Aarhus School of Business Department of Marketing and Statistics June 2010
Exploring consumer preferences in the United States wine market: Market segmentation applying best-worst scaling Generation Y emphasis
Master Thesis Master of Science in Marketing
Student: Rachel Leigh Lewis Advisor: Athanasios Krystallis
Abstract This paper provides market segmentation and targeting framework for companies in the wine market to follow. The main objective of this research is to segment the United States wine consumers based upon their preference levels of particular wine attributes. The wine attributes in an off premise purchase decision are rated using the Best-Worst scaling method. Segmentation on these scores was conducted through latent class cluster analysis resulting in four unique wine consumer segments. Based on analysis and comparisons, a detailed description of each segment is provided so company marketers know which attributes to focus on when choosing their targeting segment. A company can then implement marketing strategies that compliment their target segment and the corresponding preferences of certain wine attributes.
Keywords: Best worst scaling, Wine market, Attribute preferences, Generation Y, Consumer/Market segmentation, Latent class cluster analysis, Wine attributes
Acknowledgements I would like to start off by thanking my advisor Athanasios Krystallis for being cooperative and efficient while guiding me through this research process. I should also acknowledge Polymeros Chrysochou for being extremely helpful with the technical and analytical aspects of this report. Lastly, I am so grateful for my boyfriend, family, and friends for being so supportive during this master thesis procedure.
Rachel L. Lewis June 1st, 2010
Table of Contents Part I 1. Introduction……………………………………………………………………….……6 1.1 Problem statement……………………………………………………….……8 1.2 Aim and thesis structure………………………..……………………….……8 1.3 Delimitations……………………………………..……………………….…..10
Part II 2. Literature Review………………………………………………………………….......10 2.1 Wine market……………………………………………………………….…10 2.2 Market segmentation………………………………………………………...11 2.3 Generation Y……………………………………………………...………….14 2.4 Best-Worst scaling method…………………………………………………..16 2.5 Originality summary….……………………………………………………...19
Part III 3. Research Method………………………………………………………………………19 3.1 Research design………………………………………………………………20 3.2 Survey design and distribution……………………………………………...21 3.3 Data collection………………..………………………………………………22 4. Analysis and Results…………………………………………………………………..23 4.1 Sample description…………………………………………………..……….23 4.2 Wine purchasing behavior…..………………………………………………26 4.3 Best Worst analysis………..…………………………………………………29 4.3.1 Attribute comparisons: Demographic and geographic.……….……..33 4.3.2 Attribute comparisons: Behavioral………..…………..……………..35 4.5 Segmentation on Generation Y……………………………...………………38 4.6 Latent Class Cluster Analysis……………………………………………….41 4.6.1 Cluster 1.…………………………………………………………….44 4.6.2 Cluster 2.……….……………………………………………………44 4.6.3 Cluster 3………………………………………………………….….45
4.6.4 Cluster 4.……………………………………………………...……..46 4.6.5 Attribute Differences Among Clusters………………………………49
Part IV 5. Discussion………………………………………………………………………………52 6. Managerial Implications………………………………………………………………55
Part V 7. Limitations……………………………………………………………………………..57 8. Future Research……………………………………………………………………….58 9. Conclusion………………..……………………………………………………………60 10. Bibliography………………………………………………………………………….61
Appendix
List of figures Figure 1: Thesis structure…………………………………………...………………………9 Figure 2: Best Worst analysis on wine attributes………………….………………………31 Figure 3: Attribute importance and heterogeneity……………………………………..….32 Figure 4: Generation Y comparison on average BW scores………………………………38 Figure 5: Comparison of cluster mean BW scores……………………………………..….43
List of tables Table 1: Wine attribute list………………………………………………………..……….21 Table 2: Characteristics of the sample…………………………………………………….25 Table 3: Purchase characteristics of sample……..…………………………...……………27 Table 4: Purchase locations of sample…………………………………………………….27 Table 5: Involvement and loyalty of the sample………………………………….……….28 Table 6: BW average scores on wine attributes……………………………………..…….31 Table 7: Significant differences in Gender, Generation, Income BW scores………...…...34 Table 8: Significant differences in Region BW scores………………………………...….35 Table 9: Significant differences in Purchase variable BW scores……………………..….37 Table 10: Characteristics of Generation categories………………………………………..40 Table 11: Generation comparison on Involvement and Loyalty…………………………..41 Table 12: Latent class cluster models…………..…………………………………………42 Table 13: Latent class cluster analysis attribute mean scores……………………………..43 Table 14: Cluster demographic and geographic comparison…………….………………..47 Table 15: Cluster purchasing comparison……………..………………….……………….48 Table 16: Cluster purchase location comparison………………………………………….48 Table 17: Cluster Involvement and Loyalty comparison……………..…..……………….49
Part I
1. Introduction In such a competitive business-oriented world, companies are continuing to realize how crucial it can be to understand consumers and their corresponding behavior actions. Particular markets may be overwhelmed with excessive variations of products while other markets and products may not have been developed or even envisioned yet. Characterizing the product, the market field, and the consumer and then finally connecting consumers to the right product is a process that companies consistently have to deal with. Over the years many frameworks and advice steps have been developed to aid companies in defining these characteristics and making the connections. One particular way in which marketers define a specific market and its consumers is through market segmentation. Ideally the market segmentation should lead to “homogenous subgroups in that the members of one group should react in the same way to marketing stimuli and differ in their reactions to these stimuli from the members of other segments” (De Pelsmacker et al., 2007 p119). Segmentation continues to be one of the most powerful instruments in market strategy to understand particular markets and to supply the correct offers to selected customers at the right place of purchase (Cohen and Markowitz, 2002). The degree and levels of customer segmentation can differ greatly across a variety of markets. The greater complexity within a certain company product, the more complex the production and marketing processes can be which can lead to an even more complex market (Orth et al., 2007). The wine market is considered to be a very complex and unique market. It is extremely fragmented with numerous companies, brands, and varieties of products. Orth et al. (2007) was able to put the complexity of wine into perspective with this supermarket example. A large product category such as cereal may have up to 90 different product variants in a standard supermarket or grocery store. On the other hand wine often has well over 350 product variants. A specialty store may even carry as high as 3,500 wine product variants. Such immense variety and depth in wine products is due largely in part to the range of attributes characterizing wine in addition to the amount of
levels within each attribute (Orth et al., 2007). In addition to product complexity, wine can reach the consumer through multiple channel options. This adds to the already intricate market making things a bit more difficult and possibly confusing both for the companies and their consumers when it comes to deciding what to sell or purchase.
The wine market is filled with numerous competing companies and brands so it becomes even more important for organizations to tailor their marketing techniques towards a particular consumer segment. Today many researchers are looking into characterizing and classifying the Generation Y segment. This age group is very quickly becoming the most influential and powerful consuming generation so it is no wonder market researchers and companies are showing them interest. With the growing competitiveness in wine some wine marketers recommend finding new populations of wine consumers to target rather than focusing solely on their existing customers (Baenen, 2002; Gillespie, 2005). This Generation Y group could prove to be the key to gaining a competitive advantage in the wine market. In order to access this age group more knowledge on this segment must be obtained to understand their relationship with wine and their purchase decisions. With such a large variety in wine choices it can be confusing to understand what attributes influence consumers purchase decisions. It is particularly difficult for wine marketers to grasp and forecast consumer product preferences in the wine market (Goodman et al., 2005). One way to measure consumer preferences in wine is through the Best-Worst Scaling (BW) method otherwise known as Maximum Difference Scaling. This approach was first introduced into wine marketing by Goodman et al. (2005) and will be utilized in the U.S. wine market segmentation of this report.
By using the BW method wine
attributes will acquire a certain importance score that can allow for comparison between specific groups. It will enhance the segmentation process by illustrating variation in importance levels placed on a particular wine attributes and identifying demographic and behavioral segments based on attribute importance (Cohen and Markowitz, 2002). The details and advantages of this method will be explained further in the subsequent Literature Review section.
The intention of this report is to develop a stronger comprehension of the United States wine market consumers. Wine as the product focus for this report was decided on because of three main reasons. First off, the wine market provides a wide range and depth of attribute decisions consumers have to make when buying wine (Quester and Smart, 1998). Secondly, the wine market in the United States is growing and marketers have a huge opportunity in reaching new potential consumers. Lastly, focusing on the wine market followed along the lines of the researcher’s pure personal interest in the product and market. The report will focus on analyzing deeper into the U.S. wine market by segmenting based on the preference of wine attributes. The importance of understanding consumers and their needs is generally relevant for a wine company to succeed. The purpose of this thesis is to answer the following problem statement:
What attribute preferences and descriptive variables characterize specific consumer segments and their corresponding purchase decisions in the U.S. wine market?
Throughout the process of answering this problem statement, focus will be placed specifically on the Generation Y group and the differences in regards to the other generational groups. This is largely in place because of the compelling Generation Y attention lately and the realization of the strong strategic opportunity this generation holds in the U.S. wine market (Gorman et al., 2004).
1.2 Aim and thesis structure The main objective of this research thesis is to segment the U.S. wine market on wine attribute preferences. Making use of the BW scaling method as a segmentation platform will identify segments that put different emphasis on certain wine attributes. To further address the problem statement above, this research thesis desires to answer the following research questions through the analysis and results of this report:
1. In regards to the U.S. wine market consumers, what are the similarities and differences in BW method scores among demographic, geographic and behavioral variable categories? 2. What variable characteristics describe the Generation Y age group? What wine attributes are important (least important) to this age group and are there differences to the older generational age groups? 3. What characteristics define the wine consumer segments produced in the analysis and are there differences in attribute preferences when selecting a bottle of wine?
The research thesis structure can be seen in Figure 1. The figure displays the main chapters and focus ideas in this report. It can be seen that following this Introduction will be a basic Literature review on the prominent previous research in the following areas: the Wine market, Market segmentation, Generation Y, and the Best-Worst scaling method.
Figure 1. Thesis structure
Part I
Part II
Part III
Part IV
Part V
1.3 Delimitations The research was narrowed down to only the United States wine market. This scope was chosen based on the researcher’s basic knowledge of market and accessibility to possible survey respondents. It was convenient to utilize the network already developed by the researcher in that particular geographic and product area. Further limitations of this report will be discussed in the last section Part V of the report.
Part II
2. Literature Review This chapter provides a background on consumer segmentation in the wine market. Previous research methods and results related to the Wine market, Market segmentation, Generation Y and the Best-Worst scaling method are thoroughly discussed keeping in mind the focus of this report. The section will conclude with a short summary on the originality significance pertaining to this research report.
2.1 Wine market It is necessary to first examine the global and US wine market to get an idea of the market dynamics. Being the wine market economy is more than 4,000 years old (Spawton, 1991) it is no wonder why there have been vital changes throughout the market. The New World wine countries, such as Australia, U.S., Chile, Argentina, South Africa, and New Zealand now have a great influence on the wine market. These countries are challenging the traditional wine market and the way and reasons consumers are drinking wine (Orth et al., 2007).
In the last decade the global wine market has evolved into a unique and
complicated business market. Not only are more countries, companies, and brands playing a factor in this transformation but there have also been a shift in the overall wine business focus. Wine has become a business field focusing on management and marketing as opposed to solely viticulture (grape growing) and enology (wine making) tactics (Spawton,
1991). Spawton (1991) stresses the need for companies to have knowledge on both of these components in order to have success. A business aspect approach is a needed company element in order to compete and succeed in this highly competitive wine market.
The United States is the one of the focal points in all these wine market dynamics. Wine consumption in the United States has reached a record 745 million gallons a year. This is an increase of over 165 million gallons since 1985 (Zraly, 2009). However according to the Wine Market Council (2009) 70% of U.S adults are either non-drinkers or beer/spirits drinkers. That is a considerably large portion of the market that could potentially be future wine drinkers. Today there is so much more diversity in wine styles and wine prices, it is almost impossible to keep up with every new wine and new vintage that comes on the market (Zraly, 2009). Choosing a wine can be a difficult task for consumers so they make use of certain quality indicators (or in this case wine attributes) that are believed to be the same across countries (Aurifeille et al., 2002).
However, countries and any market
segmentation groups for that matter don’t necessarily prefer the same attributes to other attributes.
The importance level for each attribute may differ between segmentation
groups. Even though wine is a global product, variations in wine preference levels result in having a variation in the underlying marketing strategies and approaches (Orth et al., 2007). This topic is furthered discussed in the following Literature Review and throughout the remainder of this report.
2.2 Market segmentation Satisfying consumers and understanding their needs is the basis of marketing theory or marketing concept (Schiffman et al., 2008). There are differences in customer needs and it is rarely possible to satisfy all customers by treating them the same (Barber et al., 2008). Market segmentation allows for assurance that some of these different needs are being met. The process involves “dividing a market into distinct subsets of consumers with common needs or characteristics and selecting one or more segments to target with a distinct marketing mix” (Schiffman et al., 2008 p. 44). Before market segmentation, companies would take a “mass marketing approach by offering the same product to all consumers through the same marketing approaches” (Schiffman et al., 2008 p. 44).
Market segmentation is one of the three elements in a strategic marketing framework, the other two being targeting specific segments and positioning the product or service (Schiffman et al., 2008). This marketing tactic is known as the segmenting-targetingpositioning (STP) process (De Pelsmacker et al., 2007). Dodd and Bigotte (1997) point out through segmenting a market certain efficiencies can be achieved. Costs may decrease while advertising and promotions may become more effective (Dodd and Bigotte, 1997). “Segmentation can often be the key to developing and obtaining a sustainable competitive advantage” (Aaker and McLoughlin, 2007 p. 42). Barber et al. (2008) states there are two main reasons market segmentation is advantageous and helpful for any company. First, applying market segmentation allows for analyzing the needs of a particular customer segment. Secondly, any results from the analysis will provide obtainable knowledge on these identified needs in which promotional and advertising techniques can be focused on (Barber et al., 2008). Essentially the segmentation of a market will allow companies to efficiently and effectively spend their budgets while meeting the needs of the customer at the same time. According to Thomas and Pickering (2003) the interest in marketing to specific wine consumer segments is extremely new relative to the overall history of the wine market. McKinna (1987), cited in Spawton (1991), showed one of the first interests in applying segmentation to the wine market. Spawton (1991) then continued the studies and tested the four-segment solution consisting of connoisseurs, aspirational drinkers, beverage wine consumers, and new wine drinkers. Following researchers continued to test and develop consumer groups through similar lifestyle segmentations.
Constellation Wines, an
Australian wine making company, conducted one of the largest consumer research projects in the wine industry with their Project Genome where they revealed six premium wine consumer segments: Enthusiast, Image Seeker, Savvy Shopper, Traditionalist, Satisfied Sipper, and Overwhelmed (Constellation Wines, 2005). They implemented a consumer segmentation based on behavioral characteristics of the wine consumer. Many market segmentation studies have also been conducted based on consumer involvement levels in regards to a product or service.
Rothschild (1984) declared involvement to mean a
motivational and goal-directed emotional state that in turn determines the personal relevance of a purchase decision to a buyer. Larry Lockshin was part of multiple studies in which involvement was used to segment the consumer wine market both in a single country setting as well as in a global setting (Lockshin et al., 1997; Lockshin et al., 2001; etc).
Research has shown that not all wine drinkers are actually ‘involved’ in the wine product category (Lockshin et al., 1997; Quester and Smart, 1998). Quester and Smart (1998) studied consumer product involvement and certain consumption situations in relation to consumer behavior. So studies in regards to consumer involvement as a segmenting variable in the wine market have been done before (Lockshin et al., 1997; Lockshin et al., 2001; Quester and Smart, 1998; etc). For this reason, this particular study segmented the U.S. wine market on consumer preferences instead.
Involvement, however, will not
completely be disregarded because of its strong presence in the wine market. The levels of involvement will still be taken into consideration when describing the characteristics of the developed wine consumer groups. Through exploring wine market research it becomes apparent there are a wide variation of criteria to base market segmentation on (De Pelsmacker et al., 2007). As evident in the focus and aim of this research, this report will segment on consumer preferences when selecting a bottle of wine to purchase. A consumer’s purchase decision may be influenced by certain cues or wine attributes. A cue can be intrinsic meaning it is associated to the physical characteristics of the wine product itself (Schiffman et al., 2008). On the other hand, any extrinsic cues are external to the product and can be altered without actually changing the product (Schiffman et al., 2008; Lockshin and Rhodus, 1993). When a consumer has no actual experience with the product they may be more likely to rely on these extrinsic cues when deciding on a purchase or evaluating quality of the product (Schiffman et al., 2008). Perceptions of wine quality can be based on certain intrinsic cues, such as grape variety, alcohol content, and wine style (Lockshin and Rhodus, 1993). Extrinsic cues in wine may then be variables such as price, packaging, labeling and brand name (Lockshin and Rhodus, 1993). Because the wine market is filled with a plethora of
brands it possible consumers are uses different preferences to influence their purchases. This study will follow along the same lines and focus on some of these mentioned extrinsic/intrinsic cues otherwise known as wine attributes.
Quester and Smart (1998) found in their study that product involvement and consumption significantly influenced the relative importance of product attribute selection of red wine. Segmenting solely on involvement, however, could cause inaccurate results because involvement and consumption may interact during specific consumer choices (dependent on one another) (Quester and Smart, 1998). This is yet another reason why this particular research report is examining from the attribute preference perspective and using involvement and consumption later in the analysis to describe the segmented groups. It may turn out that for example the majority of highly involved consumers are in one group but this then shows a correlation of some sort to attribute preference and involvement. The base segmentation will be executed based on differences in attribute preferences.
2.3 Generation Y In the United States consumer industries may also be segmented or characterized based on generational categories. Lately a large interest has been paid to the Millennial Generation or Generation Y. There is a slight variation in the exact years of the people born into Generation Y. According to Key Findings (2004) and Harris Interactive (2001) those born in 1978 through 2000 are part of this precise generation. This age group is considered to be approximately 76 million people out of the United States population (Key Findings, 2004; Harris Interactive 2001). The total United States population is over 309,380,000 according to the U.S. Census Bureau’s population clock so Generation Y is a significant segment of the whole U.S. population (U.S. Census Bureau). For the purposes of this research report Generation Y or the Millennials were born in the 1977-1990 range making them the ages 18-32 in 2009 (Jones & Fox, 2009; Howe & Strauss, 2000; Strauss & Howe, 1992) and 26% of total adult population (Jones & Fox, 2009). One focus of this thesis is on segmenting characteristics of the Millennial Generation so it is not completely necessary to go into too much detail on the characteristics of all the remaining generations.
For more information pertaining to generational characteristics please refer to Morton (2003) and Howard and Stonier (2002).
The people of Generation Y have many characteristics specific to their age group. Charters and Fountain (2006) believe Millennials in general tend to be more confident, self-reliant and eager for a challenge. On the other hand their purchase confidence may not be as strong so they look for validation from other people (Charters and Fountain, 2006). In comparison people in Generation X tend to be known for independence and high selfassurance (Charters and Fountain, 2006).
Generation Y is considered by most main
consumer product companies as a generation with extremely high buying power (Key Findings, 2004; Harris Interactive 2001). They have grown up in a media saturated, brand conscious world (Fernandez-Cruz, 2003). Nearly 100% of the Generation Y people are connected to the Internet today (Harris Interactive 2001; Thach and Olsen, 2006). They are a unique target group because of this fact of being exposed to technology since their childhood (Djamasbi et al., 2010). The members of Generation Y believe that companies should understand them and their needs and modify appropriately when changes occur (Neuborne and Kerwin, 1999).
This generation is currently 30% of Internet using
population even though the other age groups are increasing in this area as well (Jones & Fox, 2009). This Internet access can have a major influence on consumers and greatly affect product companies because consumers are able to have access to a wide variety of information pertaining to products, competitors, reviews, etc.
Researchers have
concentrated on this relationship between technology and Generation Y members and have conducted studies for example in regards to Internet technology and web preferences (Djamasbi et al., 2010).
Until recently very little research has been done on the Generation Y connection to the wine market. The Millennial generation has shown the largest percentage increase in wine consumption over Generation X and the Baby Boomers (Zraly, 2009). In the past, the wine industry tended to focus their marketing strategies on the Baby Boomer generation because they were considered the main wine-consuming category (Lancaster and Stillman, 2002; Wine Market Council, 2003). Now the consumption of wine by generational groups
has shifted. Generation Y members are consuming larger quantities than the previous Generation X members (Jones, 2007; Wine Market Council, 2003, 2006). There are nearly 20 million members of the Millennial generation who have yet to reach the age of 21 (Wine Market Council, 2009).
This may give some assurance to wine marketers of
potential market growth in the near future. According to Barber et al. (2008) companies and marketers in the wine industry need to rethink the characteristics of the stereotypical wine drinker. Wine drinkers are becoming younger and bringing a unique set of tastes and lifestyle choices to the wine industry (Barber et al, 2006; Jones, 2006, 2007). Some researchers even believe the Millennial generation is part of the reason for the recent increased popularity of wine in the United States (Jones, 2006, 2007; Saad, 2005)
2.4 Best-Worst scaling method For many years now marketers and researchers have been exploring ways to measure consumer preferences in product markets, particularly through scaling methods (Cohen, 2009). Cohen (2009) states that Lockshin and Hall (2003) reviewed over 75 research articles concerning wine consumer behavior and preferences. Many of the studies used rating or ranking scales on survey questionnaires to measure consumer preferences for various wine attributes (Goodman 2009; Lockshin and Hall, 2003; Cohen, 2009). Measurement systems involving rankings or ratings of a product or service have supported powerful research and results in the past, however, current research is showing this form of measurement can lead to biases in the results. Respondents may not view and use the ranking or rating scale in the exact same way across all the respondents (Cohen 2003; Cohen and Neira 2003; Finn and Louviere 1992). The scaling method may have also been developed specifically for that research so the reliability and validity is lacking (Goodman et al., 2005). Using the standard scaling method also makes it hard to pinpoint the most important attribute or the most preferred product (Goodman et al., 2005). Ordinary rating or ranking scales make it difficult to measure attribute importance against the other competing attributes. Some people may be influence by all the attributes or none of them but this doesn’t provide adequate distinction to help marketers associate real influences in consumer choice (Finn and Louviere 1992). Ranking can also become exhausting for respondents as the number of attributes increases. Using solely consumer panel data only
helps to gain information on actual consumer purchases. This method is not particularly appropriate if a person is testing new concepts or a combination of preferences throughout a product field (Goodman et al., 2005). Actual preferences may be concealed because a product with a large market share may be available for more purchase and therefore purchased more frequently (Goodman et al., 2005). So consumers may not actually prefer a particular product or attribute just because it is selling more than the competitors. This establishes that consumer purchases may not necessarily reflect their true preferences. Other statistical methods such as discrete choice modeling have the capability to address consumer preferences, however, the interpretation of the data and adaptability to managerial application is remarkably more difficult.
Louviere and Woodworth (1990) formulated a scaling method in which to examine consumer preferences on products while eliminating the previously mentioned bias problems apparent in other scaling methods.
The Best-Worst (BW) Scaling method
otherwise known as Maximum Difference Scaling was developed by Louviere and Woodworth (1990) and then first presented by Finn and Louviere (1992). Since then the method has been used in multiple other studies in a variety of areas such as healthcare, social sciences, etc (Cohen, 2009). Researchers and in particular Cohen and Markowitz (2002), continued to explore it further and elaborate the true benefits from this approach. In the Best Worst scaling method respondents are provided with choice sets in which they have to compare and decide on attributes over the other options. Respondents have to choose the best/most important item and the worst/least important item from each given choice set. Through this process the bias in the rating scale is eliminated because a respondent only has one option to choose the attribute that is the `most´ and `least´ (Cohen and Markowitz 2002). Respondents are forced to make trade offs between the wine attributes (Cohen, 2009) as different attribute combinations are offered in choice sets. There is no built in assumption of the right way to read the interval scales and the differences between any of the scale points (Cohen and Neira, 2003). Not only does this method allow for the delimitation of scaling biases but it also creates a ratio-based scale with standardized scores that allow for comparisons and contrasts to be applied within the data set.
Since then multiple research papers have been published in regards to the Best-Worst Scaling method (Cohen 2003; Cohen and Markowitz 2002; Cohen and Neira 2003; Finn and Louviere 1992; Louviere and Islam 2004; etc). Goodman et al. (2005) was the first to connect the BW scaling method to the area of wine marketing in the Best-Worst Scaling: A Simple Method to Determine Drinks and Wine Style Preferences research piece. Goodman et al. (2005) demonstrated the Best Worst scaling method using various beverage types including styles and attributes of wine as focus. It was therefore a broader concentration and was principally conducted to illustrate the Best-Worst Scaling method ease and to signal areas for potential research. An opportunity for future work was stated to be in the sections of wine marketing, drivers of wine selection, and preferences for wine variety (Goodman et al., 2005). Further researchers began to take interest in this area but there are still seems to be only a limited amount of studies in marketing that use the BW method (Goodman et al., 2005, 2008).
Goodman (2009) used the BW method to his advantage in comparing wine attribute importance across cultures. As long as the choice sets were the same, the standardize scores can provide similarities and differences not only between consumer segments but between countries as well (Goodman 2009). Any cultural biases from scaling methods are eliminated. With the globalization and internationalization of many markets today this is useful to break through the barriers that are normally referred to as ‘cultural differences’ (Cohen 2009). This was extremely advantageous in the global wine market with the high variation in wine origin and the new world wine countries bringing more competition than ever to the global market. Goodman (2009) illustrated this advantage using United States BW scaling data as well as the data from 11 other countries in a comparative results paper. The comparative paper presents the analysis at a country level and proves the powerfulness of this specific method in conducting cross-cultural studies in consumer behavior. However, it was eluded that further analysis is needed, such as consumer segmentation, in order to get a better understanding of one particular countries attribute preferences.
2.5 Originality summary It is hoped that this research report provides a unique approach to base consumer segmentation on in the U.S. wine market with the use of this newer scaling method. This paper is original in that the purpose is to use the BW method to compare and contrast attribute preferences and importance levels throughout the U.S. wine market. This results in a more narrow focus by only choosing to dissect U.S. wine consumers. Through the BW score criteria it is hoped that customer segments and their particular characteristics within the U.S. consumer wine market will emerge.
Any generational differences will also be linked and examined within these segments. Studies have attempted to segment the Millennial generation (Thach and Olsen, 2006; Nowak et al., 2006; etc) but no studies were found in regards to this generation and segmenting purely on wine attribute preferences. Generation Y and other generation groups have not been clearly linked to the BW scaling method yet. The focus on the Millennial generation will provide an example of the steps a company can take to focus and characterize a particular segment of their choice. Essentially this thesis will present a combination of ideas and deeper analysis of previous studies by utilizing the BW method in the U.S. wine market to segment wine consumers. It will also provide demographic and behavioral characteristics in regards to differences in BW scores between consumer segments or clusters.
Part III
3. Research Method The following section will detail the research methods used in this wine market study of United States consumers. Consumer preference data will be collected using the BestWorst method mentioned previously. Respondents will consider their preferences of
purchasing wine in an off premise situation (not in a bar, café or restaurant) similar to the consumer choice criteria off premise study by Lockshin et al. (2006). Consumer characteristics will also be accumulated from consumer responses to questions on demographic, geographic, and behavioral features. The method was developed keeping in mind the purpose of this research being consumer segmentation based on attribute BW scores and then further describing these segments based on demographic and behavioral characteristics while paying particular attention to Generational Y.
3.1. Research Design Through the use of previous studies and experience, researchers have found consumer respondents discover the task of answering a Best Worst scaling survey as being easy and quick to complete (Auger et al., 2004; Cohen and Markowitz 2002; Finn and Louviere 1992). The Best Worst method design followed in this study is very similar to the design used by Finn and Louviere (1992) in their first presentation in the area of wine marketing and the design from the wine articles (Casini et al., 2009; Cohen, 2009; Goodman, 2009). The thirteen wine attributes selected were the same retail segment wine attributes used in Casini et al. (2009), Cohen (2009), and Goodman (2009). These particular attributes were chosen because experts have identified them as the most influential in wine selection at off premise locations (Goodman et al., 2005). The thirteen wine attributes can be found in Table 1. A balanced incomplete block design of type (13,13,4,1) was adopted to distribute the attributes into several groups of choices presented to the consumers. In this way the 13 wine attributes were used to construct 13 table choice sets (blocks), each comprising four criteria per case. Following this method ensures each attribute appears the same number of times with every other item (Orth et al., 2007). An example of the block design and one of the thirteen choice sets can be seen on in the Appendix A1 and A2 on page I. All thirteen choice sets in the survey questionnaire can be viewed in the survey questionnaire starting on page III of the Appendix A4.
Table 1: Wine attribute list Number
Wine Attribute
1 2 3 4 5 6 7 8 9 10 11 12 13
Promotional display in-store Grape variety Origin of the wine Information on the shelf Alcohol level below 13% Matching to my food Information on back label Medal/award An attractive front label Brand name Someone recommended it I read about it in a guide Tasted the wine previously
Wine attributes that influence consumers purchasing decisions and selection on wine.
After the BW choice sets, respondents were asked to rate themselves on behavioral statements. There were five involvement statements based on similar scales used in Beatty and Talpade (1994) and Mittal and Lee (1989). Respondents also rated three loyalty statements similar to scales used in Mittal and Lee (1989) and Slama and Tashchian (1985). In both cases the respondents were asked to rate the statements on a seven-point Likert scale with 1 being strongly disagree and 7 being strongly agree. With the length of the survey in mind the behavioral questions were limited and served just to get an idea of segment characteristics. It is always an area that research can concentrate on to define the segments even more in the future.
3.2 Survey design and distribution A web-based survey was used in this study because of the ease and convenience in quickly reaching respondents all over the United States. In order to avoid inaccurate or biased results, the survey demographic and behavioral questions were designed in such a way as not to influence respondents to a specific response (Sparrow, 2006). The remaining BestWorst scaling questions are already designed in a way to avoid these biases as previously
mentioned. Through this careful design based on previous studies and research it is hoped that any data errors and biases due to the design form of data collection are eliminated or at the least greatly decreased. Please view the actual distributed survey on page III in the Appendix for further details on the survey questions and layout.
The survey was designed and created through the web based survey tool SurveyMonkey (SurveyMonkey, 2010). After assembly and review of the developed survey, the program then provides a specific survey web link to distribute to potential interviewees after the survey’s completed assembly. Possible respondents were sent the survey link through email, online messaging, and other social media networks such as Facebook, Twitter, consumer blogs, etc. Due to the time length allowed for this study and the nature of this research, the data sample was collected based on convenience. The results are therefore not randomly generated and cannot necessarily be represented of the whole United States population. However it does give us an idea into the variation attribute importance in US wine consumers. Wine enthusiasts and consumers who buy wine were particularly focused. The survey was examining preferences of certain wine attributes, so the people who never buy wine were irrelevant pertaining to the focus of this study. Non-wine drinkers were avoided through the first survey question by weeding out those respondents who never bought wine. The survey link was posted on popular wine blogs, emailed to wineries, distributors, wine groups, wine tasting events, fan groups on Facebook, etc. All respondents for this wine survey completed it on voluntary free will. They were not paid nor did they receive any other reward incentive to complete the survey.
3.3 Data Collection The data was also collected through the SurveyMonkey program. A pilot study was launched on March 10th, 2010 to ensure the survey was comprehensive and working properly. 28 results were collected and apparent changes were made based on the feedback and obvious mistakes in the questions and/or data. The actual web based survey was launched on March 13th 2010 and closed to begin analysis on April 3rd, 2010. Within this three-week period potential respondents were sent the survey link or came in contact with the survey link through one of the online methods mentioned previously. With more
time a quota system could have been implemented to ensure equal and proportion respondent numbers to US demographics, per state, gender, age, etc.
4. Analysis and Results Due to the multiple components of the research analysis conducted in this report both the analysis and results will be presented together. This allows for a smoother account of steps in the analysis process and a clearer description of the corresponding results. The first section will contain a description of the sample and an explanation of any changes made to individual variables. Following will occur the actual Best-Worst method of analysis and detailing of the significant differences among certain demographic, geographic, and behavioral groups. A specific look at Generation Y in comparison to the Remaining Generations will be detailed to pinpoint any differences. The Analysis and Results section will end with the Latent Class Clustering of consumers and detailing of the cluster group characteristics, specifically keeping in mind those people in the Generation Y age group.
4.1 Sample Description The raw data set was first downloaded from SurveyMonkey and uploaded into SPSS version 17.0, a computer program used for statistical analysis (SPSS, Inc., Chicago IL). The survey resulted in 349 total respondents over the allotted three-week period. Only 260 of these responses were fully answered questionnaires and considered for analysis (N=260). Any cases with missing values were omitted in order to have consistency throughout the results and a full understanding of the details in each individual case. The first question of the survey, How often do you buy wine (not including restaurant purchases)?, served as a screening question to weed out the respondents who never buy wine. Since the focus was on the preferences of certain attributes when purchasing wine, it wasn’t applicable to include those who never purchase wine. The remaining complete 260 cases were then organized and checked over for errors to ensure the raw data was ready for descriptive and frequency analysis.
In order to develop a comparison analysis and possibly create consumer segments in the U.S. wine market, it is necessary to define and condense categories within certain variables. The Table 2 illustrates the number of responses and percentages per demographic and geographic category. The gender representation in this sample was 107 respondents (41.2%) being male and 153 respondents (58.8%) being female. Due to the focus of this study, it is obvious to create age categories based on generation titles. Making two categories of Generation Y and the Remaining Generations allows for determining differences and possible characteristics for specifically Generation Y in terms of the wine market. Consequently this results in 45% of the respondents being within the ages 18-32 otherwise known as Generation Y as stated previously in the report. Demographically respondents were from all over the United States. The U.S. Region locations and there relative states can be seen on page II in the Appendix A3. A strong representation was particularly from the Midwest region. This aspect will be detailed and explained further on in the report.
Table 2: Characteristics of the sample N
%
107 153
41.2 58.8
Generation Y: 18-32
117
45
Remaining Generations: 33-90
143
55
Education High school
41
15.8
University or higher
219
84.2
Income Low Income: $49,999 and below
114
High Income: $50,000 and above
146
43.8 56.2
Household Size One or less age 21 and over
39
Two or more age 21 and over
221
Relationship In a relationship
197
Not in a relationship
63
75.8 24.2
53 148
20.4 56.9
35
13.5
24
9.2
Gender Male Female Age Groups
US Region West Midwest South Northeast
15 85
Sample frequencies on U.S. wine consumer’s demographic and geographic characteristics.
The majority of the respondents were well educated with 84.2% being a university graduate or higher. According to the U.S. Census Bureau only 27.4% of people over 25 years of age have a Bachelor degree or higher in the United States (U.S. Census Bureau). A comparison of these two percentages just exemplifies the samples almost complete slant towards the direction of higher education respondents. With higher education tends to come a higher income. The respondent income levels ranged from less than $10,000 to more than $200,000, the majority were making over the U.S. per capita mean income of 27,466 U.S. dollars (U.S. Census Bureau). The wide ranges of income levels were condensed into two categories of $49,999 and below and $50,000 and above. For the simplicity purposes of this report the two categories will be called Lower Income and
Higher Income. The higher cutoff of $50,000 was chosen because of the high proportion of well-educated respondents. The majority of respondents were earning well over the US mean per capita income.
Over three-fourths of the sample respondents were in a relationship of some kind and living in a household containing two or more people over the age of 21. The legal drinking age in the United States is 21 so only people over 21 in the household were considered in this question (CDC). As you can see there were some variation in response rates per descriptive category but overall there was enough representation to continue on with the planned comparative analysis.
4.2 Wine purchasing behavior A respondents Purchase Frequency is characterized by how many times they buy wine during a four-week period (a month). Just a little over half (55%) of the respondents buy wine more than once per month. For comparison reasons in the analysis the respondents were separated into high and low buying frequency groups. Those buying at a higher frequency were purchasing wine once a week or more (28.5% of sample). The lower frequency group would then be purchasing a couple of times a month or less (less than once a week). Respondents of the survey were also asked to determine how many bottles of wine they had purchased within the last month. The mean number of bottles purchased over all respondents was 7,60. Being that the average number of people over 21 per household is 2,09 (68% of respondents had two or more in their household)7 bottles was chosen as the cut off between the two categories. This resulted in the Bottle Purchase categories of Low Bottle Purchase being less than seven bottles a month (65%) and High Bottle Purchase being seven or more bottles a month (35%). The Purchase Frequency and Bottle Purchase frequencies and percentages can be viewed in Table 3.
Wine purchasing in the U.S. may occur at several locations including supermarkets, restaurants, wineries, etc. Selling locations can differ by state depending on the state government laws pertaining to alcohol selling. The days and opening hours of purchase
may vary across the 50 states as well. Respondents were requested to denote in which locations they have bought their last five bottles of wine. Table 4 details each location option along with the corresponding sum and mean. Each location could have the range of 0-5 bottles for each respondent. The highest location sum could then be 1,300 bottles if every respondent bought five bottles at that particular location. The highest number of bottles was bought at the Liquor store with a sum=461 bottles and a mean=1,77 bottles. The least amount of bottles was bought at some Other location not listed in the options or on the Internet.
Table 3: Purchase characteristics of sample
Sum
%
Sample frequencies on U.S. consumer’s wine purchasing behaviors: Purchase Frequency and Bottle Purchase.
Table 4: Purchase locations of sample
Sample frequencies on U.S. wine consumer’s purchase locations for their last 5 bottle purchases.
For each individual an Involvement and Loyalty mean score was calculated based on the average of the 5 involvement statements and 3 loyalty statement responses. The results show a total Involvement mean of 5.17 meaning many of our respondents claimed to be at least slightly involved in purchasing wine as a product (Cronbach’s alpha= 0.873). This seems a bit high in regards to the respondents actual purchase behavior shown through Purchase Frequency and Bottle Purchase. This could be a sign of the respondents over estimation of their own involvement. The overall loyalty mean was 3.06 showing respondents were less loyal to specific brand names (Cronbach’s alpha= 0.032). The involvement and loyalty means were calculated to aide in describing group characteristics after the segmentation of consumers through latent class analysis. The specific statements as well as the involvement and loyalty means are displayed in Table 5.
Table 5: Involvement and loyalty of the sample
Total Involvement and Loyalty means and the individual statement means for this sample of U.S. wine consumers. The mean scores were calculated on Likert scales with 1 being strongly disagree and 7 being strongly agree. = Cronbach's alpha from the internal reliability test.
4.3 Best Worst analysis Before the actual analysis of results data can be conducted, the data has to be organized and transformed into original attributes like listed previously Table 1. A new variable on SPSS was created for each of the thirteen attributes by combining the responses from the four different times it appeared in the survey. For example in each case the Promotional Display received four different scores; 1 if respondent chose as best/most, -1 if chosen as worst/least, and zero if not chosen at all. Adding those four scores together for each case gave each respondent a total Best Worst score for Promotional Display. This resulted in each respondent having 13 new Best Worst scores, one for each attribute. Each attribute item appears four times in this design (Appendix A1 page I) so each attribute has the maximum of being chosen four times as the best and none as the worst. Or the opposite could happen with four times as the worst and none as the best. For this reason the BW scores for each attribute per individual can range from +4 to -4 (Cohen, 2009). Any individual with an attribute score beyond this range would have indicated an error somewhere within the data set. Now that the attribute variables are developed for each case, the number of times the attribute was most important (best) and the number of times the attribute was least important (worst) can be calculated overall for all respondents. These scores will from here on be known as Best and Worst Scores. The Best and Worst scores for this data set can be seen in Table 6. The total Best Worst (BW) score is calculated by subtracting the Worst score from the Best score for each particular attribute. When a BW score is positive it is because that attribute was chosen more often as ‘best’ rather than ‘worst’ (Cohen, 2009). The opposite is true for those attributes containing negative BW scores. The BW score is also an indication of the attributes level of importance in selecting a bottle of wine (Cohen, 2003; Cohen 2009; Goodman et al., 2005). The BW average is calculated by dividing the BW score by the number of respondents N=260. A fourth of this average may also be considered as the BW average because each attribute appears four times throughout the design. It will be noted in further situations when this occurs. For the most part this
study will consider those average BW scores ranging from -4 to +4. The 13 attributes along with their BW scores and averages can be furthered examined in the Table 6.
Figure 2 clearly displays the most important attribute in influencing wine choice was Tasted the wine previously with a BW average = 2.5000. The least important was Alcohol level below 13% with a BW average = -2.6808. The most important, middle ground, and least important attributes in selecting a wine can easily be in this Figure 2. These results are quite similar to those found in the Australian study (Cohen, 2009). Goodman (2009) noted this is partly why Australian wines tend to do so well in the U.S. The wine consumers in these two countries have similar attribute preferences when choosing a wine to purchase. These U.S. BW scores are pretty close to the ones collected by Goodman (2009) in his 12 country comparison study on wine attribute BW scores. The similarity in results in the Australian and USA scores helps to set up validity in this report’s results. These overall BW results wont be discussed too much further because that has already been done in previous studies (Cohen, 2009; Goodman, 2009; etc). The remaining of the BW analysis results will be discussed after the segmentation on consumers based on their attribute preferences (BW scores).
Table 6: Best Worst analysis on wine attributes No.
Attribute
Total Best
Total Worst
BW Score
Average BW Score
**Average BW Score
**Stdev of BW
13
Tasted the wine previously
679
29
650
2.5000
0.6250
0.4240
11
Someone recommended it
478
41
437
1.6808
0.4202
0.4100
2
Grape variety
330
207
123
.4731
0.1183
0.5823
12
I read about it
214
126
88
.3385
0.0846
0.4058
10
Brand name
222
142
80
.3077
0.0769
0.4270
3
Origin of the wine
235
182
53
.2038
0.0510
0.4873
6
Matching to my food
227
221
6
.0231
0.0058
0.5273
4
Information on the shelf
151
183
-32
-.1231
-0.0308
0.4145
9
182
215
-33
-.1269
-0.0317
0.4840
108
273
-165
-.6346
-0.1587
0.4283
1
An attractive front label Information on the back label Promotional display in-store
100
313
-213
-.8192
-0.2048
0.4440
8
Medal/Award
95
321
-226
-.8692
-0.2173
0.4575
7
5
Alcohol level below 13% 7 704 0.3625 -697 -2.6808 -0.6702 The ranking of BW scores indicating the importance of wine attributes by U.S. consumers. N=260 **Divided by four because each attribute was listed four times in the survey questionnaire.
Figure 2: BW average scores on wine attributes 0.8000 0.6000 0.4000 0.2000 0.0000 -0.2000
13
11
2
12
10
3
6
4
9
7
1
8
5
-0.4000 -0.6000 -0.8000 Attribute Number
Average BW scores of wine attributes that influence consumer purchase decisions in the U.S. wine market. Corresponding attribute name to the number can be found in the above Table 6.
The standard deviation for each attribute can also be seen on Table 6. A higher standard deviation indicates heterogeneity among variable groups. This variation signifies the need to segment U.S. wine consumers further. If the standard deviation was very low for each attribute, that would be a sign that all the groups feel the same way about the attribute importance levels. This is clearly not the case in this study. The Figure 3 displays the attribute importance in relation to heterogeneity for each of the 13 wine attributes. A company should always try to “optimize those attributes with high importance” levels (Mueller and Rungie, 2009 p. 29). As a person can see Tasted the wine previously had the highest level of importance and a lower standard deviation. There was not too much variation in this particular attribute signifying most people tended to agree on the importance of this preference in selecting a bottle of wine. On the other hand, Grape variety had the highest level of standard deviation in the BW Score (Standard Deviation= 0.5823). This indicates there is a large variation in the respondent’s importance levels in regards to Grape variety. This attribute is important to a portion of the consumers but may not be important to all the wine consumers (Mueller and Rungie, 2009).
Figure 3. Attribute importance and heterogeneity
0.6500 Tasted the wine previously
0.6000 0.5500
Someone recommended it QuickTime™ and a decompressor are needed to see this picture.
Heterogeneity
Grape variety QuickTime™ and a decompressor are needed to see this picture.
0.5000
I read about it
0.4500
Brand name
0.4000
Origin of the wine Matching to my food
0.3500
Information on the shelf
0.3000
An attractive front label
Importance
0.2500 Information on the back label
0.2000 -0.800 0
-0.600 0
-0.400 0
-0.200 0
0.0000
0.2000
0.4000
0.6000
0.8000
Promotional display in-store Medal/Award
Average BW Score
Alcohol level below 13%
Attribute importance and heterogeneity based on the average BW scores and the standard deviation for each wine attribute.
4.3.1 Attribute comparison: Demographic and geographic Independent sample t-tests were conducted to find statistical significant differences between the wine attributes BW means in regards to multiple demographic and geographic variables. Significant differences were found in several of the wine attribute BW means when tested against independent variables such as gender, generation, education, number in household, marital status, income, and U.S. region. For example it can be seen that in regards to gender, the wine attributes of Origin of the wine, Match to my food, and I read about it were significantly different with the male respondents having the higher BW mean of 0.5701, 0.3738, 0.6262 respectively. Attractive front label, Brand name, and Tasted it previously were significantly different resulting in females having the higher BW mean of 0.2092, 0.5098, and 2.7059 respectively. Table 7 displays the Best and Worst scores for the gender, generation, and income categories as well as their corresponding t scores and pvalues. It can be seen that Generation Y had a significantly higher mean for Promotional display, Attractive front label, and Someone recommended it. On the other hand Remaining Generations had a higher mean for Grape variety and I read about it. The corresponding t scores and p-values can be seen typed in bold lettering in Table 7. In regard to income levels Grape variety, I read about it, and Attractive front label were found to be significantly different between categories with Lower Income having the higher mean for the latter. Refer to Table 7 for the actual data values and statistical significance.
Relationship, Household and Education level categories found very little significant differences in the BW average scores. Only Attractive front label was found to be significantly different in Relationship with not being in a relationship as the higher BW mean. There were no significant differences in the Household or Education level categories. Geographically there were significant differences between the United States regions when it came to the following BW attribute means: Grape variety, Origin of the wine, Brand name, Someone recommended it, I read about it, and Tasted previously. The values obtained from the region Chi-Square test are displayed in Table 8.
Table 7: Significant differences in Gender, Generation, Income BW scores
Attribute
Tasted the wine previously Someone recommended it
t score (pvalue)
Generation Y
Remaining Generations
t score (pvalue)
Low Income
High Income
t score (pvalue)
Male
Female
2.2056
2.7059
-2,361 (,019)
2.4872
2.5105
-,110 (,912)
2.4211
2.5616
-,663 (,508)
1.7009
1.6667
,166 (,869)
1.9915
1.4266
2,799 (,006)
1.8947
1.5137
1,868 (,063)
Grape variety
0.7944
0.2484
1,869 (,063)
-0.0085
0.8671
-3,020 (,003)
-0.0702
0.8973
-3,390 (,001)
I read about it
0.6262
0.1373
2,413 (,017)
0.1111
0.5245
-2,056 (,041)
0.0439
0.5684
-2,615 (,009)
Brand name
0.0187
0.5098
-2,301 (,022)
0.2821
0.3287
-,219 (,827)
0.4035
0.2329
,799 (,425)
Origin of the wine
0.5701
-0.0523
2,561 (,011)
0.0427
0.3357
-1,207 (,229)
0.1404
0.2534
-,463 (,643)
Matching to my food
0.3738
-0.2222
2,261 (,025)
-0.2564
0.2517
-1,943 (,053)
-0.0439
0.0753
-,452 (,652)
-0.3458
0.0327
-1,819 (,070)
-0.1026
-0.1399
,180 (,857)
-0.1491
-0.1027
-,223 (,823)
-0.6075
0.2092
-3,415 (,001)
0.4957
-0.6364
4,894 (,000)
0.3333
-0.4863
3,408 (,001)
-0.4766
-0.7451
1,245 (,214)
-0.5897
-0.6713
,381 (,703)
-0.5351
-0.7123
,827 (,409)
-1.0654
-0.6471
-1,879 (,061)
-0.4444
-1.1259
3,131 (,002)
-0.7281
-0.8904
,731 (,466)
Information on the shelf An attractive front label Information on the back label Promotional display in-store
Medal/Award
-0.9533
-0.8105
-,618 (,537)
-1.0513
-0.7203
-1,454 (,147)
-0.8333
-0.8973
,279 (,780)
Alcohol level below 13%
-2.6729
-2.6863
,073 (,942)
-2.8632
-2.5315
-1,844 (,066)
-2.6491
-2.7055
,310 (,756)
Comparison between Gender, Generation, and Income variables on the attribute’s average BW scores. Bold typeface indicates significant difference between groups determined by independent sample t-tests.
Table 8: Significant differences in Region BW scores
Comparison between U.S. region categories on the attribute’s average BW scores. Bold typeface indicates significant difference between groups determined by Pearson Chi-Square test.
4.3.2 Attribute comparison: Behavioral The significant differences in wine attribute means between the Purchase Frequency and Bottle Purchase categories can be seen in Table 9. It is clear both Grape variety and Origin of the wine were found to have significantly different means within the two groups. In both cases High Purchase Frequency and High Bottle Purchase had the higher mean. So those respondents who are buying wine more often and purchasing a larger quantity of bottles are significantly more influenced by the grape variety and the origin of wine when deciding on a wine to purchase. The wine attributes of Promotional display, Attractive front label, and Brand name were also found statistical significant in regards to mean differences within the groups of Purchase Frequency and Bottle Purchase. Low Bottle Purchase and Low Purchase Frequency had the higher mean for these particular attributes. This indicates the respondents who buying wine less often and purchasing a smaller quantity of wine are significantly more influenced by the Promotional display, Attractive
front label, and Brand name selecting a wine to purchase. As a person can see, the Purchase Frequency and the Bottle Purchase category means produced very similar results when tested for statistically significant differences. Since they are both purchasing behavior variables the similar results is a good notion to the validity in the studies response results. The remaining behavioral variables, Involvement and Loyalty means, will be explored further in the following Generation Y Segmentation section.
Table 9: Significant differences in Purchase variable BW scores Purchase Freq
Bottle Purchase
No.
Attribute
Low Purchase Frequency
High Purchase Low Bottle t-test (p-value) Frequency Purchase
High Bottle Purchase
t-test (pvalue)
13
Tasted the wine previously
2.4839
2.5405
,243 (,808)
2.5207
2.4615
,663 (,508)
11
Someone recommended it
1.7419
1.5270
-,953 (,341)
1.8462
1.3736
2,040 (,042)
2
Grape variety
0.1935
1.1757
3,506 (,001)
0.0828
1.1978
-4,845 (,000)
12
I read about it
0.2312
0.6081
1,696 (,091)
0.2130
0.5714
-2,762 (,006)
10
Brand name
0.4731
-0.1081
-2,501 (,013)
0.4970
-0.0440
2,824 (,005)
3
Origin of the wine
0.0000
0.7162
2,976 (,003)
-0.0710
0.7143
-3,376 (,001)
6
Matching to my food
-0.1022
0.3378
1,522 (,129)
-0.1243
0.2967
-1,868 (,063)
4
Information on the shelf
-0.0161
-0.3919
-1,654 (,099)
-0.0828
-0.1978
,751 (,453)
9
An attractive front label
0.0753
-0.6351
-2,702 (,007)
0.1479
-0.6371
3,188 (,002)
7
Information on the back label
-0.6344
-0.6351
-,003 (,998)
-0.6686
-0.5714
-,462 (,644)
1
Promotional display in-store
-0.6667
-1.2027
-2,213 (,028)
-0.6095
-1.2088
2,899 (,004)
8
Medal/Award
-0.7957
-1.0541
-1,027 (,305)
-0.7219
-1.1429
1,916 (,057)
5
Alcohol level below 13%
-2.6559
-2.7432
-,438 (,662)
-2.6568
-2.7253
1,155 (,249)
Comparison between Purchase Frequency and Bottle Purchase variables on the attribute’s average BW scores. Bold typeface indicates significant difference between groups determined by independent sample t-test
4.5 Segmentation on Generation Y A main focus of this study was to segment and define characteristics of wine consumers based on the resulting BW scores. Exploring the BW scores throughout demographics provided a good start in discovering differences among particular groups. Before segmenting the US wine consumers using latent class analysis it seemed necessary to examine and get a better idea of the Generation Y category. This is largely in part because Generation Y was a specific focal point for this research paper. Developing these characteristics on Generation Y will help to foresee possible characteristics in the consumer segments as well. 68% of Generation Y respondents are making an income of $49,999 or less. Close to three-fourths of the Generation Y respondents are buying wine less than once a week and 80% are buying less than seven bottles a month. Generally, it seems the majority of the Generation Y respondents are of lower income buying a less amount of bottles less often. The number of respondents in Generation Y compared to the Remaining Generations for each variable category can be seen in the first two columns on Table 10.
Figure 4: Generation Y comparison on average BW scores 0.8000 0.6000 Generation Y 0.4000 Remaining Generations
0.2000 0.0000 13
11
9
10
12
3
2
4
6
1
7
8
5
-0.2000 -0.4000 -0.6000 -0.8000
Attribute Number
Comparison of Generation Y and Remaining Generations on the average BW scores of wine attributes that influence consumer purchase decisions in the U.S. wine market. Corresponding attribute name to the number can be found in the previous Table 9.
Figure 4 compares the average BW scores on wine attributes between Generation Y and the Remaining Generations. A person can see there is a variance in particular to the middle attributes. Chi-square tests were conducted to find significant differences on the between the generation categories (Generation Y and Remaining Generations) against other variables. The Chi-square analysis results are displayed in Table 10 along with all the chi-square values and p-values for each comparison. When tested against Generation, no statistical significant differences were found between the variables in Region, Purchasing Frequency (high or low), and Household (living alone vs. with one or more people). However there were statistical significant differences between the generational and income groups. This makes logical sense since people in the 20s and early 30s age group may still be in school or just starting out in their career. On the other hand, someone in the 40s or 50s age group have more work experience and are probably therefore making a significant amount more in income. There were also statistical differences in Gender, Education, Relationship and Bottle Purchase. These differences and their corresponding values can all be seen in Table 10.
An ANOVA was conducted on the Involvement and Loyalty mean for significant differences between Generation groups. The values and significance can be seen in Table 11. Statistically, neither was found to have significant differences between groups. Comparing these generation groups to other variables provided some significant differences. However, a person would think the involvement levels would differ as well. These results all just signify the importance of further analysis. In many of the previous research studies this is the point in which the analysis would stop and the discussion on the comparisons would happen. Previous authors also mention that the analysis needs to be taken a step further in order to segment the U.S. wine market and compare for similarities and differences (Goodman ; Cohen et al., 2009). Using the BW scores to segment the U.S. wine market will help to produce knowledge particular variables important in understanding consumer purchase selection of wine (Lockshin et al., 2001).
Table 10: Characteristics of Generation categories Variable
Demographic Gender Male Female Education High school University or higher Income Low: 49,999 and below High: $50,000 and above Household Living alone Living with one or more Relationship In a relationship Not in a relationship Geographic Region West Midwest South Northeast Behavioral Purchase Frequency Low: Less than once a week High: Once a week or more Bottle Purchase Low: Less than seven/month High: Seven or more/month
Remaining Generation Generations Y (N) (N)
38 79
69 74
11 106
30 113
79 38
35 108
19 98
20 123
77 40
Pearson ChiSquare
p-value
Degrees of Freedom
6.611
0.010
1
6.494
0.011
1
48.428
0.000
1
0.256
0.613
1
11.488
0.001
1
7.235
0.065
3
1.411
0.235
1
22.009
0.000
1
120 23
18 77 14 8
35 71 21 16
88 29
98 45
94 23
75 68
Total 117 143 Comparison between generation categories and remaining variables for significant differences. Bold typeface indicates significant difference between groups determined by Pearson Chi-Square test.
Table 11: Generation comparison on Involvement and Loyalty Variable
Involvement Loyalty
Gen Y
Remaining Generations
F
p-value
5.03
5.28
2.147
0.144
2.9516
3.1562
2.990
0.085
ANOVA test on Generation categories and the Involvement and Loyalty means. Bold typeface indicates significant difference between groups.
4.6 Latent Class Cluster Analysis Now that the average BW scores have been discussed and variable groups have been compared, it is time to segment the sample based on the average BW scores. There are multiple analytical ways in which to segment consumers but Latent Class Cluster analysis (LCA) will be implemented in this report. LCA positioned wine consumer respondents into groups, known as clusters, based on average BW scores. The latent class cluster analysis software program called Latent Gold 4.5 was used to conduct the cluster segmentation. Its quick and easy applicability to practicality is a major reason to why LCA was used in this study as well as why LCA has become an increasingly popular research tactic (Vermunt & Magidson, 2002; Magidson & Vermunt, 2004). LCA was also appealing to this particular study because of its clear advantages over other standard clustering techniques (e.g. K-means). First off a statistical model is made for the population in which the sample was taken and the individual cases are placed into a cluster based on probability distributions (Vermunt & Magidson, 2002). LCA also uses particular diagnostic conditions and fit statistics to determine the number of clusters that should be in the model (Askegaard et al., 2010; Vermunt & Magidson, 2002). This process doesn’t assume linearity, normally distributed data, or homogeneity of variances (Askegaard et al., 2010; Vermunt & Magidson, 2002). Lastly, LCA can be projected using any combination of variable scale types such as categorical or continuous (Askegaard et al., 2010; Vermunt & Magidson, 2002).
The Latent Cluster Analysis determined models differing in the number of clusters developed. Table 12 displays the models and their corresponding Bayesian Information Criterion (BIC) and Akaike’s Information Criterion (AIC) scores (Askegaard et al., 2010). The BIC and AIC are selection criteria that measure the goodness of fit for a particular model (Kuha, 2004). It can be seen that the BIC decreases as another cluster is added. As a guideline to the best fitting model, the BIC should be small compared to the other models (Askegaard et al., 2010). The model that best fits this criterion was the four-cluster model (BIC=13256.70 and AIC=12754.65). This is the best fitting model because it contains the lowest BIC and AIC right before the BIC starts increasing again.
Table 12: Latent class cluster models Model
BIC
AIC
One-cluster model
13450.95
13098.44
Two-cluster model
13351.87
12949.51
Three-cluster model
13302.47
12850.26
*Four-cluster model
13256.70
12754.65
Five-cluster model
13286.20
12734.30
Cluster models proposed from the latent class cluster analysis. The Bold typeface indicates the best fitting model.
Now that the four-cluster model is developed, the mean BW score can be compared throughout the clusters. The individual cluster sizes and attribute’s average BW score are shown in Table 13. Figure 5 makes it visually easier to see the variation in the four cluster’s average BW scores. The four-cluster model variable was then inputted into SPSS to allow for group comparisons to be made for significant differences in means. This will further help to characterize the clusters developed by this model.
Table 13. Latent class cluster analysis attribute mean scores Cluster 1
Cluster 2
Cluster 3
Cluster 4
p-value
R2
Promotional display
-0.8776
-1.8299
0.5423
-0.2630
3.80E-08
0.2093
Grape variety Origin of the wine Information on the shelf Alcohol level below 13 Match to my food Information on back label Medal or Award Attractive front label Brand name Someone recommended it I read about it Tasted the wine previously Segment size N (=260)
0.7523 0.3626 -0.8597 -2.7384 -0.2738 -1.4886 -0.9591 -0.1418 1.4725 1.9574 0.2303
1.4240 0.9005 0.1203 -2.7707 0.9243 0.0858 -0.8405 -1.9142 -1.2679 1.8252 1.4306
-2.5595 -2.2203 0.8762 -2.2874 -0.8787 -0.3668 -0.6524 1.2053 0.7985 2.5072 0.1961
1.4894 1.3350 0.2653 -2.8194 0.1512 0.0238 -0.9389 1.9457 -0.4353 -0.4566 -1.4277
0.00016 1.60E-05 3.30E-05 0.27 0.00028 8.00E-06 0.72 1.60E-10 3.50E-08 1.80E-06 6.20E-09
0.3749 0.3549 0.1567 0.0158 0.0928 0.1759 0.0037 0.4863 0.4731 0.2938 0.2977
3.2029
2.1072
2.7048
1.0121
5.30E-05
0.1942
Wine Attribute
74
44
35
41 28 17 13 Table 13 shows the latent class cluster analysis attribute mean scores, indicator estimates and the segment size for the four-cluster model.
Figure 5: Comparison of cluster mean BW scores 4.0000 3.0000 2.0000 1.0000 Cluster 1
0.0000 -1.0000
Cluster 2
-2.0000 Cluster 3 -3.0000 Cluster 4
ly io us pr ev
Ta st ed
Ir
ea d
ab ou t
it
it
e m en de d
na m So m
eo ne
re co m
an d Br
ab el
ar d
nt l fro e
At tra ct iv
la be l
M
on at io n
In fo rm
ed al or Aw
fo od m
ba ck
y
13 to
M at ch
th e
Al co ho ll
In fo rm
at io n
on
of t rig in O
ev el be lo w
sh el f
w in e he
va r ra pe G
m ot io na l
di sp
la y
ie ty
-4.0000
Pr o
%
107
Wine Attributes
Figure 5 displays the wine attribute BW score pertaining to the individual clusters in the fourcluster model.
The last portion of the Analysis and Results section combines the Latent Class Cluster Analysis along with the Chi-square and ANOVA analysis between variable groups to develop a description of these four clusters of U.S. wine consumers. The four segments can now be characterized and discussed based upon the study results. The demographic, geographic, and behavioral characteristics and/or significant differences will first be specified for each cluster. During the cluster descriptions please refer to the following tables for more information. The demographic, geographic, and purchasing characteristics per individual cluster can be seen in Table 14 and Table 15 as well as the significant differences among category groups. Duncan post hoc tests were performed to find significantly different means in Purchase location, Involvement, and Loyalty and the results can be seen in Table 16 and Table 17. Lastly, the individual BW attribute scores for each cluster will then be indicated and compared among clusters
4.6.1 Cluster 1 Cluster 1 is the largest segment (N=107 respondents) with 41% of the total consumer respondents. Majority of the respondents in this segment were female (N=73, 68.2% within the cluster) with representation of Generation Y (N=44) and the Remaining Generations (N=63) as well as Low (N=45) and High Incomes (N=62). Most of the respondents in this group are living in a household with two or more people (N=88, 82.2% within the cluster) and are in a relationship of some kind (N=79, 73.8% within the cluster). This particular segment contains both High and Low purchase frequency and bottle purchases but in respect to the other segments, it has the majority of the high purchase frequencies and high bottle purchases along with Custer 2. Cluster 1 has an individual involvement mean of 5,10 meaning the respondents consider themselves involved with the wine product by slight agreement to the involvement statements. At the same time they had the highest loyalty mean (3.29, significantly higher then Cluster 4). The significance can be seen in the Duncan post hoc test results in Table 17.
4.6.2 Cluster 2 The second largest developed segment is Cluster 2 with 28% of the total consumer respondents (N=74). This segment had the largest representation of male (N=45) and was
skewed to the Remaining Generation respondents. This segment tends to have a higher income (N=49, 66.2% within the cluster) as well as being in a relationship and living in a household with two or more people over 21. They buy more bottles and more often compared to the other clusters (High Purchase Frequency N=30, High Bottle Purchase N=38). Cluster 2 is buying significantly more wine at wine shops compared to Clusters 3 and 4. On the other hand this segment is buying significantly less wine at grocery stores compared to the other three clusters. This segment is also buying significantly more wine at a restaurant in comparison to Cluster 4. These significances are further detailed in Table 16. This segment was considered to be the most involved by having a higher degree of agreement to the involvement statements (mean=5.76). On the other hand they don’t consider themselves to be extremely brand loyal (mean=2.95).
4.6.3 Cluster 3 17% of the respondents were segmented into Cluster 3 (N=44). This segment is slightly more of females but this is probably just due to the fact that there were more female respondents than male in the total sample. This segment is the only segment that is largely represented by Generation Y (N=29, 65.9% within the cluster). 29 respondents may not seem very high but when considering the Remaining Generation amount in this group N=15) and the Generation Y total of 117 it is significant enough. The remaining of the Generation Y respondents are subgroups within the other clusters that will be discussed in more detail further on in the Discussion chapter of this report. This segment tends to be of lower income (N=26, 59.1% within the cluster), which makes sense being the younger age group is more likely to make a lower income. Opposite from Cluster 2, almost all the consumer respondents in Cluster 3 are buying a low amount of wine and at a less often frequency (Low Purchase Frequency =90.9% respondents within the cluster, Low Bottle Purchase=84.1% respondents within the cluster). Cluster 3 bought the most amount of wine at liquor stores. The members of this segment were significantly low in buying wine at a wine shop compared to Cluster 2 (Table 16). Compared to the other clusters, Cluster 3 has the lowest Involvement mean (4.33). The cluster still may be slightly involved (scale 1=strongly disagree and 7=strongly agree on involvement statements) but the mean is considered significantly lower than the other three clusters (p-value=0.000, Duncan post
hoc test, Table 17). Cluster 3’s loyalty mean was on the low end (mean=2.98) but wasn’t considered to be significantly different than the other clusters (Duncan post hoc test, Table 17). So far Cluster 3’s characteristics sound similar to the Generation Y characteristics previously mentioned.
4.6.4 Cluster 4 The last and smallest group is Cluster 4 with 13% of the consumer respondents (N=35). This cluster again had a larger portion of female representation. So only Cluster 2 was unique in that it was largely male respondents. Cluster 4 was almost split 50/50 in regards to generational age. All of the clusters had mostly a high level of education but Cluster 4 had the highest percentage of respondents with a higher education (97.1%, N=34). However, according to the Chi-Square analysis test there were no significant differences in regards to Education level between the cluster groups (see Table 14 for significance). Again this can be explained because almost all the respondents were at a high education level to begin with (N=219, 84.2% of total respondents). Cluster 4 was split on income levels and showed similar results to other clusters in regards to Household size and Marital Status. As far as purchasing behavior goes, Cluster 4 had a mix of levels but leaned more towards the lower purchase frequency and lower bottle purchase. The segment’s involvement mean was 5.17 and significantly different from Cluster 2 and Cluster 3 (p=0.000, Post Hoc test Table 17). Cluster 4 had the lowest loyalty mean (Mean= 2.73) that significantly differed from Cluster 1’s much higher mean of 3.29 (Table 17). It should be noted that both the Household size and Education level were found to have no significant differences among the category groups. This could mainly due to the fact that the majority of the respondents in this research sample were not living alone and were a university graduate or higher. In each cluster segment, all four Regions were also represented. However, the Midwest was the highest in every segment due to the large amount of results from Midwest consumers.
Table 14: Cluster demographic and geographic comparison Cluster 1
Cluster 2
Cluster 3
Cluster 4
34 31.8 73 68.2
45 60.8 29 39.2
17 38.6 27 61.4
11 31.4 24 68.6
44 41.1 63 58.9
26 35.1 48 64.9
29 65.9 15 34.1
18 51.4 17 48.6
19 17.8 88 82.2
13 17.6 61 82.4
8 18.2 36 81.8
1 2.9 34 97.1
45 42.1 62 57.9
25 33.8 49 66.2
26 59.1 18 40.9
18 51.4 17 48.6
19 17.8 88 82.2
8 10.8 66 89.2
7 15.9 37 84.1
5 14.3 30 85.7
79 73.8 28 26.2
65 87.8 9 12.2
31 70.5 13 29.5
22 62.9 13 37.1
21 19.6 68 63.6 15 14 3 2.8
21 28.4 30 40.5 9 12.2 14 18.9
6 13.6 31 70.5 6 13.6 1 2.3
5 14.3 19 54.3 5 14.3 6 17.1
Gender Male % within cluster Female % within cluster Age Groups Generation Y: 18-32 % within the cluster Remaining Generations: 33-90 % within the cluster Education High school % within the cluster University or higher % within the cluster Income Low Income: $49,999 and below % within the cluster High Income: $50,000 and above % within the cluster Household Size One or less age 21 and over % within the cluster Two or more age 21 and over % within the cluster Marital Status In a relationship % within the cluster Not in a relationship % within the cluster US Region West % within the cluster Midwest % within the cluster South % within the cluster Northeast % within the cluster
Pearson Chi-Square
Degrees of Freedom
0.001
3
0.008
3
0.166
3
0.043
3
0.637
3
0.019
3
0.001
9
Table 14 shows the differences between clusters in their demographic and geographic variables. Bold typeface indicates significant difference between groups by Pearson Chi-Square tests.
Table 15: Cluster purchasing comparison Cluster 1 Purchase Frequency Low Purchase Frequency: Less than once a week % within the cluster High Purchase Frequency: Once a week or more % within the cluster Bottle Frequency Low Bottle Purchase: Less than seven bottles/month % within the cluster High Bottle Purchase: Seven or more bottles/month % within the cluster
Cluster 2
Cluster 3
Cluster 4
77 72
44 59.5
40 90.9
25 71.4
30 28
30 40.5
4 9.1
10 28.6
73 68.2
36 48.6
37 84.1
23 65.7
34 31.8
38 51.4
7 15.9
12 34.3
Pearson Chi-Square
Degrees of Freedom
0.004
3
0.001
3
Table 15 shows the differences between clusters in their behavioral variables. Bold typeface indicates significant difference between groups by Pearson Chi-Square tests.
Table 16: Cluster purchase location comparison Cluster 1
Cluster 2
Cluster 3
Cluster 4
Purchase Standard Standard Standard Standard Mean Mean Mean Mean Location Deviation Deviation Deviation Deviation Liquor Store 1.71 1.92 1.49 1.87 2.14 2.09 2.11 2.01 Wine Shop 0.89 1.44 1,27a 1.57 0,70b 1.03 0,46b 0.92 Winery 0.47 1.02 0.68 1.20 0.27 0.87 0.43 0.92 Grocery Store 1,27a 1.65 0,45b 0.92 1,25a 1.62 1,63a 1.94 a Restaurant 0.52 0.94 0,68 0.95 0.50 0.85 0,20b 0.47 Internet 0.05 0.25 0.20 0.64 0.09 0.47 0.03 0.17 Other 0.05 0.25 0.18 0.67 0.05 0.21 0.14 0.85 Table 16 shows the differences across the clusters in their purchasing locations. Different superscripts indicate significantly different means following a Duncan post hoc test.
F
pvalue
1.44 3.37 1.49
0.232 0.019 0.218
6.58 2.33 2.28 1.19
0.000 0.075 0.080 0.313
Table 17: Cluster Involvement and Loyalty comparison Cluster 1
Cluster 3
Cluster 4
Mean
Standard Deviation
Mean
Standard Deviation
Mean
Standard Deviation
Mean
Standard Deviation
F
pvalue
5.10a
1.32
5.76b
1.39
4,33c
1.37
5.17a
1.06
11.09
0.000
3.29a 1.04 2.95 0.83 2.98 0.78 2.73b 0.99 3.96 Table 17 shows the differences across the clusters in their involvement and loyalty means. Different superscripts indicate significantly different means following a Duncan post hoc test.
0.009
Individual Involvement Mean Individual Loyalty Mean
Cluster 2
4.6.5 Attribute Differences Among Clusters The variation between the cluster’s 13 attribute BW scores is evident back in Figure 5. All the changing and crossing of the graph lines signifies these differences between the cluster groups. The description of attribute preference differences will be detailed in the order of this studies total BW scores for U.S. wine consumers starting with the most preferred attribute, Taste the wine previously (BW average=2.5000) to the least preferred attribute, Alcohol level below 13% (BW average=-2.6808). Remember the average BW scores can range from -4 to +4 indicating a strong preference for that attribute or a very weak preference respectively.
All fours clusters differed in their average BW scores in regards to the attribute Tasted the wine previously. Cluster 4 in particular was low with a BW average score of 1.0121. This attribute isn’t really important to Cluster 4 respondents when deciding on a wine to purchase. On the other hand Cluster 1 places Tasted the wine previously at a much higher importance level (BW average=3.2029). Cluster 4 also scored the attribute Someone recommended it very low and actually negative in comparison to the other clusters (BW average= -0.4566). The remaining three clusters scored this attribute significantly higher than the total BW average= 1.6808, with Cluster 1 being the highest at BW average=2.5072. In fact these three other clusters resulted in the Someone recommended it
attribute as the most preferred or most important attribute overall. Grape variety was preferred on a very low level (BW average=-2.5595) by Cluster 3, the Generation Y cluster. This segment was the only cluster to rate this as a negative score meaning it was more of a least preferred attribute. Many of the other attributes would influence Cluster 3 purchasing decision on wine over this Grape variety attribute. Cluster 2 and Cluster 4 rated this attribute quite important (BW average= 1.4240; 1.4894 respectively) with Cluster 1 being just under them (BW average= 0.7523). Cluster 4 placed the I read about it attribute very low and negative importance level (BW average= -1.4277) while Cluster 2 placed it quite high (BW average= 1.4306 ) especially considering the total BW average amount (BW average= .3385). The Brand name attribute was found to be less important for Cluster 2 (BW average= -1.2679) and Cluster 4 (BW average= -0.4353). On the other hand the Brand name was a more important preference for Cluster 1 (BW average=1.4725) and Cluster 3 (BW average= 0.7985). The Origin of the wine attribute was a strong influencer in deciding on a wine to purchase for the respondents in Clusters 1, 2, and 4. However, the Generation Y cluster (Cluster 3) ranked this attribute extremely low (BW average= -2.2203) showing it was chosen as the least or worst attribute preference in many of the cases. This is a similar situation to the outcome of the Grape variety attribute.
Cluster 2 graded Matching to my food attribute as a higher preference in choosing their wine (BW average= 0.9243). This particular attribute rated as the least important by Cluster 3 (BW average= -0.8787). Cluster 3 does view Information on the shelf as being pretty important when selecting wine (BW average= 0.8762). The Information on the shelf is least important for those respondents in Cluster 1 (BW average= -0.8597). The four clusters vary greatly on the An attractive front label attribute. It can be seen back in Table 6 that this attribute has a high standard deviation compared to the other attributes. Cluster 4 views An attractive front label as their most imperative attribute in choosing a wine. This segment preferred this attribute the most to all the other attributes (BW average= 1.9457). Cluster 2 was the opposite and gave this attribute one of their lowest preference marks (BW average= -1.9142). Information on the back label was considered the second least important wine preference for Cluster 1 (BW average= -1.4886). The remaining three clusters saw this attribute as right in the middle of their preferences. The Promotional
display in-store seemed to have the most influence on Cluster 3’s wine choices (BW average= 0.5423) and the least influence on Cluster 2’s preference of wine choice (BW average= -1.8299). Cluster 3 was the only cluster to view it positively as being a preference of choice over other preference attributes. Despite the strong presence for choosing this attribute over others, the total BW average was -.8192 showing an overall lean towards Promotional display in-store being less important or preferred in choosing a wine. All four clusters didn’t strongly label the Medal/Award attribute as being the most or least important. It tended towards the negative side but Cluster 3 had the highest score (BW average= -0.6524) meaning they were just slightly more influenced by a wine with a medal or an award. However the differences between the clusters BW means in regards to this attribute were not found to be statistically significant (p-value=XXXXX). For Clusters 1, 2, and 4 the Alcohol level below 13% was considered the least important attribute. The attribute was slightly higher for Cluster 3 respondents making it only their third least important attribute.
Any of these variances can be connected back to Figure 3. The attributes towards the top of this graph (higher standard deviations) are the same attributes with some differences between clusters. None of the clusters exactly matched the total BW average scores calculated in the beginning of the analysis. All of the clusters seemed to score higher or lower, in other words more extreme, than those scores. This is sign that those scores are dulled down a bit because of the averaging of some of the segmented groups more extreme scoring. This just proves the need to segment the U.S further and not just solely rely on the overall BW average scores to develop marketing strategies. This issue along with the cluster characteristics and overall study conclusions will further be discussed in the following section.
Part IV
5. Discussion By first looking at the results of the overall Best Worst method analysis, a person can get a general idea of the US wine attribute preferences that affect consumer purchasing decisions. This portion of the study obtained similar basic results to the previous research mentioned in the Literature Review section. For example these results were similar to Goodman (2009) U.S. and Australian results in that Tasting the wine previously and Someone recommend it scored high in attribute importance along with Grape variety and Origin of wine. The pertinent similarities are a strong reason why Australian wines are so successful in the U.S. (Goodman 2009). Because of the nature of the method, these scores can easily be compared across cultures (as in Goodman, 2009). This can be extremely advantageous in such a strong global wine market. Other countries could compare their own attribute preferences to the U.S. and therefore realize where they need to change their marketing strategy and tactics in order to make their wines successful in the U.S. wine market.
Investigating the U.S. Best Worst scores further shows large variation among certain variable groups. In particular Generation Y and the Remaining Generations had multiple significant differences in regards to the attribute scores and the other demographic and behavioral variables. These differences along with others indicate a need for further segmentation. The BW scores became too general when combining all U.S. wine consumers into one large segment. The variation of U.S. wine consumers’ views on wine attributes is too high to treat them all the same. The high variation in the products choices and the large fragmentation in the wine market might have a large affect on these differences in groups. A different simpler product may be able to group the U.S. as one whole segment when using Best Worst analysis. As seen in this research, really high attribute scores of a certain group canceled out the really low scores of another when they were all viewed together. For example this could be a crucial issue if a company decides to place their focus on the wine grape variety in their marketing strategy. As seen in this
study, the overall Grape variety scored a high BW score relative to some of the other attributes. However after further analysis it was found that Cluster 3 sees Grape variety as one of their worst preference attributes. The company would then be marketing their specific strategy to a group that wouldn’t respond and buy the product on those tactics. Broad mass marketing solely on the overall U.S. BW scores could have potential to create great inefficiency and ineffectiveness in the companies marketing strategy. For this reason it is believed that marketing based solely on Best Worst analysis may not be the most profitable strategy for a wine company. A similar segmentation process, as seen in this report, needs to be conducted to further define segments based on the differences in the Best Worst scores or attribute preferences.
This report using Latent Class Cluster Analysis to segment the U.S. wine market resulted in four cluster segments. Each cluster has certain characteristics and attribute differences that could aide marketers in developing their strategies and tactics. Cluster 1 is mostly of female representation and a combination cluster in regards to the other demographics. This cluster contains the most brand loyal consumers. It had a higher Brand name attribute score compared to any other cluster. Consumers in Cluster 1 are buying more wine, more often and look to wine attributes referring to the whole wine experience, knowledge and the actual wine product qualities itself (intrinsic cues). The wine preference attributes of Tasted previously, Someone recommended it, Brand name, Grape variety, and Origin of wine were the most influential in Cluster 1’s wine choices. Members of Cluster 2 are buying a high volume and frequency of wine as well but this segment contains more of the older (variable group Remaining generations) male population in some sort of a relationship. They are the wine enthusiasts who are the most involved and not extremely loyal to one brand of wine. They exemplify this by buying significantly more wine at the more specialized locations of wine shops. They are going out of their way to get a certain wine for example and not just conveniently purchasing wine grocery stores. They are knowledgeable on wine and don’t rely on the extrinsic attributes of Promotional display, Attractive front label, and Brand name to influence their wine decisions. They place a much higher preference level on Grape variety, Origin of the wine, and Match to my food and the attributes relying wine knowledge or experience such as Tasted previously,
Someone recommended it, and I read about it. Cluster 3 is the general Generation Y segment. The consumers in this group have a lower income and are buying low amounts of wine at a low frequency. They also hold the lowest involvement mean denoting this group of consumers to be taking less of an interest in wine as a product. This could probably be the reason they rely more on extrinsic attribute cues such as Information on the shelf, Promotional display, and Attractive front label. The attributes Someone recommended it and Tasted previously were also strong influencers in this consumer groups wine decisions. The younger group is looking to these attribute cues to help them with their decision process because they don’t have the involvement knowledge pertaining to Origin of wine and Grape variety for example. Lastly Cluster 4 is almost 100% of higher education members. The members are not very loyal in addition to being slightly involved with wine as a product. This group seems to rely more on intrinsic cues with the Grape variety and Origin of wine attributes as being some of their most important in choosing a wine. However, Attractive front label was their most important preference in wine selection. Medal/Award, Someone recommended it, I read about it attributes were some of their least important. That being said they seem to want to try new wines and not rely on other people’s judgment in their decision. This is more than likely the reason the Tasted previously attribute obtained a higher BW score. It could be they look at those intrinsic cues to narrow the search and then the Attractive front label attribute is the final cue that leads to the wine choice of purchase. They are a bit more adventurous in their buying habits in that respect compared to the other three clusters. A notable result in this study is that consumers from the Generation Y group are slightly present in the clusters other than Cluster 3 as well. So in essence the Generation Y group in regards to the product of wine can be separated into further subgroups based on their behavioral (involvement) characteristics. There is a strong Generation Y group with their characteristics specific to Generation Y. Then there are also some Generation Y members who may be more involved in wine and more of a wine enthusiast causing a subgroup in another cluster. So it seems inefficient for marketers to link the whole United States together as one consumer group and in smaller terms even the whole Generation Y wine consumers into one group. These subgroups should be taken as indications of some people
in Generation Y having more of an interest in wine. They may not be the most involved overall but according to the Wine Market Council (2009) they have the largest percentage of their members (20%) out of the generational groups belonging to at least one wine club. It clearly shows the interest and potential within the generation. The size and growing marketing power in Generation Y wine consumers make this age group an attractive focus. However not all those in Generation Y are full blown wine enthusiasts purchasing large amounts all the time. For this reason there is a strong need for a portion of (particularly those in Cluster 3) these consumers to become more aware of wine as a product. Marketers could grasp this younger age group of lower consumption wine drinkers and even potential wine drinkers by creating that awareness and incentive to purchase wine in the minds of the consumer. The vivid opportunity is there to capture this Generation Y segment and the companies who tailor their marketing tactics to the specific segment characteristics will more than likely gain the most benefit.
6. Managerial Implications One of the main reasons for doing research is to offer recommendations and suggestions for way in which managers can implement the research process or results into their own companies. In this case it was important the research was applicable and beneficial to the wine market and its players in some way. Some examples of managerial implications have already been mentioned throughout the previous Discussion section. The information and results presented in this report provide further implications for the marketer or manager on how exactly to implement their segmentation strategy. This research study helped to develop the first criteria element in Schiffman et al. (2008) strategic marketing framework mentioned in the Literature Review. Now that the US wine consumers are segmented in some way, company managers and marketers can move on to the details of the other two elements; targeting a specific segment and positioning the product of wine in the mind of the consumer. When looking at targeting a wine segment the company can choose differentiate marketing or concentrated marketing. In differentiate marketing a company decides to target several of the segments or clusters and use personalized marketing mixes for each individual group (Schiffman et al., 2008 p 41). On the other hand the company
could chose to be more focused on one segment for example the Cluster 3 or Generation Y segment with a unique marketing mix as in concentrated marketing (Schiffman et al., 2008 p 41).
In regards to this report a company may decide to target only the Generation Y cluster because of the strong market potential and opportunity of gaining market share in an already extremely fragmented market. The Generation Y group seems to match the criteria for effective targeting of a market segment. The research results show it is an identifiable market with a sufficient growing size but at the same time it needs to be accessible in terms of both media and cost (Schiffman et al., 2008). It wouldn’t necessarily make sense for a company to market to a group that is declining or couldn’t be reached through any media measures. The company could use the characteristics and attribute information specific to this group to help choose how to most efficiently market to this group of wine consumers as they choose which wine to buy. It could continue to look at the Generation Y segments through other preference angles and other variable perspectives to obtain an even clearer notion of the segment and their purchase behavior. Knowing what the different Generation Y subgroups value in wine attributes will aide a company to better fit their wine products to consumer’s expectations and demands (Cohen, 2009). The steps followed in this research are easily something a company could implement into their own research strategies.
The third element of the marketing strategy involves positioning the product in the Generation Y consumers mind. This will be a key element as mentioned before because a certain general segment of Generation Y consumers are not the strongest wine enthusiasts. In the potential Generation Y consumers, wine as a product probably isn’t even positioned in their mind to purchase yet. So marketers face the challenge of convincing this Generation Y group that wine is the answer and essentially just making them aware. Similarly a company may want to shift the attitude of this general Generation Y segment so that wine becomes more common, everyday enjoyment (Wine Market Council, 2009). This third element is where companies need to do additional work with their marketing strategies in order to take the research results a set further. The Generation Y subgroup
characteristics have been identified in regards to preferences of attributes and it is now up to the company to chose the segment in which to market to and the media communication and strategy approaches in which they will do this.
Having companies focus on particular segment results such as this may assist the wine industry in producing wines that are the style that match the segments preferences. It will greatly aid in “minimizing the risk of losing existing customers through not meeting their preferences” (Goodman 2005). Following these results and the targeting and positioning steps allows for the wine market to be more efficient and effective. Ideally the company gets what they are aiming for and the customers are happy to be purchasing wine that matches their preferences. Overall it is a more accurate utilization of resources to match the wine market.
Part V
7. Limitations Similarly to any other research study, there were certain limitations in regards to this particular thesis report. Neither quota sampling nor random sampling was logistically possible so the study is based on a convenience sample of U.S. wine drinkers. Therefore it wasn’t exactly demographically representative of the whole United States. Respondents were contacted through wine networks and through the Internet in general. So those not connected to the Internet were obviously not represented in these results. More time was needed in order to contact and collect more responses. A greater number of respondents would have helped to reach a more accurate U.S. representation. For this reason the results give us a good idea of United States wine consumers but they are not exactly statistically representative of the United States. Due to the convenience factor a high portion of the respondents were from the Midwest region of the U.S. Although the U.S. is typically see
as one segment, it may not necessarily work to generalize these results if the Midwest provides a certain bias concerning the wine market.
This study only focused on segmenting the wine consumers and describing Generation Y drinkers so it didn’t go as deep into consumer characteristics as it could have. For example it is unclear in this study on the exact relationship the consumers have with wine; winemaker, distributor, just a drinker, etc. This may have put an influence on some of the consumer respondent answers. Didn’t specify too deeply into relation with wine example wine maker, producer, buyer, seller, etc. These ideas will be discussed more in the following Future Research section.
8. Future Research Beyond the managerial implications and limitations there are indications in this research report that could possible provide support for further research. Again, this study was to show the process of segmentation on BW attribute preferences and describe the Generation Y consumers more but it would always be useful to collect more descriptive data on the consumers. The more depth of characteristics on the consumer you can get, the better you know the consumer and their attribute preferences on wine. It may also be beneficial to explore geographic differences more in depth. There really haven’t been any studies proving no significant differences in wine consumers based on geographic U.S. regions. It could be possible those living in areas of prominent wine culture, wineries, etc have different wine attribute preferences then other areas. The respondents in this report were highly skewed to consumers in the Midwest region such. If this were the case it would be necessary to ensure significant consumer representation from each United States area or region. Now that this research has taken a step deeper into segmentation based on preference scores, it is possible to characterize these segments even more. The use of the BW method emphasizes the ease of a similar application deeper into the wine market and even into other areas. Other companies in the wine market could easily apply a similar study
particular to their own interests and goals. The attribute preferences used in this research where general and taken from previous research. However, other attributes could be studied even those attributes that are less concrete such as quality and price. Wine varieties may also play a part in differences among attribute preferences. Future researchers could explore more in the areas of wine knowledge and involvement levels in regards to these attribute preferences. Characteristics dealing more with psychographic and psychological aspects could be explored and added to the already BW analysis and segmentation. The person’s relationship to wine should be noted in further research. For example a wine maker or someone in the wine industry itself may place importance on certain attributes that a regular wine lover wouldn’t. This would allow for even a greater understanding of the wine consumer segments. In a broader sense a replication of this study could be made in other countries, other markets and even in other products. Comparisons can easily be made then across cultures as long as the attributes in the BW analysis are consistent. The level of applicability is quite large in this sense.
More and more research will continue to come out in regards to the Generation Y because of its popular focus now. It would be interesting to see more on the technology/Internet/social media and wine correlations in the U.S. Generation Y consumers. Applying more of a media approach to different marketing segments especially with the media’s strong presence in consumer’s everyday lives. This research study showed the process and advantages of using a method like this and provided information on the different Generation Y subgroups and the wine attribute preferences. It is beneficial to know more about the Gen Y relationship to wine and how exactly they should be marketed to in regards to wine and obtaining wine knowledge. Like many other studies, this research could be extended into longitudinal research. This would help to understand any changes or steadiness of the clusters consumer preferences, specifically the Generation Y cluster, over a period of time. In general research in this specific combination of areas is semi-limited so future similar studies could be conducted in order to check the validity of this report’s sample and results.
9. Conclusion Conclusions in regards to the three research statements mentioned in the Introduction section have already been discussed in the previous discussion section. The following are however, some general conclusions this thesis research can draw upon.
Further analysis and segmentation into the U.S. wine consumer attribute preferences shows that the U.S. has too high of a variance in its preferences to be linked together as one segment. In terms of wine a company needs to segment their market and chose a particular segment or segments to market to. This report provides possible steps to follow in order to create this segmentation and market on the characteristics perceived in the results.
It may not be beneficial for a company to solely segment on generational terms because as this research concluded, Generation Y for example has particular subgroups within the other segments. In the wine market, there is no distinct Generation Y age group that contains all Generation Y members. This could be true for other variables as well. It more than likely doesn’t work just to segment based on one variable. This thesis broadcasts this with its segmentation based on the wine attribute preferences and then further descriptions based on all the other demographic, geographic and behavioral variables. This just shows how important it is for a company to be specific in their segmenting and targeting strategies in order to gain the greatest efficiency and effectiveness. Marketing needs to be done based on the variance in the wine criteria importance between generational groups or any of the segment groups and not based on a mass marketing approach.
By taking this previous research a step further and following a portion of the research recommendations, the United States wine consumer market has the opportunity to be segmented even further and the segments to become even more detailed. As a result of the Generation Y segmentation into subgroups marketers now have a better idea of how to approach this growing and opportunistic segment in regards to the wine market and how to adhere their marketing tactics to properly fit this strategy.
10. Bibliography Aaker, D. A., McLoughlin, D. (2007). Strategic Market Management. European Edition. West Sussex, England: John Wiley & Sons, Ltd. Arnold, E. (2006). Younger drinkers reaching for wine. Wine Spectator, 31(2), p.15-15. Askegaard, S., Chrysochou, P., Grunert, K. G., & Kristensen, D. B. (2010). Social health practices in eating behaviour: A market segmentation approach. Results from: Health Branding project. Auger, P., Devinney, T., & Louviere, J. (2007). Using best-worst scaling methodology to investigate consumer ethical beliefs across countries. Journal of Business Ethics, 70, p. 299-326. Aurifeille, J. M., Quester, P.G., & Lockshin L. (2002). Global vs. International involvement-based segmentation: a cross national study. International Marketing Review, 19(4), p.369-386. Baenen, S. (2002) The long and winding road: US consumers and their changing attitudes towards wine, Wine Business Monthly, 9(6), 41-42 Barber, N. & Almanza, B. A. (2006). Influence of wine packaging on consumers' decision to purchase. Journal of Foodservice Business Research, 9(4), p.83-98. Barber, N., Almanza, B., & Donovan, J. (2006). Motivational factors of gender, income, and age on selecting a bottle of wine. International Journal of Wine Marketing, 18, p. 218-232. Barber, N., Almanza, B., & Dodd, T. (2008). Relationship of wine consumers' selfconfidence, product involvement, and packaging cues. (cover story). Journal of Foodservice Business Research, 11(1), p. 45-64. Barber, N., Dodd, T., & Ghiselli, R. (2008). Capturing the Younger Wine Consumer. Journal of Wine Research, 19(2), p. 123-141. Barber, N., Taylor, D. C., & Strick, S. (2010). Selective marketing to environmentally concerned wine consumers: A case for location, gender and age. Journal of Consumer Marketing, 27(1), p. 64-75. Beatty, S. & Talpade, S. (1994). Adolescent influence in family decision making: A replication with extension. The Journal of Consumer Research, 21(2), p. 332-341. Bruwer, J. & Johnson, R. (2010). Place-based marketing and regional branding strategy perspectives in the California wine industry. Journal of Consumer Marketing, 27(1), p. 5-16.
Bruwer, J., Li, E., & Reid, M. (2002). Segmentation of the Australian wine market using a wine-related lifestyle approach. Journal of Wine Research, 13(3), p. 217-242. Casini, L., Corsi, A. M., & Goodman, S. (2009). Consumer preferences of wine in Italy applying best-worst scaling. International Journal of Wine Business Research, 21(1), p. 64-78. Centers for Disease Control and Prevention (CDC). Department of Health and Human Services Alcohol Quickstats. Page last modified: 3 September 2008. Charters, S., & Fountain, J. (2006). Younger Wine Tourists: A study of generational differences in the cellar door experience. Global Wine Tourism Research Management and Marketing, p. 153-160. Cohen, E. (2009). Applying best-worst scaling to wine marketing. International Journal of Wine Business Research, 21(1), p. 8-23. Cohen, E., d’Hauteville, F., & Sirieix, L. (2009). A cross-cultural comparison of choice criteria for wine in restaurants. International Journal of Wine Business Research, 21(1), p. 50-63. Cohen, S.H. (2003). Maximum difference scaling: improved measures of importance and preference for segmentation. Sawtooth Software Conference Proceedings, Sequim, WA. Cohen, S.H. & Markowitz, P. (2002). Renewing market segmentation: some new tools to correct old problems”. ESOMAR 2002 Congress Proceedings. Amsterdam, The Netherlands: ESOMAR p. 595 -612. Cohen, S.H. & Neira, L. (2003). Measuring preference for product benefits across countries: overcoming scale usage bias with maximum difference scaling. ESOMAR 2003 Latin America Conference Proceedings, Amsterdam, The Netherlands. Constellation Wines (2005). Constellation Wines US releases results of landmark consumer study: Project Genome SM reveals much about premium wine consumers motivations, behaviors. Available at: www.cwinesus.com/ProjectGenome.html (accessed 23 February 2010). De Pelsmacker, P., Geuens, M., & Van den Bergh, J. (2007). Marketing Communications: A European Perspective. Essex, UK: Pearson Education Limited. Djamasbi, S., Siegel, M., & Tullis, T. (2010). Generation Y, web design, and eye tracking. International Journal of Human-Computer Studies, 68(5), p. 307-323. Dodd, T. & Bigotte, V. (1997). Perceptual differences among visitor groups to wineries. Journal of Travel Research, 35(3), p. 46.
Fernandez-Cruz, M. (2003). Advertising agencies target Generation Y. U-Wire: University of Kentucky. Available at: www.youngmoney.com/lifestyles/campus_life/031202_01. (Accessed 2 April 2010). Finn, A. & Louviere, J.J. (1992). Determining the appropriate response to evidence of public concerns: the case of food safety. Journal of Public Policy and Marketing, 11(1), p. 12-25. Gillespie, J. (2005) Broadening and strengthening the consumer base of the U.S. wine market. Presentation made at the 2nd International Wine Marketing Symposium, Rohnert Park, CA, July Gorman, P., Nelson, T., & Glassman, A. (2004). The Millennial Generation: A strategic opportunity. Organizational Analysis, 12(3), p. 255-270. Goodman, S. (2009). An international comparison of retail consumer wine choice. International Journal of Wine Business Research, 21(1), p. 41-49. Goodman, S., Lockshin, L., & Cohen, E. (2005). Best-Worst Scaling: A Simple Method to Determine Drinks and Wine Style Preferences. Paper presented at the 2nd International Wine Marketing Symposium Proceedings, Sonoma, CA, July. Goodman, S., Lockshin, L., & Cohen, E. (2006). Using the Best-Worst method to examine market segments and identify different influences of consumer choice. Paper presented at the 3rd International Wine Marketing Symposium Proceedings, Montpellier, July. Goodman, S., Lockshin, L., & Cohen, E. (2008). Examining market segments and influencers of choice for wine using the best worst choice method. Revenue internationale des Sciences Sociales- Special Wine Marketing Edition, 8(1), p. 94112. Harris Interactive (2001). Millennium’s first college grads are ‘connected, career-minded and confident-way’. Available at: http://www.harrisinteractive.com/news /allnewsby date. asp? NewsID=292. (Accessed 12 April 2010). Howard, R. & Stonier, J. (2002) Marketing wine to generation X. Australia and New Zealand Wine Industry Journal, 17(3), p. 78-82. Howe, N. & Strauss, W. (2000). Millennial rising: The next great generation. New York: Random House. Hussain, M., Cholette, S., & Castaldi, R. (2007). Determinants of wine consumption of US consumers: An econometric analysis. International Journal of Wine Business Research, 19(1), p. 49. Jones, J. (2006). U.S. drinkers consuming alcohol more regularly. Available at: http://www.gallup.com/poll23935/US-Drinkers-Consuming-Alcohol-More
Regularly.aspx. Jones, J. (2007). Beer again edge out wines as Americans drink of choice. Available at: http://www.gallup.com/poll/28234/Beer-Again-Edges-Wine-Americans-DrinkChoice.aspx. Jones, S. and Fox, S. (2009). Generations Online in 2009. Pew Internet & American Life Project. Available at: http://pewinternet.org/Reports/2009/Generations-Online-in2009.aspx. (Accessed 14 April 2010). Key Findings (2004). Understanding the echo boomers: Who they are and how you can reach this young, affluent market segment. Available at: www.keyfindings.com/healthcare /article2.htm. (Accessed 12 April 2010). Kuha, J. (2004). AIC and BIC. Sociological Methods and Research, 33(2), p. 188-229. Lancaster, L.C., & Stillman, D. (2002). When generations collide: How to solve the generational puzzle at work. New York: HarperCollins. Lockshin, L. & Hall, J. (2003). Consumer purchasing behaviour for wine: what we know and where we are going, paper presented at the International Wine Marketing Colloquiium, Adelaide, July, CD-ROM. Available at www.unisa.edu.au/winemarketing/conferences/ proceedings.asp. Lockshin, L., Jarvis, W., D’Hauteville, F., & Perrouty, J. (2006). Using simulations from discrete choice experiments to measure consumer sensitivity to brand, region, price, and awards in wine choice. Food Quality and Preference, 17, p. 166-178. Lockshin, L., Quester, P., & Spawton, T. (2001). Segmentation by involvement or nationality for global retailing: A cross-national comparative study of wine shopping behaviours. Journal of Wine Research, 12(3), p. 223-236. Lockshin, L., & Rhodus, T. (1993). The effect of price and oak flavor on the quality perceptions of consumers and wholesalers. International Journal of Wine Marketing, 5(2), p. 13-25. Lockshin, L. S., Spawton, A. L., & Macintosh, G. (1997). Using product, brand and purchasing involvement for retail segmentation. Journal of Retailing and Consumer Services, 4(3), p. 171-183. Louviere, J.J. & Islam, T. (2004). A comparison of importance weights/measures derived from choice-based conjoint, constant sum scales, and best-worst scaling. CenSoC Working Paper No. 04-003, (September). Louviere, J.J. & Woodworth, G.G. (1990). Best worst scaling: a model for largest difference judgments. Working paper, Faculty of Business, University of Alberta, Edmonton.
Magidson, J., & Vermunt, J. K. (2004). Latent class models. In: Kaplan, D.W., editor. Sage Publications, Inc. p. 175-198. The Sage Handbook of Quantitative Methodology for the Social Sciences. Mittal, B. & Lee, M. S. (1989). A causal model of consumer involvement. Journal of Economic Psychology, 10, p. 363-389. Morton, L. P. (2003). Targeting Generation X. Public Relations Quarterly, 48(4), p. 43-45. Mueller, S. & Rungie, C. (2009). Is there more information in best-worst choice data? International Journal of Wine Business Research, 21(1), p. 24-40. Neuborne, E., and Kerwin, K. (1999) Generation Y. Business Week, 15 Feb. 1999, p. 4650. Nowak, L., Thach, L., & Olsen, J. E. (2006). Wowing the millennials: creating brand equity in the wine industry. The Journal of Product and Brand Management, 15(5), p. 316. Orth, U.R., Lockshin, L., & d’Hauteville, F. (2007). The global wine business as a research field. International Journal of Wine Business Research, 19(1), p.5-13. Quester, P. G. & Smart, J. (1998). The influence of consumption situation and product involvement over consumers' use of product attribute. The Journal of Consumer Marketing, 15(3), p. 220. Rothschild, M.L. (1984). Perspectives on involvement: current problems and future directions. Advances in Consumer Research, Volume 11, p. 216-217. Saad, L. (2005). Wine gains momentum as American’s favorite adult beverage. The Gallup Poll Survey, July 18, 2005. Available at: http://www.gallup.com/poll/17335/WineGains-Momentum-Americans-Favorite-Adult-Beverage.aspx. Schiffman, L. G., Kanuk, L. L., Hansen, H. (2008). Consumer Behavior: A European Outlook. Essex, UK: Pearson Education Limited. Slama, M. E. & Tashchian, A. (1985). Selected socioeconomic and demographic characteristics associated with purchasing involvement. Journal of Marketing, 49, p. 72-82. Sparrow, N. (2006). Developing reliable online polls. International Journal of Market Research, 48(6). Spawton, T. (1991). Marketing planning for wine. European Journal of Marketing, 25(3), 6-48. SPSS for Windows version 17.0, Chicago: SPSS Inc.
Strauss, W. & Howe, N. (1992). Generations: The History of America’s Future, 15842069. William Morrow & Co. SurveyMonkey 2010. Available at: http://www.surveymonkey.com. (Accessed starting 5 March 2010). Thach, E. C. & Olsen, J. E. (2006). Market segment analysis to target young adult wine drinkers. Agribusiness, 22(3), p. 307-322. Thomas, A. & Pickering, G. (2003). Behavioural segmentation: A New Zealand wine market application. Journal of Wine Research, 14(2), p. 127-138. U.S. Census Bureau: Population Division. Available at: http://www.census.gov/population/www/ index.html. (Accessed 19 April 2010). Vermunt, J. K. & Magidson, J. (2002). Latent class cluster analysis. In: Hagenaars J. A., & McCutcheon A., editors. Applied Latent Class Analysis, p. 89-106. Cambridge: Cambridge University Press; 2002. Wine Market Council (2003). Wine Market Council’s 2003 Consumer Tracking Study. Wine Market Council (2006). Wine Market Council’s 2006 Consumer Tracking Study. Wine Market Council (2009). Wine Market Council’s 2009 Consumer Tracking Study. Available at: http://www.winemarketcouncil.com/research_summary.asp. (Accessed 3 March 2010). Zraly, K. (2009). Windows on the world complete wine course. 25th Anniversary Edition. New York, NY: Sterling Publishing Co., Inc.
