P A L E R M O 2010
w
ww
ww
w .. vv dd qq ss .. nn ee tt
A Better Understanding of the Structure of a Wine Market
Using the Attribute of Variety
Justin COHEN, Larry LOCKSHIN
Ecole Supérieure d'Agriculture d'Angers, Laboratory GRAPPE;
U. South Australia, Ehrenberg-Bass Institute for Marketing Science, School of Marketing
Introduction
This research explores buying patterns in the Australian retail wine market, a complex and potentially
highly fragmented product category. This study utilises revealed preference data obtained over a threeyear period (October 2000-September 2003) from members of a loyalty program from an Australia-wide
high-end retail chain. It addresses the question of whether product attributes such as variety, region of
origin, country of origin, Australian state, brand and $2 price tiers, rather than brand alone, better
describe the structure of this market. The stock-keeping units (SKUs) as defined by Fader and Hardie
(1996), which number in the thousands, form the basis of this analysis.
In the Australian wine market there are numerous varieties that are intended to cater to various palates
and gastronomic pairings. These together with the wide range of price and the multiple regions where
wine is produced combine to contribute to the great diversity and possible fragmentation of this
market. Most of the research that exists in the repeat purchase behaviour stream focuses on brand
buying. The true extension of this research is its exploration of attributes in an attempt to answer the
question as to whether attributes behave in the same manner as brands and whether their buying
propensities are moderated by size? Some of the methods used in this research have already been
applied to the Australian retail wine market (Romaniuk and Dawes 2005; Jarvis 2006; Murphy 2006;
Jarvis, Rungie et al. 2007). However, these methods have never been used comprehensively to describe
the product category of wine in its totality. This research provides a comprehensive review of marketing
metrics applied to one set of data, extended to attributes, in a potentially highly varied category.
This paper is part of a larger research project conducted to satisfy the requirements of a doctoral thesis
by the primary author of this work utilising all 5 methods to evaluate the 6 attributes being studied. This
larger body of work evaluates the enhanced value of exploring attributes rather than brands; compares
the value and applicability of the various methods; and compares the findings from polarisation with the
most recently described study of Jarvis (2006) using different methods of data handling and
aggregation. The attributes are evaluated using ‘observed’ performance measures, which are compared
with their ‘theoretical’ Dirichlet estimates. In addition, three methods arising from the Dirichlet, namely
the ‘double jeopardy’ phenomenon, polarisation and the duplication of purchase law are utilised to
stratify the market in a managerially useful form. The more enhanced value of exploring the Australian
wine market based upon these attributes rather than by brand alone has been clearly demonstrated.
In the industry description of the Australian wine market, there is adherence to the conventional
separation into red and white wine categories (Euromonitor International 2004). Information provided
for variety highlights only the major players and gives little attention to the smaller varieties
(Euromonitor International 2004). Prior work by Jarvis (2006) followed the accepted but not proven
notion that white and red wine are separate categories. In this research, a decision was made to look at
wine holistically and determine whether this market segmentation was warranted. It was decided that
the attribute of variety would include not only red and white wines, but also sparkling and fortified
wines, since a contribution to the wine market would be to properly understand how each of these
classifications contributes to the market as a whole.
Variety is highlighted as the attribute to be investigated in this research. Rasmussen (2001) showed that
in many countries, including Australia, the variety of grape is an important attribute of wine. Resnick
(2008), in her discussion of the place of wine and food in traditional and emerging markets, commented
on the replacement of regional branding by varietal choice currently favoured in the ‘new world’.
Comparison of the wine attributes in this research, using polarisation, has shown that variety has the
highest levels of loyalty compared to the other attributes studied. This paper will show the greatly
enhanced value of analysing variety in the description of the Australian retail wine market and will
dispel the myth of a separate white and red wine category that is prevalent in the industry literature
and in recent research (Jarvis 2006).
Data and Analysis
Revealed preference data from a loyalty program of one of Australia’s largest wine retailers, spaning a
three-year period from October 1, 2000 through September 30, 2003, provides the information for this
research. Purchase records of a sample (n=4,768) of wine buyers are analysed and the middle year of
results is presented. Each buyer had to have made a purchase at least once in each of the three years to
be included. For the attribute of variety, there were 28 levels, which included 15,944 SKUs
encompassing the categories of white, red, sparkling and fortified wines. There were a total of 564,078
transaction lines in the three-year period. There were 746,469 units of wine sold during the time period
with an average of 1.3 units per purchase. 86% of wine sold is in bottles. 11% of wine sold is soft pack
(bag in box). All data was coded, manipulated and prepared for analysis using SAS. The data was pretested using Excel in order to calculate basic measures for comparison. Automated programs were
developed through the use of Matlab and utilised for the analysis conducted for this research.
Methodology
The attribute of variety is isolated and 4 different outputs are presented to demonstrate what can be
learnt about a product category taking an attribute-based approach. The ‘observed’ performance
measures are calculated and tabulated. The phenomenon of ‘double jeopardy’ (Chatfield and Goodhardt
1975; Goodhardt, Ehrenberg et al. 1984; Ehrenberg, Goodhardt et al. 1990; Fader and Schmittlein 1993;
Jarvis 2006; Li, Habel et al. 2006) as well as the method of polarisation (Sabavala and Morrison 1977;
Fader and Schmittlein 1993; Rungie 2000; Rungie, Stern et al. 2003; Rungie, Brown et al. 2005; Rungie
and Laurent 2006) are utilised to identify the existence of ‘excess loyalty’(Fader and Schmittlein 1993),
‘niche’ and ‘change of pace’ behaviour (Kahn, Kalwani et al. 1988). The duplication of purchase law
(Goodhardt, Ehrenberg et al. 1984; Ehrenberg 1988; Ehrenberg, Uncles et al. 2004; Mansfield 2004) is
presented to illustrate repertoire buying and the level of sharing amongst varieties. The Dirichlet model
(Chatfield and Goodhardt 1975; Bass, Jeuland et al. 1976; Goodhardt, Ehrenberg et al. 1984; Uncles,
Ehrenberg et al. 1995; Ehrenberg and Uncles 1999), which is described as a comprehensive model of
buyer behaviour is central to the understanding of ‘double jeopardy’, polarisation and duplication of
purchase. Table 1 below contains the equations for the ‘observed’ performance measures (Ehrenberg
1988; Ehrenberg, Uncles et al. 2004; Rungie and Goodhardt 2004).
Table 1: How to Calculate ‘Observed Performance Measures
Market Share = Total Purchases of the Brand
Total Purchases of the Category
Penetration = The Number of Buyers of the Brand
The Total Number of Shoppers in the Category
Purchase Frequency = Average Number of Units of the Brand Purchased By All
Buyers in a Given Time Period
Sole (100%) Loyals = Percentage of Buyers Buying Only That Brand
Share of Category Requirements (SCR) = Purchase Frequency of Brand
Category Purchase Rate for Buyers of Brand
(In this research ‘attribute levels’ are substituted for brands)
www.vdqs.net/2010Palermo
2/10
The results obtained using the above five performance measures give a descriptive picture of the actual
behaviour of the attribute of variety and its levels. This forms the basis of comparison to ‘theoretical’
NBD-Dirichlet estimates (Goodhardt, Ehrenberg et al. 1984) subsequently used to formulate ‘double
jeopardy’ and deviations from it.
This paper will represent the ‘double jeopardy’ phenomenon in a graphical fashion as described by
Habel (2007) who elaborated on how it is possible to draw a ‘double jeopardy’ line (curvilinear) based
on the estimated Dirichlet parameters K, A, S and the brand alphas. The points represent the plotting of
the observed cross-points of penetration and purchase frequency and the purple curvilinear line is the
‘theoretical’ double jeopardy line as estimated by the Dirichlet model (Rungie 2003). The % differences
for each pair of ‘observed’ versus ‘theoretical’ purchase frequencies provide a more accurate
classification of the exceptions to the ‘double jeopardy’ phenomenon (> +/- 20% are classified as
noteworthy).
‘Excess loyalty’ (Fader and Schmittlein 1993) is exhibited if a high penetration level has markedly greater
purchase frequency than would be expected on the basis of the Dirichlet model. ‘Niche’ behaviour
(Kahn, Kalwani et al. 1988) is demonstrated if a low penetration level has a markedly higher purchase
frequency than would be expected on the basis of the Dirichlet model. ‘Change of pace’ behaviour is
displayed when there is a markedly lower purchase frequency than would be expected on the basis of
the Dirichlet model (Kahn, Kalwani et al. 1988; Fader and Schmittlein 1993).
The method of polarisation (Rungie and Goodhardt 2004; Rungie 2005; Jarvis 2006; Li, Habel et al. 2006)
provides a means to remove the effect of market share and also stratifies the market into the categories
‘borrowed’ from ‘double jeopardy’ namely ‘excess loyalty, ‘niche’ and ‘change of pace’(Jarvis, Rungie et
al. 2003; Jarvis 2006; Jarvis, Rungie et al. 2006; Jarvis, Rungie et al. 2007; Jarvis, Rungie et al. 2007). The
DMD ϕc provides a baseline loyalty level for an attribute and allows for comparisons across attributes.
The BBD ϕ shows variation from the DMD ϕc baseline and allows for comparison between attribute
levels. It is calculated by comparing each level of the attribute with all the other levels of the attribute
grouped together.
Casini et. al. (2008) provides a manual for the interpretation of the graphical representation of market
share and polarisation, both DMD ϕc and BBD ϕ. A scatter plot is drawn for each attribute by plotting
polarisation (DMD ϕc and BBD ϕ) against market share. A horizontal line is drawn through the DMD ϕc. A
vertical line is drawn through the midpoint of the range of market shares. This divides the scatter plot
into four quadrants. Attribute levels in the upper right hand quadrant will be considered to exhibit
‘excess loyalty’. Attribute levels in the upper left hand quadrant will be considered to demonstrate
‘niche’ behaviour. Attribute levels in both lower quadrants will be considered to display possible ‘change
of pace’ tendencies (> +/- 20% are classified as noteworthy).
The duplication of purchase law is an empirical generalisation that uses the patterns of repeat purchase
to identify the competitive structure of a market (Goodhardt, Ehrenberg et al. 1984). It suggests that
penetration levels moderate the amount of sharing between competing brands (Ehrenberg 1988;
Ehrenberg, Uncles et al. 2004; Mansfield 2004). The aberrations, which occur when brands (or attribute
levels) compete more or less than expected, are termed partitions (Wright, Sharp et al. 1998).
Duplication of purchase provides a means to look at ‘consumer-buying based’ patterns rather than
market-based patterns (Sharp 2009; Ehrenberg, Uncles et al. 2000). A duplication of purchase
(deviations) table is constructed by calculating the percent deviations of the observed from the
expected duplications in the duplication table (using the D coefficient)(Ehrenberg and Goodhardt 1970;
Murphy 2006). Deviations of greater than or equal to 10% are highlighted in green, whereas deviations
of greater than or equal to -10% are highlighted in red. Deviations of greater than or equal to +/- 20%
are considered noteworthy and if they are reciprocal are deemed to be a partition.
www.vdqs.net/2010Palermo
3/10
Results
Table 2 below (ordered by decreasing penetration) contains the attribute level results for the ‘observed’
performance measures of market share, penetration (b), purchase frequency (w), sole (100%) loyals and
share of category requirements (SCR).
Table 2: Variety Year 2 ‘Observed’ Performance Measures
‘All Other White Wine’ and ‘All Other Red wine’ are aggregated levels, which were not given further classification by the
provider of the data, and do not warrant further individual comment.
Noteworthy findings:
• 4 varieties (Chardonnay, Sparkling White, Shiraz and Cabernet Sauvignon) account for nearly 40%
of the market and are each bought by more than half of the buying population.
• Cask wine (white and red) accounts for 12% of the market (white 7% & red 5%). White cask wine
has a penetration of 20%, purchase frequency of 8.2, sole loyals of 3% and SCR of 22%. Red cask
wine has a penetration of 17%, a purchase frequency of 6.7, sole loyals of 2% and SCR of 19%.
• Lambrusco has a purchase frequency of 3.0 and an SCR of 8% relative to its penetration of 2%.
www.vdqs.net/2010Palermo
4/10
• The following varieties, not already mentioned with market shares below 5%, display possible
‘change of pace’ tendencies: Sparkling Red, Cabernet Sauvignon, Pinot Noir, Spumante, Merlot,
Verdehlo, Shiraz Blends, Cabernet Blends, Muscat, Semillon, Dessert, Grenache, Chardonnay
Blends and Tokay.
The Duplication of Purchase law generally appears to hold. Table 3 below (ordered by decreasing
penetration) contains the results of the deviations from the duplication of purchase law. Deviations of
10% or more in the positive direction are shaded green. Deviations of 10% or more in the negative
direction are shaded red.
Table 3: Variety Year 2 Duplication of Purchase (Deviations)
Figure 1 below contains the graphical representation of the ‘double jeopardy’ phenomenon.
Figure 1: Variety Year 2 ‘Double Jeopardy’
The ‘double jeopardy’ phenomenon generally appears to hold.
Noteworthy exceptions (> +/- 20%) to the ‘double jeopardy’ phenomenon through the use of %
difference:
• Chardonnay (30%) exhibits ‘excess loyalty’.
• White (161%) and Red (119%) Cask wine demonstrate ‘niche’ behaviour.
• Cabernet Blends (-26%), Shiraz Blends (-21%), Semillon (-25%), Sparkling Red (-27%), Champagne (33%), Grenache (-33%), Dessert (-45%), Chardonnay Blends (-48%), Muscat (-48%), Spumante (38%) and Tokay (-55%) suggest possible ‘change of pace’ behaviour.
Figure 2 below depicts the scatter-plot of the BBD ϕ and DMD ϕc values relative to market share.
Figure 2: Variety Year 2 BBD ϕ / DMD ϕc
‘All Other White Wine’ and ‘All Other Red wine’ are aggregated levels, which were not given further
classification by the provider of the data, and do not warrant further individual comment.
Noteworthy deviations (greater than or equal to +/-20%):
• There appears to be one wine market with no evidence of a separate red and white wine
category.
• There is significantly higher than expected sharing between White and Red Cask Wine and this
sharing is reciprocal. This suggests the possibility of a partition.
• There is significantly higher than expected sharing (suggesting the possibility of partitions)
between the following:
• Grenache with Cabernet Blends, Shiraz Blends and Merlot.
• Chardonnay Blends with Verdehlo.
• Tokay with Port and Muscat.
• There is significantly less than expected sharing between the following:
• White Cask Wine with Shiraz/Hermitage and Cabernet Sauvignon.
• Muscat and Tokay with Chardonnay.
• Spumante with Shiraz/Hermitage.
Noteworthy findings:
• Chardonnay and Sparkling White exhibit ‘excess loyalty’.
• White and Red Cask wine as well as Lambrusco and Sherry demonstrate ‘niche’ behaviour.
www.vdqs.net/2010Palermo
5/10
www.vdqs.net/2010Palermo
6/10
Discussion
Twenty-six individual varieties were evaluated with the addition of two aggregated levels, namely ‘all
other white wine’ and ‘all other red wine’, not given further classification by the provider of the data.
Exploring the ‘observed’ performance measures, it is seen that 4 varieties namely Chardonnay, Sparkling
White, Shiraz/Hermitage and Cabernet Sauvignon are the market leaders and account for 40% of the
market. If White and Red cask wine (12% of the market) and blended red wine (10% of the market) are
included, these together account for 2/3 of the wine market. There are a lot of varieties to choose from,
but half of them have less than 2% market share and occupy a very small position in the market. The 4
major varieties mentioned are each bought by more than half of the buying population. There is a large
variation in penetration with a consequently big difference in the numbers of consumers buying each of
these wines. For example, half of the wine buying population in Australia buys Chardonnay, but only 8%
buy Grenache giving it a much smaller customer base.
Chardonnay exhibits ‘excess loyalty’ which correlates with the fact that it had a high purchase frequency
and the greatest market share and may be benefiting from the ‘triple jeopardy’ effect (Fader and
Schmittlein 1993). It could also have a flavour profile most conducive to the Australian palate or it may
just go better with the Australian diet. Alternatively, it may be easier to drink a white than a red wine or
because of the hot Australian climate more white wine than red wine is consumed. Sparkling White also
has a higher level of loyalty compared to the category, exhibiting ‘excess loyalty’. It has a relatively high
position in the market and has increased sole loyals because it is a different type of product than still
wine and has something unique about it (with bubbles). A small group of consumers may prefer this
style or be purchasing it for different reasons such as celebrations or gift giving.
White and Red Cask wine demonstrate ‘niche’ behaviour because they are potentially functionally
different even though they are classified as a variety, or because they are packaged (size and container)
or priced differently. It may simply be that a certain class of people drink wine as a commodity with
their meals without a need for more complex or expensive bottled wine or because it is used for
cooking. Even though cask wine accounts for about 45% of volume sales in Australia, this data shows
that Vintage Cellars, by catering to a more involved consumer, has a disproportionately lower share of
the market for cask wine. However, looking at this differently, perhaps the real story is that these more
involved wine buyers are still buying cask wine and it represents a significant portion of sales in this
retail chain and specifically more than the artisanal varietals that one might expect these more involved
consumers to purchase.
There is also greater loyalty to Lambrusco and Sherry than the category level and this is described as
‘niche’ behaviour due to their smaller market shares. Lambrusco is a different product, being a sparkling
wine from Italy. Most Australians do not buy Lambrusco, but Australia has a very large immigrant
community from Italy and the panellists who buy Lambrusco could be of Italian descent. Sherry is a
different product because it is often used for cooking or drunk as an aperitif. Champagne has a higher
than expected level of loyalty because there is something different about the product and is universally
recognised as the choice for celebrations. Spumante is also a sparkling Italian wine and may be
purchased for the same reasons as stated above for Lambrusco.
‘Change of pace’ behaviour occurs for 19 of the 26 varieties with market shares less than 5% because
after the top 4 to 6 varieties there is a conglomerate of smaller varieties. For example, Verdehlo may
only be available in 6 possible choices while there may be dozens of different Chardonnay bottles on the
shelf. These ‘change of pace’ varieties are all very different from one another in style and are not bought
for regular consumption. Over a year, people buy different wines for many different reasons such as
cooking, drinking, gifts, holidays and even possibly related to the climate at the particular time of year.
This category includes some of the blended wines like Cabernet and Shiraz blends. All small varieties
have a chance of being purchased by these consumers. Marketers need to determine who the buyers
are and specifically target them, and they need to realise that small varieties like Grenache may be
change of pace, and not worth targeting. Compared to the other attributes using DMD ϕ c , the loyalty to
variety is the highest.
www.vdqs.net/2010Palermo
7/10
There is one wine market with no evidence of a separate white and red wine partition. Exploring the
deviations from duplication of purchase shows a significant white and red cask wine partition due to
their significantly higher sharing which is reciprocal and the fact that they also share less with most large
varieties. As previously stated, Cask wine drinkers may be buying strictly for the cheaper cost per unit
and that may be a sufficient reason for a partition or it is simply a functional difference. There is
significantly higher sharing between Tokay and Port and Muscat because they are all functionally similar
to each other and different from still or sparkling wines. There is also significantly higher sharing
between Chardonnay blends and Verdehlo because Verdehlo is often blended into Chardonnay blends.
Finally, there is significantly higher sharing between Grenache and Cabernet and Shiraz blends and
Merlot because Grenache is also quite heavily blended with other red varietals. This sharing may fall
short of the computational definition of a partition, but is nevertheless important.
When one considers each one of these examples, one can attribute the higher sharing to similarities in
flavour or style eg. tannic, acidic and sweet, or functionality eg. dessert. There is a proclivity to buy
similar things so white wines tend to share together and red wines tend to share together but not
enough to be considered partitions. There is also the possibility that some higher than expected sharing
occurs because varieties are produced in the same region or due to price-based factors. There is less
than expected sharing occurring because a specific variety such as Sparkling White is functionally
different to all the others. The same fact explains the reciprocally lesser sharing between White and Red
Cask wine and the other big players. It is important to note that the fact that bottled wine consumers
drink less cask wine does not mean that they do not drink cask wine. There is one wine market with no
evidence of a separate red and white wine category as predetermined by Jarvis (2006) in his thesis.
Limitations
The limitations of this research are related to the data. Typically research in this field requires true panel
data, which maintains a sample of buyers and non-buyers. Due to the fact that this data was obtained
from a specialty wine and spirits retailer and that this was a loyalty program, the existence of nonbuyers was nil. Although the sample of buyers obtained does cover a representative sample across
Australia, these wine buyers could be described as more involved consumers because of the type of
wine store they frequent. Comparison of these findings to general industry data does validate that there
are large differences between the structure of the market in this retail chain and that of Australia as a
whole. These differences are likely due to the involvement of the consumer, their socio-economic
background and marketing strategy implemented by the retailer. There is a lack of demographic data
associated with the sample, which prevents segmentation of any kind.
Implications
Marketing is the late arrival to the wine industry, to some it is even the bastard child, but marketing is
now integral to the business of wine. (Hall and R 2008)
The findings of this research will resonate with small producers (ie. the family owned winery) and large
producers (ie. the corporate winery). The information provided by the classification of variety, in this
paper, into the categories of ‘excess loyalty’, ‘niche’ and ‘change of pace’ is of immense benefit (Jarvis and
Goodman 2005). For example, Chardonnay and Sparkling White display ‘excess loyalty’. This means that
bringing a product such as this to market does have a higher likelihood of success and frequent purchase
as opposed to other varietals. In order to sell wine as a mainstream mass-marketed product, a company is
better off selling one of the larger varietals. If a vineyard is going to produce one of the smaller varietals
and have a smaller position in the market place, they have to be able to successfully compete with their
product. In this regard, they have to be certain that their particular variety demonstrates ‘niche’
behaviour, which is indicative of potential success in the market and not ‘change of pace’ of behaviour
which is generally regarded as unsustainable financially.(Jarvis and Goodman 2005).
Truly deconstructing the market would make it possible for a more efficient design of retail space both
in terms of brick and mortar retailers and internet sales. Gaining a better understanding of the
duplication of purchase of varieties can add superior value to shelf-space ordering. Utilising the
knowledge acquired from the duplication of purchase method, it can be surmised that it might be more
appropriate to stock Sauvignon Blanc adjacent to Pinot Noir as these varieties have higher than
www.vdqs.net/2010Palermo
8/10
expected sharing. Understanding the structure of a market based upon its attributes is vital for wine
labels (front and back) as described in the summary of key design elements for a wine label (Hall and R
2008). Having the knowledge as to what aspects of an SKU are the most popular with the market and
have the highest levels of loyalty will allow for guidance to those designing the labels for a wine. This
can also assist in the story written on the back label. Awareness of a variety as ‘niche’ or ‘change of
pace’ can affect the label commentary. For example, a cabernet blend that is known to display changeof-pace tendencies can be addressed in the back label with “try something different for a change”.
Conclusion
Evaluation of the attribute of variety greatly enhances the description of the Australian retail wine
market. The ‘observed’ performance measures, though simple in concept, provide extremely useful
insight into the structure of the market for both mainstream and artisanal varieties. Through the use of
the ‘double jeopardy’ phenomenon and polarisation, the market has been classified into managerially
useful segments of ‘excess loyalty’, ‘niche’ and ‘change of pace’ behaviour. There is one wine market
with no evidence of a separate white and red wine category. Dispelling this myth should play an
important role in the way market research is conducted in the Australian wine market in the future
especially if these findings can appropriately be disseminated to industry and academia.
References
Bass, F. M., A. Jeuland, et al. (1976). "Equilibrium Stochastic Choice and Market Penetration Theories: Derivations
and Comparisons." Management Science 22(No. 10, June): 1051-1063.
Casini, L., A. Corsi, et al. (2008). The impact of wine attributes on the behavioural loyalty of italian consumers. 4th
International Conference of the Academy of Wine Business Research.
Chatfield, C. and G. Goodhardt (1975). "Results concerning brand choice." Journal of Marketing Research
12(February): 110-113.
Ehrenberg, A. S. C. (1988). Repeat-buying: facts, theory and applications. London, Oxford University Press.
Ehrenberg, A. S. C., G. Goodhardt, et al. (1990). "Double Jeopardy revisited." Journal of Marketing 54(3): 82-91.
Ehrenberg, A. S. C. and G. J. Goodhardt (1970). "A model of multi-brand buying." Journal of Marketing Research
7(February): 77-84.
Ehrenberg, A. S. C., M. Uncles, et al. (2000). Armed to the Teeth - An Exercise in Understanding Buyer Behaviour Part A and Answers to Part A. London, South Bank University.
Ehrenberg, A. S. C. and M. D. Uncles (1999). Understanding Dirichlet-type markets. Report 1 for Corporate
Members. Adelaide, Ehrenberg-Bass Institute for Marketing Science: 1-26.
Ehrenberg, A. S. C., M. D. Uncles, et al. (2004). "Understanding brand performance measures: using Dirichlet
benchmarks." Journal of Business Research 57(12): 1307-1325.
Euromonitor International (2004). Wine - Australia: 1-15.
Fader, P. S. and B. G. S. Hardie (1996). "Modeling Consumer Choice Among SKUs." Journal of Marketing Research
XXXIII: 442-452.
Fader, P. S. and D. C. Schmittlein (1993). "Excess Behavioral Loyalty for High-Share Brands: Deviations from the
Dirichlet Model for Repeat Purchasing." Journal of Marketing Research 30(November): 478-493.
Goodhardt, G. J., A. S. C. Ehrenberg, et al. (1984). "The Dirichlet: A comprehensive model of buying behaviour."
Journal of the Royal Statistical Society 147(5): 621-655.
Habel, C. A. (2007). Formalising double jeopardy and deconstructing dynamics in repeat purchase markets. School
of Marketing. Adelaide, University of South Australia: 218.
Hall, C. M. and M. R (2008). Wine marketing: a practical guide. Oxford, UK, Elsevier Ltd.
Jarvis, W. (2006). An analysis of variation in attribute level loyalty using polarisation. School of Marketing.
Adelaide, University of South Australia.
Jarvis, W. and S. Goodman (2005). "Effective marketing of small brands: niche positions, attribute loyalty and
direct marketing." Journal of Product & Brand Management 14(5): 292-299.
www.vdqs.net/2010Palermo
9/10
Jarvis, W., C. Rungie, et al. (2006). "Using polarisation to identify variations in behavioural loyalty to price tiers."
Journal of Product & Brand Management 15(4): 257-264.
Jarvis, W., C. Rungie, et al. (2003). Analysing wine behavioural loyalty. International Wine Marketing Colloquium,
Adelaide, South Australia, University of South Australia.
Jarvis, W., C. Rungie, et al. (2007). "Revealed preference analysis of red wine attributes using polarisation."
International Journal of Wine Business Research 19(2): 127-138.
Jarvis, W., C. Rungie, et al. (2007). "The polarisation method for merging data and analysing loyalty to product
attributes, price and brands in revealed preference." International Journal of Market Research 49(4: Data
Integration Special Issue): 489-513.
Kahn, B. E., M. U. Kalwani, et al. (1988). "Niching Versus Change-of-Pace Brands: Using Purchase Frequencies and
Penetration Rates to Infer Brand Positionings." Journal of Marketing Research 25(November): 384-390.
Li, F., C. Habel, et al. (2006). Excess loyalty - how often does it occur? Australian and New Zealand Marketing
Academy (ANZMAC) Conference. Brisbane, Queensland.
Mansfield, A. (2004). How do tourism destinations compete? an analysis of the duplication of purchase law.
Marketing Science Centre. Adelaide, University of South Australia.
Murphy, T. C. (2006). Buying across price piers in the Australian wine market. The Ehrenberg-Bass Institute for
Marketing Science. Adelaide, The University of South Australia. Master of Business (Research): 134.
Rasmussen, M. (2001). The effects of Region on consumer choice behaviour. School of Marketing, Wine
Marketing Research Group. Adelaide, University of South Australia.
Resnick, E. (2008). Wine brands success strategies for new markets, new consumers and new trends. New York,
Palgrave MacMillan.
Romaniuk, J. and J. Dawes (2005). "Loyalty to price tiers in purchases of bottled wine." Journal of Product & Brand
Management 14(1): 57-64.
Rungie, C. (2000). Heterogeneity in Brand Choice. Marketing Science Centre. Adelaide, University of South
Australia.
Rungie, C. (2003). "How to Estimate the Parameters of the Dirichlet Model using Likelihood Theory in Excel."
Marketing Bulletin 14.
Rungie, C. (2005). A simple method for analysing loyalty is growing in popularity Adelaide, School of Marketing,
University of South Australia.
Rungie, C., B. Brown, et al. (2005). "A standard error estimator for the polarization index: Assessing the
measurement error in one approach to the analysis of loyalty." Marketing Bulletin 16.
Rungie, C. and G. Goodhardt (2004). "Calculation of theoretical brand performance measures from the
parameters of the Dirichlet model." Marketing Bulletin 15: 1-19.
Rungie, C. M. and G. Laurent (2006). Polarization is the Best Measure of Brand Loyalty. Working Paper: University
of South Australia and Groupe HEC.
Rungie, C. M., P. Stern, et al. (2003). Loyalty Changes Through Time. EMAC.
Sabavala, D. J. and D. G. Morrison (1977). "A Model of TV Show Loyalty." Journal of Advertising Research
17(December, No. 6): 35-43.
Uncles, M., A. Ehrenberg, et al. (1995). "Patterns of Buyer Behavior: Regularities, Models, and Extensions."
Marketing Science 14(3, Part 2 of 2): G61-G70.
Wright, M., A. Sharp, et al. (1998). "Are australasian brands different?" Journal of Brand and Product
Management 7(6): 465-480.
www.vdqs.net/2010Palermo
10/10