Investigating Situational Effects in Wine Consumption: a Means-End Approach

ABSTRACT - This study groups consumer’s means-end chains according to the consumption situation, rather than by consumer characteristics. It relies on a predictive clustering technique, learning vector quantization (LVQ), to form well differentiated clusters which could be used by marketers to position their product for different usage situations. 648 different means-end chains, corresponding to 356 different occasions, were collected from 223 respondents. Using LVQ, an initial 8-cluster solution was found which fit the data well. However, a better predictivity was obtained by increasing the number of clusters to 14. The implications of these results are discussed in the conclusion of this paper along with directions for future research.



Citation:

Jean Marie Aurifeille, P.G. Quester, John Hall, and Larry Lockshin (1999) ,"Investigating Situational Effects in Wine Consumption: a Means-End Approach", in E - European Advances in Consumer Research Volume 4, eds. Bernard Dubois, Tina M. Lowrey, and L. J. Shrum, Marc Vanhuele, Provo, UT : Association for Consumer Research, Pages: 104-111.

European Advances in Consumer Research Volume 4, 1999      Pages 104-111

INVESTIGATING SITUATIONAL EFFECTS IN WINE CONSUMPTION: A MEANS-END APPROACH

Jean Marie Aurifeille, Universite de la Reunion, Reunion

P.G. Quester, The University of Adelaide, Australia

John Hall, Victoria University, Australia

Larry Lockshin, University of South Australia, Australia

ABSTRACT -

This study groups consumer’s means-end chains according to the consumption situation, rather than by consumer characteristics. It relies on a predictive clustering technique, learning vector quantization (LVQ), to form well differentiated clusters which could be used by marketers to position their product for different usage situations. 648 different means-end chains, corresponding to 356 different occasions, were collected from 223 respondents. Using LVQ, an initial 8-cluster solution was found which fit the data well. However, a better predictivity was obtained by increasing the number of clusters to 14. The implications of these results are discussed in the conclusion of this paper along with directions for future research.

INTRODUCTION

Recent use of the means-end chain methodology for predicting brand choice has shown some promise in grouping consumers based on linking their personal values to the desired roduct attributes (Aurifeille and Valette-Florence 1995; Reynolds and Gutman 1988). An alternative method for clustering markets, situation, was proposed earlier by Dickson (1982), but no empirical evidence was cited. A better understanding and more accurate prediction of behaviour in the marketplace may benefit from both a consumer means-end and situational perspective. Until now, the concept of situation was seldom operationalised, mostly because of its descriptive nature and its potentially unlimited range.

The goals of this paper are two-fold: first to investigate the suitability of means-end data for analysing situations, rather than products; second to assess the grouping of these chains using a new analytical technique, learning vector quantization. The paper first reviews the literature on situation or occasion-based behaviour and means-end research. A brief summary of the interview process for the data collected based on wine consumption occasions is provided before the analysis. A novel analytical approach is adopted in this paper, relying on the predictive clustering technique of learning vector quantization (LVQ). The results of this empirical study, along with their implications for both using means-end methods for situational analysis and for the LVQ approach, are presented in conclusion of this paper.

LITERATURE REVIEW

Situation

Situational influences have a theoretical foundation in Lewin’s field theory (1936) and the modern interactionism conception of human behaviour. These perspectives asserted that human motivations, intentions, and behaviour are a function of the interaction between consumers and situations. According to these theories, each individual views each physical and social setting somewhat differently.

A fairly limited number of researchers have investigated situational factors as a determinant of choice behaviour. Sandell (1968) presented subjects with an inventory of beverages and found that personal differences and differences in situations, considered separately, were poor predictors of product preference. Their interaction, however, provided a better predictor of beverage preference. This same type of interaction between product choice and usage situation was found by Green and Rao (1972), Belk (1974ab), and Srivastava, Shocker, and Day (1978). In a later study, Srivastava (1980) examined the appropriateness of financial services in a particular situation and found it to be relatively stable across situations, thus providing further support for using consumption situations as a basis for segmenting the market. Dubow (1992) compared occasion-based and user-based segmentation for the jug wine market in the US and concluded that the occasion-based segmentation was richer and more relevant for brand positioning and advertising strategy.

Clearly, there is merit in including product characteristics, consumer characteristics and specific situations in a combined analysis. Although little conclusive research has been reported in using consumption situation to group consumers, the above review indicates that adding situation to either product or consumer characteristics may improve the predictive nature of such market clustering techniques.

Values

Values are responsible for the selection and maintenance of the ends or goals toward which individuals strive (Vinson, Scott and Lamont 1977). A value is a centrally held, enduring belief which guides actions and judgements across specific situations and beyond immediate goals to more ultimate end states of existence (Kamakura and Mazzon 1991). Various combinations of values significantly differentiate individuals (Rokeach 1968). Personal values therefore have a major influence on a person’s lifestyle, interests, outlook and consumption priorities and thereforecan play an important role in the development of strategies to understand markets (Muller 1991).

Studies using value orientations to enrich the segmentation process have become increasingly popular, (Boote 1981, Holman 1984, Kahle 1986, Muller 1989, 1991, Kamakura and Novak 1992, Blamey and Braithwaite 1997, Thrane 1997, Jago 1998). The most frequently used instrument for measuring values is the Rokeach Value Survey which consists of 18 instrumental values and 18 terminal values (Kamakura and Mazzon 1991). The List of Values (LOV) developed by Kahle (1983) modifies Rokeach’s scale of terminal values into a smaller set of nine primarily person-oriented terminal values more directly related to a person’s daily life roles and situations (Beatty, Homer, and Shekhar 1985, Kamakura and Mazzon 1991) and as such, it has been utilised in a variety of segmentation studies (Kahle 1986, Muller 1989, 1991, Kamakura and Mazzon 1991, Kamakura and Novak 1992, Blamey and Braithwaite 1997, Jago 1998). In order to identify values and value chains, means-end analysis (Gutman 1982, Reynolds and Gutman 1988) provides a methodological approach used for identifying values as well as the attributes, benefits and consequences related to these values.

Means End Chains

The means-end chain is a conceptual model that relates salient values of the consumer with evaluative criteria (attributes) of the product (Howard, 1977; Vinson, Scott, & Lamont, 1977; Reynolds and Gutman, 1984). The model offers a procedural guide that establishes linkages connecting values important to the consumer to specific attributes of products, through psycho-sociological and functional benefits (called 'consequences’).

TABLE 1

OCCASION CLASSIFICATION

A sequence of in-depth probes traces the network of connections or associations in memory that eventually lead to values. This laddering process is accomplished by asking a "why is that most important to you" question at each level and uses the response as the basis for the next probe. The process continues until both a consequence and a personal value are elicited from the consumer, or the consumer has no further answers to the probes (Reynolds and Perkins 1987).

Gutman’s original model (1982) used situation in the theoretical description as one part of the matrix. Situation was deemed an input to the process of consumer decision-making. However, in various empirical examinations of the model, situation was not included (Reynolds and Gutman 1984; Reynolds and Perkins 1987; Reynolds and Gutman 1988; Gengler and Reynolds 1989). This research proposes to use situation instead of product or brand as the central focus of the means-end analysis.

THE STUDY

As previously noted, no prior reported study has, to the authors’ knowledge, attempted to examine the link between situations and the means-end chains associated with them. Our main objective, therefore, is to look for a relationship between consumption situations on the one hand, and the resulting means-end chain characterising a consumer’s value chain on the other hand. To do this, we propose the use of learning vector quantization (LVQ), a predictive clustering technique that can be applied to whole means-end chains, as opposed to other specific characteristics. Specifically, in order to identify whether particular patterns of means-end chains are associated with particular situations, we seek to test empirically whether the LVQ approach would provide a predictive clustering methodology such that any consumer’s means-end chain could be associated and therefore allocated to a specific consumption situation.

This approach, clustering means-end chains based on consumption situations rather then consumers, is therefore quite distinct from a more traditional segmentation approach that would aim at grouping consumers exhibiting similar means-endchains. We contend, consistent with Dickson (1982), that the same consumer may well exhibit a different means-end chain when facing a different consumption situation, making clustering the means-end chains more meaningful than clustering consumers.

Data collection method

The product selected for this study is wine. Previous research in occasion-based segmentation has shown that wine is chosen and consumed for different reasons in different situations (Dubow 1992; Lockshin, Macintosh and Spawton, 1997). Wine has a wide variety of attributes and as shown by Dubow (1992), a number of different consequences and values associated with its use. Therefore, the means-end approach adopted for this study used occasion as a factor for each ladder, rather than brand as in previous research. Indeed the seminal article in the means-end chain literature relies on data collected about wine coolers (Reynolds and Gutman, 1988).

A sample of 233 respondents was interviewed using a means-end analysis procedure. The sample was a convenience sample: interviewers used the phone book in a random manner in order to make contacts with respondents and also used personal contacts; some restrictions were placed on the sample. Respondents were required to be over 25 years of age and to have consumed wine which they had purchased in the last three months. Interviewers were asked to follow the means-end procedure for a specific purchase and consumption situation (some respondents discussed the last two occasions). The interviews thus produced 648 ladders for 356 occasions. The interviews were undertaken by trained researchers, the majority of interviews were undertaken by post graduate students in Marketing Research while approximately one third were undertaken by professional market research interviewers. All interviews were recorded and transcribed, all interviews were checked for authenticity and accuracy. As a result of reading the transcripts one of the authors (John Hall), identified the attributes, consequences and values associated with each wine consumption situation. A significant portion of these interviews were reviewed by colleagues to validate and support the process.

The values were coded to reflect the categories of the LOV scale. The 356 individual consumption occasions were also aggregated into 8 specific occasions that summarised and reflected the occasions presented by respondents (Table 1). A realistic representation of age and gender was obtained in the sample and respondents had purchased a reasonable amount of both red and white wines.

The List of Values (LOV) was used to measure values in this study. Based largely on the work of Maslow and Rokeach, it uses nine terminal values that are based on the Rokeach Value Survey, namely Fun and enjoyment in life, Being well respected, Warm relationships with others, Self-fulfilment, Security, Self-respect, Sense of belonging, Sense of accomplishment, and Excitement. The items can be interpreted in terms of interest served and motivational domains (Kamakura and Novak 1992). It is important to note that the attributes, consequences and values elicited in this study are reflective of those found in other studies. Each of these has been regarded as important by other authors in the field as summarised in Table 2.

TABLE 2

ATTRIBUTES AND CONSEQUENCES USED IN THIS STUDY

TABLE 3

ATTRIBUTES IDENTIFIED AND NUMBER OF LADDERS

TABLE 4

CONSEQUENCES IDENTIFIED AND NUMBER OF LADDERS

RESULTS

The approach adopted in this study involved the use of a bio-mimetic method of predictive clustering coined 'learning vector quantization’ (LVQ). According to this method, each object is assigned to the cluster, the center of which resembles most the object’s profile. The center of the cluster is then moved (i.e. its coordinates are changed) to allow it to come closer to the object if they are both connected to the same dependant variable (in this case the occasion). If this is not the case, the discrepancy of the cluster center with the profile of the object is increased. In ther words, the clusters are built up incrementally with the addition of each object, around centers that reflect the similarities and differences in each additional profile. This results in differentiated clusters, while at the same time allowing some predictivity: as a new object is presented, its profile (i.e. coordinates) should enable an accurate allocation to a given segment. This approach is described in more detail elsewhere (Aurifeille and Deissenberg, 1998).

Most frequently cited attributes, consequences and values

Respondents were asked what had influenced their selection of a particular wine for a particular occasion. Following the interviews, the attributes were categorised. Table 3 shows these categories and the number of chains on which each attribute occurred. Taste (285), price (221), type (215), and brand (111) were the attributes of wine most frequently listed.

Table 4 lists the consequences identified and the corresponding number of ladders. A number of consequences were frequently suggested as a result of attributes associated with selected wines. Selected attributes were indicators of quality -a consequence appearing in most means-end chains (212). Other frequently cited important consequences included: Socialise (168), Complement Food (135), Impress Others (131), Value for Money (123), and Mood Enhancement (118).

Table 5 lists the values (from the LOV Scale) and corresponding number of ladders. All values except Excitement (6) and Sense of Accomplishment (35) were well represented. Fun and enjoyment in Life (226) was the most represented value on the means-end chains. Other popular values were: Being well respected (148), Warm relationship with others (121), Self fulfilment (119) and Security (109).

LVQ Clustering

A minimum requirement of the LVQ analysis is to define a number of clusters at least equal to the number of categories in the dependant variable, in this case, occasion. Therefore, in order to segment consumers’ means-end chains according to the eight types of occasions identified, the initial LVQ analysis sought an 8-cluster solution (ie. one for each consumption situation). Results of this initial LVQ are shown in Tables 6 and 7.

While the 8-cluster solution provides well differentiated segments, its ability to predict accurately the membership of any object remains modest. Indeed, there are only 199 ladders accurately assigned to the right occasion cluster. Considering that the worst possible score would involve 139 ladders accurately assigned (to a single occasion cluster), it appears that 7 additional clusters were required to assign a small proportion (less than 12 %) of the objects.

We therefore undertook to improve the predictivity of the solution by increasing the number of clusters. This led to a 14-cluster solution (larger solutions appeared less predictive) for which 385 ladders remained ill-assigned. In other words, this solution assigned 24% of ladders accurately, a marked improvement over the 8-cluster solution. As a result, however, our 8 original occasions were split in 14 clusters as shown in Table 8.

The corresponding cross-tabulations are shown in Table 9 (eg. 59 ladders corresponding to occasion 2 are assigned to it from three different clusters, namely clusters 2, 3 and 4). It is clear that the 14-cluster LVQ solution, while more predictive, reveals less homogeneity within each occasion than was achieved with the 8-cluster solution.

Clearly, this 14-cluster solution suggests that 14 occasions would provide a better fit, as well as more interpretable results.

TABLE 5

VALUES IDENTIFIED AND NUMBER OF LADDERS

TABLE 6

CLUSTER DIFFERENTIATION TESTS

TABLE 7

FREQUENCIES OF OCCASION PER CLUSTERS

TABLE 8

CLUSTER SPLITTING BASED ON ORIGINAL OCCASIONS

TABLE 9

FREQUENCIES CROSS TABULATIONS (CLUSTERS BY OCCASIONS)

CONCLUSIONS AND DIRECTIONS FOR FUTURE RESEARCH

This paper presented a novel approach that connects stuation with consumers’ means-end processes. The method of analysis, learning vector quantization, was selected for its ability to optimise the resulting segments in terms of their differentiation and predictivity. It was found that predictivity with the proposed 8-occasion framework was poor and that 14 clusters were necessary to optimise the predictivity of the segmentation. However, interpretability of the 14 cluster solution is less clear.

These results suggest that the original categorisation of occasions into the 8 proposed consumption situations may need further examination. The 14-cluster solution demonstrated that only a few of these eight occasions were well defined, as evidenced by a match with a single cluster. The other occasions are represented by the sum of several clusters, which indicates, at least mathematically, a mixture of occasions in a single cluster. Future research should therefore seek to validate whether the 8 occasions used in this study are in fact multi-dimensional. For example, Eating with Friends (3 clusters) or Eating with Family (2 clusters) may in fact involve other latent but important factors that would influence the means-end chain associated with them.

Wine managers seeking to differentiate their markets based on the eight consumption situations would need to examine the content of these situations in order to uncover other aspects which may determine why certain choices are made, following distinct means-end chains. The strong fit of the eight cluster solution may provide a new means of segmenting wine markets, but the less definable 14 clusters may indicate some inherent problems with predictivity. However for practical purposes, the 8 cluster outcome provides some valuable insights and potential uses for marketing managers.

On a more general level, the paper shows the recurring compromise which must be made between predictivity and differentiation, two desirable outcomes sought when undertaking any clustering. Using an innovative data analysis technique enabled the researchers to question some of their assumptions regarding their categorisation of situations, while at the same time confirming the feasibility and the usefulness of clustering means-end chains for the purpose of understanding product choices. Further research, based on correspondence analysis of the occasions and the predictive means-end clusters, should assist in identifying the dimensions underlying the concept of situation. While further refinement is required, some optimism is warranted for a more detailed examination of the value oriented occasion-based clustering procedure.

REFERENCES

Aurifeille J-M. and Deissenberg C., (1998), Bio-mimetic Approaches in Management. Kluwer, Dordrecht.

Aurifeille J-M and Valette-Florence P., (1995), 'Determination of the Dominant Chains’, International Journal of Research in Marketing, 12, 267-278.

Beatty S. E., Kahle L., Homer P.and Shekhar M. (1985), 'Alternative Measurement Approaches to Consumer Values: The List of Values and the Rokeach Value System’, Psychology and Marketing, 2(Fall), 181-200.

Belk R.W., (1974a), 'An Exploratory Assessment of Situational Effects in Buyer Behavior’, Journal of Marketing Research, 11, 156-163.

Belk R.W (1974b), 'Application and Analysis of the Behavioural Differential Inventory for Assessing Situational Effects in Consumer Behaviour’, Advances in Consumer Research, 1, S. Ward and P. Wright, eds., Urbana Association for Consumer Research, 37-388.

Blamey R.K. and Braithwaite V.A., (1997), 'A Social Values Segmentation of the Potential Ecotourism Market’, Joural of Sustainable Tourism, 5(1), 29-45.

Boote A.S., (1981), 'Market Segmentation by Personal Values and Salient Product Attributes’, Journal of Advertising Research, 21, 29-35.

Dickson P.R. (1982), 'Person-Situation: Segmentation’s Missing Link’, Journal of Marketing, (Fall), 56-54.

Dubow J., (1982), 'Occasion Based vs User Based Segmentation’, Journal of Advertising Research, 2, 11-18.

Gengler C., and Reynolds T., (1989), 'Means End Structural Analysis: Computer Generated Hierarchical Value Maps’, Workshop on Consumer Behavior, Extending the Cognitive Structure Perspective, European Institute for Advanced Studies in Management, Brussels.

Gluckman R.L.,(1990), 'A Consumer Approach to Branded Wines’, European Journal of Marketing, 24(4), .27-46.

Green,P.E. and Rao V., (1972), 'Configuration Synthesis in Multidimensional Scaling’, Journal of Marketing Research, 9, 65-68.

Gutman, J., (1982), 'A Means-end Chain Model Based on Consumer Categorisation Process’, Journal of Marketing, 1, 23-24.

Hall J, Doole I., Shaw M., (1997), 'Cross Cultural Segmentation of the Wine Market’, International Journal of Wine Marketing, 9 (2/3), 83-90

Holman R.H. (1984), 'A Values and Lifestyles Perspective on Human Behavior’, in Pitts R.E. and Woodside. A.G., Personal Values and Consumer Psychology, Lexington Books, Lexington, MA, 35-54.

Howard, J. A., (1977), Consumer Behaviour: Application of Theory. New York: McGraw-Hill.

Jago L. (1988), The Analysis of Values in Major Events, Unpublished Doctoral Thesis , Victoria University of Technology, Melbourne Australia.

Judica F.; Perkins W. S.,(1992), 'A Means-End Approach to the Market for Sparkling Wines’, International Journal of Wine Marketing, 4(1), 10-18

Kahle L.R., (1986), 'The Nine Nations of North America and the Value Basis of Geographic Segmentation’, Journal of Marketing, 50, 37-47

Kahle L.R. (1983), Social Values and Social Change: Adoption to Life in America. New York, Warner.

Kamakura W. A., Mazzon J.A.(1991), 'Value Segmentation: A Model for the Measurement of Values and Value Systems’, Journal of Consumer Research, 18, 208-218.

Kamakura W. A, Novak T. P., (1992), 'Value-System Segmentation: Exploring the Meaning of LOV’, Journal of Consumer Research, 19, 119-132.

Koewn C., Casey M. (1995), 'Purchasing Behaviour in the Northern Ireland Wine Market’, British Food Journal, 97(11), 17-20.

Lewin K., (1936), Principles of Topological Psychology. New York, McGraw-Hill.

Lockshin L. Macintosh G. and Spawton A. L. (1997), "Using Product, Brand, and Purchasing Involvement for Retail Segmentation," Journal of Retailing and Consumer Services, Vol. 4 (3), 171-183.

Mitchell V. W., Greatorex M., (1988), 'Consumer Risk Perception in the UK Wine Market’, European Journal of Marketing, 22(9), 5-15 .

Mitchell V.W., Greatorex M., (1989), 'Risk Reducing Strategies Used in the Purchase of Wine in the UK’, European Journal of Marketing, 23(9), 31-46.

Muller T. E., (1991), 'Using Personal Values to Define Segments in an International Tourism Market’, International Marketing Review, 8(1), 57-70.

Muller T. E., (1989), 'The Two Nations of Canada versus The Nine Nations of North America: A Cross-cultural Analysis of Consumer’s Personal Values’, Journal of International Consumer Marketing, 1(4), 57-79.

Nerlove M., (1995), 'Hedonic Price Functions and the Measurement of Preferences: The case of Swedish Wine Consumers’, European Economic Review, 39(9), 1687-1716,

Reynolds T., Gutman, J., 1984 ’Advertising is Imae Management’, Journal of Advertising Research, 24, 27-36.

Reynolds T., Perkins W. S. (1987), 'Cognitive Differentiation Analysis: A New Methodology for Assessing the Validity of Means-End Hierarchies’, Advances in Consumer Research, 14, 109-113.

Reynolds T. J. , Gutman J., (1988), 'Laddering Theory, Method, Analysis and Interpretation’, Journal of Advertising Research, 28, 11-31.

Rokeach M., (1968), Beliefs Attitudes and Values. Jossey Bass, San Fransisco.

Rokeach M., (1973), The Nature of Human Values. Collier Macmillan, London.

Sandell R.G., (1968), 'The Effects of Attitudinal and Situational Factors on Reported Choice Behavior’, Journal of Marketing Research, 4, 405-408.

Sharp A., Smith J., (1991), 'Champagne’s Sparkling Success’, European Journal of Marketing, 25(3), 13-19.

Sharp B., (1991), 'Marketing Orientation: More than Just Customer Focus,’ European Journal of Marketing, 25(3), 20-25.

Spawton T.,(1991a), 'The Wine Marketing Environment’, European Journal of Marketing, 25(3), 12-18.

Spawton T., (1991b), 'Wine and the Marketing Mix’, European Journal of Marketing, 25(3), 19-31.

Srivastava R. K., (1980), 'Usage-Situational Influences on the Perceptions of Product Markets Response Homogeneity End Its Implications for Consumer Research’, Advances in Consumer Research, Jerry C. Olson, ed., Ann Arbor, MI, Association for Consumer Research, 644-649.

Srivastava R. K., Shocker A.D., Day G. S., (1978), 'An Exploratory Study of the Influences of Usage Situation on Perceptions of Product Markets’, Advances in Consumer Research, H. Keith Hunt, ed. Ann Arbor, MI, Association for Consumer Research, 32-38.

Thompson K., Vourvachis A., (1995), 'Social and Attitudinal Influences on the Intention to Drink Wine’, International Journal of Wine Marketing, 7(2), 35-45.

Thrane C., (1997), 'Values as Segmentation Criteria in Tourism Research: The Norwegian Monitor Approach’, Tourism Management, 18( 2), 111-113.

Vinson D. E. , Scott J. E., Lamont L. M., (1977), 'The Role of Personal Values in Marketing and Consumer Behavior’, Journal of Marketing, April, 44-50.

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

Authors

Jean Marie Aurifeille, Universite de la Reunion, Reunion
P.G. Quester, The University of Adelaide, Australia
John Hall, Victoria University, Australia
Larry Lockshin, University of South Australia, Australia,



Volume

E - European Advances in Consumer Research Volume 4 | 1999



Share Proceeding

Featured papers

See More

Featured

P6. Marginal Cost Consideration

Ethan Pew, Stony Brook University
Hyunhwan Lee, University of Miami, USA

Read More

Featured

Sizes are Gendered: Impact of Size Cues in Brand Names on Brand Stereotyping

Kuangjie Zhang, Nanyang Technological University, Singapore
Shaobo (Kevin) Li, Nanyang Technological University, Singapore
Sharon Ng, Nanyang Technological University, Singapore

Read More

Featured

Brands as Mediators: A Research Agenda

Philipp K. Wegerer, University of Innsbruck, Austria

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.