An exploration of loyalty determinants in Greek wine varieties

University of Aarhus From the SelectedWorks of Polymeros Chrysochou 2009 An exploration of loyalty determinants in Greek wine varieties Athanasios Krystallis Polymeros Chrysochou Available at: https://works.bepress.com/chrysochou/23/

An exploration of loyalty determinants in Greek wine varieties Athanasios Krystallis a, * and Polymeros Chrysochou a MAPP, Aarhus School of Business, Aarhus University, Denmark a. MAPP, Aarhus School of Business, Aarhus University, Haslegaardsvej 10, DK-8210 Aarhus V, Denmark. Email addresses: A. Krystallis (atkr@asb.dk) and P. Chrysochou (polyc@asb.dk) * Corresponding author TRACK: 53 Wine marketing and branding in the global arena 1

An exploration of loyalty determinants in Greek wine varieties Abstract Purpose: This paper aims to provide a deeper understanding of the market structure of red and white wines from Greek wine varieties and measure loyalty behaviour of frequent wine buyers in Greece. Design/methodology: The study concerned measuring brand performance and loyalty of 4 different Greek wine varieties. Based on stated preference data, basic brand performance measures are estimated through Juster purchase probabilities of brand choice. To measure loyalty behaviour, the polarisation index φ (phi) is used as a measure to model loyalty both at the brand name and specific wine attributes and attribute-levels. Findings: The findings of the present study point to the conclusion that each one of the four Greek wine varieties under examination exhibits its own market structure and loyalty profile, whereas price, quality certification and winemaker s size seem to function as loyalty stimulators more effectively for white wines. Moreover, it is also clear that the origin or type of the wine variety per se does not constitute a particularly important loyalty component in the wines marketing mix. Research limitations/implications: The wine category has always been one of the most challenging product categories to investigate. Many attributes contribute to building loyalty that is often hard to delineate and take into account each. Moreover, the present methodology is based on stated preference data, whereas revealed preference data could be the ideal for applying the specific methodology. Originality/value: Few studies, if any, have explored the issue of loyalty using the present methodological approach in the case of wine. 2

1 Introduction Wine is often characterised as a difficult and confusing product for consumers. In addition, the wine market is much diverse and competitive, with an enormous number of niche brands, making consumer choice even more complex. When consumers are overloaded with information that exceeds their processing limits, they tend to simplify their decision-making based on heuristics (Bettman et al., 1998). In the case of wine, such heuristics can often be extrinsic quality cues (e.g. brand, price, variety, VQPRD certification, etc.). Certain past studies have explored the importance of such cues on consumers wine choice (e.g. Fotopoulos et al., 2003; Jarvis et al., 2007). However, less is known on whether consumers remain loyal within a specific wine category, and moreover what influences their loyalty behaviour mostly. Loyalty has mostly been viewed from the brand level. However, there is the possibility to account for loyalty broken down into product attributes, but such an approach has rarely been measured, analysed or discussed, apart from a few recent exceptions (Jarvis et al., 2007; Singh et al., 2008). Singh et al. (2008) suggested that measuring loyalty to product attributes is urgently needed, since manufacturers, retailers and marketers generally seem to have little empirical knowledge on this issue. Furthermore, a better understanding of which attributes are drivers of loyalty will provide further input for the design and implementation of loyalty strategies. On this basis, the aim of this paper is to provide a deeper understanding of the market structure of red and white wines from Greek wine varieties and measure loyalty behaviour of frequent wine buyers. Basic brand performance measures are estimated through Juster-scale purchase probabilities of brand choice. The Dirichlet-NBD model and the polarisation index φ (phi) are used to model loyalty both at the a) brand name; and b) wine attributes level. The wine attributes and their respective levels used for the purposes of this study are: a) price 3

(low, medium, high tiers), b) winemaker (small, medium, large corporate size), and c) quality certification (VQPRD wine or not). The paper develops as follows: in the Material and Method section a brief description of the concept of behavioural loyalty and its measurement, as well as a detailed description of the research design, the data collection procedure and the sample(s) are included; in the Results section the main empirical outcomes of the research explained in relation to its aims; finally, in the Discussion and Conclusion section the practical implications of the current research findings are highlighted. 2 Material and Method 2.1 Measuring loyalty Brand loyalty is nowadays a well established concept. Copeland (1923) first defined loyalty as an extreme attitude towards a product which might have a special effect on buyer behaviour. This definition is considered as one of the most cogent descriptions of loyalty ever proposed. It took 50 years of research until researchers suggested that loyalty is a composite measure, comprising behavioural and attitudinal loyalty (Day, 1969; Dick and Basu, 1994; Jacoby and Chestnut, 1978). Behavioural loyalty has been suggested to be measured at a product s brand level with the use of various parameters commonly known as Brand Performance Measures (BPMs). Typical BPMs are market share, penetration, and purchase frequency. Certain BPMs can be used as inputs in fitting the Dirichlet model (Fader and Schmittlein, 1993; Bhattacharya, 1997). Models like this are often called zero-order models due to their assumption that each purchase is unrelated to the previous one ( as if random ). The Dirichlet model has been shown to provide useful benchmarks and offer a natural baseline for the repeat-purchase loyalty each brand enjoys (Ehrenberg et al., 2004; Goodhardt et al., 1984). Empirical regularities 4

associated with the Dirichlet model have been widely tested and supported in marketing, and observed for over 30 years in a large number of product categories and across different countries in Europe, the USA, Asia and Australia (Ehrenberg et al., 2004; Uncles et al., 1995). Brand performance measures are usually obtained from revealed preference data (e.g. scanner data, panel data). However, such data can be expensive and time-consuming to acquire and process and may not be readily available for certain product categories and markets (Uncles and Lee, 2006). An alternative approach for empirically estimating similar measures is by using the Juster-scale, according to which respondents are asked to rate the probability of purchasing selected brands within a product category in a future time purchasing period ( What are the chances that you, personally, will buy/purchase <brand j> in the next <period>? ). Answers to the previous question are provided in an 11-point probability scale ranging from 0 to 10, where 0 denotes no chance, almost no chance (1 in 100) and 10 denotes certain, almost practically certain (99 in 100). The attraction of the Juster-scale is its use to develop observed estimators for a set of various brand performance measures which can further be used as inputs to fit the Dirichlet model (Wright et al., 2002). The Juster scale is preferred over other purchasing intention measures, since it has been proven to be more reliable and precise, constituting a direct means to estimate real purchasing behaviour (Wright and MacRae, 2007). A measure stemming from the Dirichlet model used to model loyalty is the polarisation index φ that was initially proposed by Sabavala and Morisson (1977). Polarisation φ is estimated by the following equation: φ = 1/(1+S), where S is a parameter of the Dirichlet model (see Ehrenberg, 1988). Both indices (φ and S) capture changes in heterogeneity of consumer choice as purchase incidence changes. The primary benefit of φ is that it is easier to interpret, since it varies from zero to one, whereas S varies from zero to infinity. Values of φ close to 5

zero indicate pure homogeneity in consumer choice, denoting high switching levels within a product category, where all buyers have the same propensity to buy individual brands. Values of φ close to one signify the existence of maximum heterogeneity, indicating high levels of loyalty in a product category within which each consumer buys only his/her favourite brand (Fader and Schmittlein, 1993; Stern and Hammond, 2004). 2.2 Study design For the purposes of the study, brand performance on a set of 40 wine brands was investigated using the Juster-scale. The period for measuring the probability of purchase was set to four weeks. Four Greek wine varieties were chosen (two white varieties: Asyrtiko and Mosxofilero ; and two red varieties: Agiorgitiko and Xinomavro ) and 10 known brand names were selected from each variety according to their retail sales and the attributes under examination (price, winemaker and certification), so to include a reasonable number of brands for each attribute. Thus, 12, 15 and 13 brands belonged to the low, medium and high price tiers respectively, 35 brands were VQPRD wines, and 12, 6 and 22 brands were produced by big, medium and large winemakers respectively. In order to simplify the selection of brands, no reference to vintage was made. Each brand was presented as pictures to the participants to further stimulate brand recognition and increase realism. In addition, for each variety brands were shown to participants randomly to avoid any order bias effects. Finally, apart from the section with the Juster-scale, the questionnaire included: a) Questions about wine purchasing frequency of respondents (replies varied from once a week to at least once a month ; b) A measurement of attitudinal loyalty towards Greek wine varieties (2 items per variety: I am committed to this <wine variety> and I am willing to pay a higher price for this <wine 6

variety> over other <red/white wine varieties>, adapted from Chaudhuri and Holbrook, 2001). Answers were measured on 7-point, Likert agreement scales with end-points 1 = totally disagree to 7 = totally agree ; c) A measurement of the importance of 10 wine purchasing criteria (e.g. How important are the following criteria for your wine purchasing selection? ). Answers were measured on a 7- point, Likert importance scale with end-points 1= completely unimportant to 7= absolutely important. The 10 criteria were: grape variety, price, winemaker s fame, vintage, prize won, quality certification, region of origin, brand name, label, and fermentation in barrel; and d) A part aiming to describe participants socio-demographic profile. 2.3 Sample A web-based administered survey was undertaken in Greece during March 2009. Participants were selected according to their purchase of wine (at least one purchase a month), in order to ensure a minimum level of respondents knowledge and involvement in wine purchasing. In total, 408 respondents participated in the survey, from which 205 answered questions referring to white varieties (N W ) and 203 to red varieties (N R ). Table 1 presents descriptive and group comparison tests between the two samples. Both white and red wine samples are skewed towards relatively younger age groups, higher educational levels and higher income tiers, although the participation of male and female consumers is almost equal between the two samples. Moreover, the majority of consumers in both samples are married, with one or two children, live in large cities across the country and work in paid types of employment. Regarding the statistically significant differences between the two samples, more red wine consumers belong to the tertiary educational level and live in smaller cities, while less are in a paid type of employment compared to white wine consumers. --- Insert Table 1 about here --- 7

The sample was further segmented to frequent and non-frequent buyers according to their stated frequency of buying (red or white) wine. Frequent buyers are those reporting buying wine at least once a week, whereas non-frequent those at least once a month. Frequent buyers of red wine were 61 (N RF) and of white wine 59 (N WF ). The socio-demographic profile of the frequent buyers and the statistically significant differences from the overall samples for white and red wines appear in Table 2. No statistically significant differences emerge between the two pairs of groups per wine category (white or red), apart from gender (i.e. more male consumers belonging to the frequent group) for both varieties and age groups for white wine buyers sample (more frequent buyers in the 25-34 and above 55 age groups). --- Insert Table 2 about here --- 2.4 Procedure Answers to the Juster-scale were used for the estimation of BPMs, such as penetration and purchase frequency at the brand level, as well as at the product category level, and market shares at the category level. Theoretical BPMs are relatively easy to estimate and are analytically described by Wright, Sharp and Sharp (2002). The above mentioned BPMs are those necessary to fit the Dirichlet model, from which additional measures of loyalty can be estimated. In addition, each brand was categorized based on the following attributes: a) price tier (low, medium, high); b) winemaker (small, medium, large corporate size); and c) certification (VQPRD wine or not) allowing analyses to be performed on an attribute level as well. Finally, the polarization index φ was estimated from the Dirichlet model allowing for modelling loyalty both at a brand and attribute level. All analyses were performed in MATLAB and the DIRICHLET software (Kearns, 2000). All analyses were conducted on the frequent buyer group. The main reasoning behind this decision was twofold. First, wine is by default a less frequent bought product thus exhibits 8

low loyalty levels. Therefore, from a technical point of view, it is not sure, and not aim of the present paper, whether this was due to low purchase frequencies or variety seeking behaviour. Thus, the frequent buyer group was selected instead. Second, from a marketing point frequent buyers comprise the main target segment for wine companies. Therefore, it is interesting to explore loyalty behaviour of such buyers. 3 Results In relation to the criteria for wine purchasing included in the questionnaire, the five most important ones are region of origin, variety, price, winemaker s fame and quality certification (Table 3). Brand name did not receive high score indicating its low importance in the choice of wine. No significant differences were found between different types of wine apart from vintage and label on which the red wine consumers consider as more important. --- Insert Table 3 about here --- For the frequent buyer group the total penetration (% of category buyers) is 89% for Asyrtiko, 92% for Moshofilero, 87% for Xinomavro and 89% for Agiorgitiko (Table 4). Furthermore, the average purchasing frequency during the 4-week period under examination is 7.1 times for Asyrtiko, 6.3 times for Moshofilero, 5.8 times for Xinomavro and 6.8 times for Agiorgitiko. In other words, more than 90% of the frequent wine buyers purchase one of the 20 white wine brands at least once a month, while a frequent wine buyer buys any of the 20 white wine brands around 7 times a month. The relevant figures for the red wines are penetration 88% at average and average purchasing frequency around 6 times a month (4 weeks). --- Insert Table 4 about here --- The third column of Table 4 indicates the polarisation indices φ for the four Greek wine varieties under investigation. It is clear that loyalty towards all varieties is very low, where as there are no differences between the two white and the two red wine varieties. The same low 9

loyalty to all the wine varieties under examination is exhibited by the rather neutral attitudinal loyalty measures shown in the last column of Table 4. Basic brand performance measures per wine variety for each wine attribute under examination are shown in Table 5. In terms of price, larger market shares appear in relation to the medium tier across all wine varieties, whereas comparably high market shares appear for the high price tier in the Asyrtiko variety (white wine) and the low price tier in the Xinomavro variety (red wine). Accordingly, higher penetrations and purchase frequencies appear in relation to the medium price tier across all wine varieties, but equally high measures appear for the high price tier in the white wine varieties and for the low price tier in the red wine varieties. --- Insert Table 5 about here --- In terms of certification, much larger market shares, penetrations and purchase frequencies emerge for the VQPRD wines than the non-certified ones, across all white or red varieties. Furthermore, in relation to the corporate size of the winemaker the results seem to be varietyspecific, since market share, penetration and purchase frequency are usually higher for the big winemakers in relation to Xinomavro (red) wine, where as the relevant measures for Moshofilero (white) and Agiorgitiko (red) wines are usually higher for small winemakers. For Asystiko (white) wines, both big and small winemakers seem to acquire equally large market shares and penetrations. However, the differences among varieties in market share, penetration and purchase frequency in relation to the size of the winemaker are not usually too large, especially in relation to the comparison between big and small winemakers. The fourth column of Table 5 indicates the polarisation indices φ for the wine varieties under investigation per attribute and attribute-level. It is clear that loyalty to price, as well as to all price tiers is very low, especially for red varieties. Moreover, loyalty to certification is considerably higher for VQPRD wines in white wines and especially the Asyrtiko variety. 10

Finally, loyalty to the size of the winemaker is also considerably higher for the white wine varieties; in particular, loyalty to big winemakers is high for Asyrtiko wines, and loyalty to small winemakers is high for Moshofilero wines. However, substantial loyalty appears in relation to small winemakers also for Agiorgitiko (red) wines. 4 Discussion Although the nature of the data collection procedure followed in the present survey does not allow for wide generalisations, the socio-demographic profile of the two samples indicates that younger, more educated, urban consumers constitute the most typical wine buyers. Moreover, it is worth noting that the main discriminating characteristic between white and red wine buyers is the educational level, with red wine being a more preferable choice of the more educated consumers than white wine. However, the general convergence of the socio-demographic profiles of both white and red wine buyers possibly indicates that the same consumers are purchasers of both wine categories, usually substituting the one for the other according to criteria that prevail under different contextual conditions (e.g. occasion a bottle of wine is purchased for, availability of wide range of wines in relevant outlets and convenience of accessing them, etc.). Also striking is the finding that the heavy frequency of wine purchasing does not seem to be depended on socio-demographic characteristics of the buyers. Again, more customised, personality-related characteristics (e.g. consumers attitudes and beliefs) or even purchasing context-specific factors as above might offer more valuable explanatory insights than sociodemographics as to the reasons that make a consumer to be a frequent wine buyer. The importance ranking of the wine selection criteria included in the present survey points out to the accuracy of the wine attributes selected as potential drivers of consumer loyalty to a wine brand, and offers face validity to the study. In this respect, it is shown that Greek 11

consumers select wines based on region of origin, wine variety, price and winemaker s fame. The fact that variety/region combinations are more important criteria than brand name constitutes both a reason for and an indication of the large fragmentation of the wine market at the brand level. This fragmentation justifies to a substantial extent the low loyalty levels exhibited in the present survey, as it will be discussed further in this section. One of the most interesting parts of the findings is what concerns consumer loyalty towards the Greek varieties under examination. Following the indications about large fragmentation of the wine market at the brand level, loyalty towards all varieties is found very low, with no differences between white and red wine varieties. This implies that the consumers of the survey are not fond of any (Greek) variety in particular, thus substitution of one wine brand from one variety (red or white) for another wine brand of another variety can usually happen, especially since the same consumer can equally well buy red or white wine, as shown before. Hence, the (Greek) variety of wine does not constitute an extrinsic cue that can create loyalty to the category. In this respect, the high market shares, penetrations and purchase frequencies found at the variety level do not seem to owe their existence to the origin or type of the varieties. Moreover, the neutral level of attitudinal loyalty to the varieties under examination further justifies the above finding and offers additional face validity to the research. Nevertheless, regarding in particular the comparison between attitudinal and behavioural loyalty, the high reliability coefficients of the attitudinal loyalty scales for each variety indicate that the specific loyalty measurements can function as a robust basis to compare polarisation-type of loyalties steaming from Juster estimates. On the other hand, the results regarding the VQPRD quality certification are more straightforward. Quality certification seems to constitute a very powerful attribute irrespective of wine category (white or red) or variety. However, results in this respect should be interpreted with caution, since the representation in the research design of non-certified wines, 12

or of wines with other types of quality certification (e.g. regional wines) is particularly small, meaning that the results are dominated by VQPRD wine brands across categories and varieties. The importance of the winemaker s corporate size is hard to grasp through the present findings. A certain trend that can be supported is the fact that the (consumer inferred) average corporate size of the winemaker is the less successful winemaker-related characteristic of a wine product across categories and varieties. This might again relate to a perception trade-off consumers engage in during the wine selection process, since both small- and large winemaker size might be perceived as a positive heuristic that is associated with a certain expertise or fame on behalf of the wine producer (e.g. big size might imply a successful, experienced producer with accumulated market expertise; and small size might indicate technical/oenological expertise, specialisation in specific production methods and low volume-high quality market orientation). In general, however, it needs to become clear that the results in respect to winemakers corporate size are proven to be variant and too much variety-depended. It is also possible that the winemaker s fame, which is also found to be among the most important self-reported wine selection criteria as explained above, is not represented well by the winemakers corporate size (in other words, the mental distance in consumers associations between corporate size and market value is too large). Finally, the underrepresentation of medium-sized winemakers can also be a limitation that changes brand performance measures in favour of the small- and (mainly) large-sized winemakers. Focusing on the price attribute, the preference for usually average price tiers across all wine varieties analysed may indicate a type of value-cost trade-off resulting to low and high prices aversion; low prices may indicate lack of product quality in the eyes of the customers leading to a high-risk buying decision; and high-price aversion may be the outcome of the widely held perception among Greek consumers that wine is often a low value-for-money product. After 13

all, price is found to be one of the most important self-reported criteria of wine selection, as described above. Nevertheless, the present results in relation to price imply a certain positioning of the Greek white wine brands at a more premium market level compared to the Greek red wine brands examined here (i.e. higher price tiers in Asyrtiko white wine acquire higher market share, penetration and purchase frequency, where as for Xinomauro red wines the case is the opposite). It thus seems that although medium prices are usually preferred, the importance of price is overall category-depended, closely mirroring value-cost considerations. In terms of loyalty, price is an important factor primarily for the white wine category; additionally, not only the high-price tier in Asyrtiko and Moshofilero wines, but also the lowprice tier in the former and the medium-price tier in the latter can also stimulate consumer loyalty. As regards red wines, the medium-price tier creates more loyalty, especially in the Xinomauro variety, which is in accordance with the above-described findings about mediumpriced red wines usually achieving higher market shares, penetrations and purchasing frequencies. All in all, although there is a certain variation at the variety level, price considerations follow the red-white wine category differentiation, with red wines being better positioned at medium price levels and white wines at both medium and high price levels. 5 Conclusions Wine marketers have always been eager in building strong bonds with their customers. Therefore, identifying which wine attributes make consumers loyal and increase their chances of repeat purchase is always of high importance. Moreover, delineating the issue of loyalty to a wine brand and to the cues (i.e. extrinsic attributes) that are found on the label (e.g. variety, winemaker, VQPRD sign, etc.) can further help wine practitioners to determine which product attributes stimulate loyalty and consequently create tailor-made products that fit the market better and increase market shares. In this respect, the findings of the present study 14

point to the conclusion that each one of the four Greek wine varieties under examination exhibits its own market structure and loyalty profile, whereas price, quality certification and winemaker s size seem to function as loyalty stimulators more effectively for white wines. Moreover, it is also clear that the origin or type of the wine variety per se does not constitute a particularly important loyalty component in the wines marketing mix. 15

References Bettman, J.R., Luce, M.F. and Payne, J.W. (1998), Constructive consumer choice processes, Journal of Consumer Research, Vol. 25, No. 3, pp. 187-217. Bhattacharya, C. (1997), Is your brand s loyalty too much, too little, or just right? Explaining deviations in loyalty from the Dirichlet norm, International Journal of Research in Marketing, Vol. 14, No. 5, pp. 421-435. Chaudhuri, A. and Holbrook, M.B. (2001), The chain of effects from brand trust and brand affect to brand performance: The role of brand loyalty, Journal of Marketing, Vol. 65, No. 2, pp. 81-93. Copeland, M.T. (1923), Relation of consumer s buying habits to marketing methods, Harvard Business Review, Vol. 1, No. 2, pp. 282-289. Day, G.S. (1969), A two-dimensional concept of brand loyalty, Journal of Advertising Research, Vol. 9, pp. 29-35. Dick, A.S. and Basu, K. (1994), Customer loyalty: Toward an integrated conceptual framework, Journal of the Academy of Marketing Science, Vol. 22, No. 2, pp. 99-113. Ehrenberg, A.S.C. (1988), Repeat Buying Facts, Theory and Applications, 2 nd Ed., Oxford University Press, New York. Ehrenberg, A.S.C., Uncles, M.D. and Goodhardt G.J. (2004), Understanding brand performance measures: Using Dirichlet benchmarks, Journal of Business Research, Vol. 57, No. 12, pp. 1307-1325. Fader, P. and Schmittlein, D. (1993), Excess behavioural loyalty for high-share brands: Deviations from the Dirichlet model for repeat purchasing, Journal of Marketing Research, Vol. 30, No. 4, pp. 478-493. Fotopoulos, C., Krystallis, A. and Ness, M. (2003), Wine produced by organic grapes in Greece: using means--end chains analysis to reveal organic buyers purchasing motives in comparison to the non-buyers, Food Quality and Preference, Vol. 14, No. 7, pp. 549-566. 16

Goodhardt, G., Ehrenberg, A.S.C. and Chatfield, C. (1984), The Dirichlet: A comprehensive model of buying behaviour, Journal of the Royal Statistical Society, Vol. 14, No. 7, pp. 621-655. Jacoby, J. and Chestnut, R.W. (1978), Brand Loyalty Measurement and Management, Wiley, New York. Jarvis, W., Rungie, C. and Lockshin, L. (2007), Revealed preference analysis of red wine attributes using polarization, International Journal of Wine Business Research, Vol. 19, No. 2, pp. 127-138. Kearns, Z., (2000), Dirichlet No Solver Software, Massey University, New Zealand. Sabavala, D. and Morrison, D. (1977), A model of TV show loyalty, Journal of Advertising Research, Vol. 17, No. 6, pp. 35-43. Singh, J., Ehrenberg, A.S.C. and Goodhardt, G. (2008), Measuring customer loyalty to product variants, International Journal of Market Research, Vol. 50, No. 4, pp. 513-532. Stern, P. and Hammond, K. (2004), The relationship between consumer loyalty and purchase incidence, Marketing Letters, Vol. 15, No. 1, pp. 5-19. Uncles M.D., Ehrenberg, A.S.C. and Hammond K. (1995), Patterns of buyer behavior: Regularities, models and extensions, Marketing Science, Vol. 14, No. 3, pp. G71-G78. Uncles, M.D. and Lee, D. (2006), Brand purchasing by older consumers: An investigation using the Juster scale and the Dirichlet model, Marketing Letters, Vol. 17, No. 1, pp. 17-29. Wright, M. and MacRae, M. (2007), Bias and variability in purchase intention scales, Journal of the Academy of Marketing Science, Vol. 35, No. 4, pp. 617-624. Wright, M., Sharp, A. and Sharp, B. (2002), Market statistics for the Dirichlet model: Using the Juster scale to replace panel data, International Journal of Research in Marketing, Vol. 19, No. 1, pp. 81-90. 17

Tables Table 1. Socio-demographics of wine type segments White wine (N 1 =205) Red wine (N 2 =203) Gender Male 0.47 0.51 Female 0.53 0.49 Marital status Married 0.52 0.47 Single 0.40 0.40 Other 0.08 0.13 Live in a place of Less than 10.000 inhabit. 0.02 0.03 10.000 100.000 inhabit. 0.11 0.21 More than 100.000 inhabit. 0.87 0.75 Education Elementary school or less 0.05 0.01 Technical or high school 0.47 0.26 University or higher 0.53 0.73 Household size 1 0.24 0.16 2 0.27 0.24 3 0.16 0.24 >3 0.33 0.36 Age groups 24 or less 0.06 0.12 25-34 0.38 0.37 35-44 0.39 0.29 45-54 0.12 0.15 55 or more 0.04 0.06 My net monthly income is... than 1.000 Less 0.16 0.15 Almost equal 0.42 0.35 More 0.42 0.50 Occupation In paid work 0.69 0.59 Self-employed 0.24 0.24 Other 0.07 0.17 I buy wine More than once a week 0.12 0.06 Once a week 0.17 0.24 2-3 times a month 0.32 0.35 Once a month 0.40 0.35 χ 2 p-value 0.79 0.428 2.37 0.305 9.00 0.011 18.98 0.000 7.41 0.060 8.33 0.080 3.05 0.217 9.85 0.070 8.19 0.042 18

Table 2 Socio-demographics of the frequent buyer group per wine type White wine χ 2 Red wine p-value (N WF =59) (N RF =61 Gender Male 0.59 0.69 5.19 0.030 Female 0.41 0.31 Marital status Married 0.47 0.43 Single 0.39 3.08 0.214 0.44 Other 0.14 0.13 Live in a place of Less than 10.000 inhabit. 0.02 0.02 10.000 100.000 inhabit. 0.07 1.60 0.448 0.21 More than 100.000 inhabit. 0.91 0.77 Education Elementary school or less 0.00 0.00 Technical or high school 0.44 0.71 0.703 0.25 University or higher 0.55 0.75 Household size 1 0.24 0.21 2 0.27 0.23 1.40 0.705 3 0.20 0.30 >3 0.29 0.26 Age groups 24 or less 0.07 0.10 25-34 0.44 0.38 35-44 0.24 21.51 0.000 0.26 45-54 0.12 0.18 55 or more 0.14 0.08 My net monthly income is... than Less 0.22 0.16 Almost equal 0.32 3.98 0.137 0.38 More 0.46 0.46 Occupation In paid work 0.58 0.69 Self-employed 0.34 5.11 0.078 0.18 Other 0.08 0.13 Note: Chi-squares report differences from the non-frequent buyer groups χ 2 p- value 10.84 0.001 0.82 0.662 0.86 0.650 1.01 0.603 5.26 0.154 1.88 0.757 066 0.718 3.76 0.152 19

Table 3 Importance of criteria in wine purchasing for whole sample and wine type segments All sample (N = 408) White wine (N 1 =205) Red wine (N 2 =203) F (p-value)* Region of origin 5.31 5.26 5.36 0.56 (0.456) Variety 5.30 5.18 5.42 3.23 (0.073) Price 5.18 5.22 5.15 0.22 (0.641) Winemaker s fame 5.12 5.11 5.13 0.01 (0.923) Quality certification 5.04 5.08 5.00 0.24 (0.622) Brand name 4.92 4.92 4.92 0.00 (0.996) Vintage 4.84 4.66 5.02 5.94 (0.015) Barrel fermentation 4.84 4.74 4.94 1.52 (0.219) Label 4.31 4.02 4.59 11.48 (0.001) Medal 4.13 4.03 4.23 1.44 (0.231) * Differences between two groups of wine segments. 20

Table 4 Brand performance measures and polarisation indices for Greek varieties, frequent buyers Penetration Purchase Frequency φ Mean loyalty (attitudinal) 1,2 White wine Asyrtiko 0.89 7.1 0.05 3.23 Moshofilero 0.92 6.3 0.05 3.85 Red wine Xinomavro 0.87 5.8 0.04 3.26 Agiorgitiko 0.89 6.8 0.04 3.50 1. Likert agreement scales with end-points 1 = totally disagree to 7 = totally agree 2. Cronbach alpha coefficients: Asyrtiko = 0.878; Moshofilero = 0.831; Xinomavro = 0.859; and Agiorgitiko = 0.781. 21

Table 5 Brand performance measures and polarisation indices for wine attributes, frequent buyers Attribute Market share Penetration Purchase Frequency φ As M X Ag As M X Ag As M X Ag As M X Ag Price 0.06 0.05 0.03 0.03 Low 0.24 0.15 0.35 0.26 0.59 0.51 0.67 0.63 2.64 1.66 2.61 2.46 0.09 0.02 0.02 0.02 Medium 0.27 0.44 0.34 0.38 0.64 0.77 0.63 0.71 2.68 3.30 2.71 3.23 0.04 0.08 0.08 0.05 High 0.29 0.23 0.15 0.20 0.64 0.59 0.44 0.57 2.88 2.24 1.73 2.17 0.08 0.07 0.02 0.02 Certification 0.12 0.06-0.02 VQPRD 0.69 0.67 -* 0.73 0.82 0.86-0.85 5.30 4.51-5.17 0.21 0.12-0.02 No VQPRD 0.12 0.14-0.11 0.41 0.52-0.41 1.79 1.60-1.61 0.03 0.02-0.02 Winemaker 0.03 0.12 0.02 0.04 Big 0.32 0.36 0.59 0.21 0.61 0.72 0.79 0.56 3.37 2.90 3.82 2.31 0.22 0.08 0.02 0.05 Medium 0.16 - - 0.29 0.50 - - 0.65 1.97 - - 2.67 0.02 - - 0.02 Small 0.32 0.52 0.24 0.34 0.69 0.79 0.58 0.66 2.97 3.82 2.09 3.07 0.02 0.20 0.02 0.10 Note: Key to wine varieties: As=Asyrtiko, M=Moshofilero, X=Xinomavro, Ag=Agiorgitiko * All brands in the Xinomavro variety are VQPRD wines 22

Contact Details /Biographical Notes Athanasios Krystallis is Associate Professor of Marketing at the Department of Marketing and Statistics, Aarhus School of Business, and co-responsible for consumer research at the Centre for Research on Customer Relations in the Food Sector (MAPP). Athanasios has an MSc degree in food marketing and a PhD in consumer behaviour (both from the University of Newcastle upon Tyne, UK). His scientific interests emphasize on consumer behavior towards extrinsic food quality cues (focusing on quality logos and health claims, such as organic, PDO, genetically modified, functional and light) and the implications on marketing strategy of the way extrinsic cues are perceived by customers. His scientific interests recently focus also on branding issues and the determinants of consumer loyalty. His inclusive work in the area of food consumer behavior comprises 35 peer reviewed papers since 2001, in Journals such as European Journal of Marketing, Agribusiness, Food Quality and Preference, Appetite, Risk Analysis, Journal of Risk Research, Health Risk and Society, British Food Journal, Sociologia Ruralis, Journal of Consumer Behavior, etc. His publications also include 10 books and book chapters, and numerous papers in peer-reviewed conference proceedings. A. Krystallis can be contacted at: atkr@asb.dk. Polymeros Chrysochou is a Doctoral Researcher at the Centre for Research on Customer Relations in the Food Sector (MAPP), Department of Marketing and Statistics, Aarhus School of Business, Denmark. His academic research focuses in the area of branding and consumer behaviour in food. He holds an MBA in Agribusiness Management from Agricultural University of Athens. Prior to joining Aarhus School of Business, he participated in several European funded FP projects such as TRACE, TYPIC, SafeFOODS and CONDOR. P. Chrysochou can be contacted at: polyc@asb.dk. 23