The current research developed an instrument to tap into multiple dimensions of what makes for an entertaining mall visit. Possibly the most important finding of the research was that only two dimensions of mall entertainment appeared to be driving the number of visits to the mall. One of these dimensions could be easily categorized as having a hedonic experience while shopping. However, the other appears to largely be a utilitarian dimension of the shopping trip. Since both of these dimensions had a direct and significant relationship with the number of shopping visits, it is important for mall developers and managers to remember that shopping is an economic activity which provides the shopper with desired items. Making sure that the mall has the right mix of stores to address the needs of the target market should remain a prime consideration. The current research also demonstrated that while entertaining the shopper is important, it does not necessarily have to include large amusement areas, theme parks, or strolling troubadours. Shoppers just wanted a place to get away from their daily grind, a place to visit with friends and family. There are a number of ways that malls help provide the means for consumers to engage in these activities, such as staying open at night, having food courts where multiple types of food are available, having special exhibits keyed to consumer interests and activities (e.g., boat shows, camping shows, bridal shows), and having seating areas where people can rest and visit. These features are not necessarily expensive to produce, but they provide great value and incentive for the consumer to visit the mall. While this research provides a clear cut of relationship between the entertainment aspects in the shopping mall and shopping behavior, it suggests several additional avenues for future research. For example, at present, it is unknown what portion of influence to the shopping behaviors comes from the children in a family. Obviously, a significant portion of household shopping behavior may be influenced by children. The age and gender composition of the familys children may also play a role. For example, a family with two teen-aged girls may be encouraged to go shopping more frequently than a family with a single male toddler at home. Further, the gender difference in shopping behavior is not discussed in the study. It is well known that males are not gnerally shoppers, but identifying key entertainment aspects for males may help mall managers develop programs encouraging greater participation in shopping by males. The mall has become the downtown for many suburban communities. As such, it is expected to provide more than just necessary products for the consumers in the community. It has become the preferred gathering place for teenagers, a place to meet friends after work, a place for early morning physical conditioning, as well as a convenient location for buying merchandise. The mall is a hub of both economic and social activity (Feinberg, Meoli, and Sheffler 1989) and what encourages people to engage in consumer activity in these locations is an important area for further research. REFERENCES Babin, Barry J., William R. Darden, and Mitch Griffin (1994), "Work and/or Fun: Measuring Hedonic and Utilitarian Shopping Value," Journal of Consumer Research 20 (4), 644-656. Bellenger, Danny N., and Pradeep K. Korgaonkar (1980), "Profiling the Recreational Shopper," Journal of Retailing, 56 (Fall), 77-91. Berman, Barry and Joel R. Evans (2001), Retail Management: A Strategic Approach, 8th ed. Upper Saddle River, NJ: Prentice Hall. Bloch, Peter, Nancy Ridgway and Scott Dawson (1994), "The Shopping Mall as Consumer Habitat," Journal of Retailing, 70(1), 23-42. Christman, Edward (1988), "Mixing Entertainment," Retail Shopping Centers Today. 1, 4-5. DeVellis, Robert F. (1991), Scale Development: Theory and Applications. Newbury Park, CA: Sage Publishing. Eastlick, Mary A., Sherry Lotz and Soyeon Shim (1998), "Retail-Tainment: Factors Impacting Cross Shopping In Regional Malls," Journal of Shopping Center Research, 5(1), 7-33. Edwards, David (2000). Introduction to Graphical Modeling, New York, NY: Springer-Verlag. Feinberg, Richard A., Jennifer Meoli and Brent Sheffler (1989), "Theres Something Social Happening at the Mall," Journal of Business and Psychology, 4(Fall), 44-63. Harmon, Harry H. (1976), Modern Factor Analysis, Chicago, IL: The University of Chicago Press, Chicago. Hirschman, Elizabeth C. and Morris Holbrook (1982), "Hedonic Consumption: Emerging Concepts, Methods and Propositions," Journal of Marketing, 46(Summer), 92-101. Hoban, Susan (1997), "Retail Entertainment: New Developments Are Making Shopping Fun Again," Commercial Investment Real Estate Journal, March/April, 24-29. Holzinger, Karl J. and Harry H. Harmon (1941), Factor Analysis, Chicago, IL: The University of Chicago Press, Chicago. Langrehr, Frederick W. (1991), "Retail Shopping Mall Semiotics and Hedonic Consumption," Advances in Consumer Research, 18, 428-433. Nunally, Jum C. and Ira H. Bernstein (1994), Psychometric Theory, New York, NY: McGraw-Hill. Random House Dictionary (1993), Avenel, NJ: Random House. Sherman, Elaine and Ruth B. Smith (1987), "Mood States of Shoppers and Store Image: Promising Interactions and Possible behavioral Effects," Advances in Consumer Research, 14, 251-254. Wakefield, Kirk L. and Julie Baker (1998), "Excitement at the Mall: Determinants and Effects on Shopping Responses," Journal of Retailing, 74(Winter), 515-540. ----------------------------------------
Authors
Iksuk Kim, California State University B Los Angeles
Tim Christiansen, University of Arizona
Richard Feinberg, Purdue University
Hyunjip Choi, KyongGi University
Volume
NA - Advances in Consumer Research Volume 32 | 2005
Share Proceeding
Featured papers
See MoreFeatured
M5. The More Expensive a Gift Is, the More It Is Appreciated? The Effect of Gift Price on Recipients’ Appreciation
Jooyoung Park, Peking University
MENGSHU CHEN, Tencent Holdings Limited
Featured
I1. Blaming Him or Them? A Study on Attribution Behavior
Chun Zhang, University of Dayton
Michel Laroche, Concordia University, Canada
Yaoqi Li, Sun Yat-Sen University, China
Featured
How Categories Transform Markets through Non-Collective, Non-Strategic Collaboration
Pierre-Yann Dolbec, Concordia University, Canada
Shanze Khan, Concordia University, Canada