Sunteți pe pagina 1din 12

Food Quality and Preference 25 (2012) 171–182

Contents lists available at SciVerse ScienceDirect

Food Quality and Preference


journal homepage: www.elsevier.com/locate/foodqual

Market price differentials for food packaging characteristics


Simone Mueller Loose a,b,⇑, Gergely Szolnoki c
a
MAPP Centre for Research on Customer Relations in the Food Sector, Aarhus University, Haslegaardsvej 10, 8210 Aarhus V, Denmark
b
Ehrenberg-Bass Institute for Marketing Science, University of South Australia, P.O. Box 2470, Adelaide SA 5000, Australia
c
Geisenheim Research Center, Section of Economics and Market Research, von-Lade-Str. 1, 65366 Geisenheim, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Consumers’ experimentally measured food preferences are strongly influenced by extrinsic product char-
Received 26 December 2011 acteristics such as labelling and packaging. So far it is unknown if and to what degree these preferences
Received in revised form 20 February 2012 for food packaging are also reflected in market price differences. Using a scanner data set for red wine
Accepted 23 February 2012
market transactions in two US markets, this study estimated price premiums and discounts related to
Available online 3 March 2012
a range of packaging characteristics employing a hedonic pricing model. Models partitioning market
prices into attribute related price components were first estimated, giving each product the same weight.
Keywords:
Additionally, products were weighted by unit sales, whereby higher demanded products were more
Hedonic pricing
Label design
strongly reflected in marginal price estimates. Separate models were estimated for domestic and
Label colour imported products to assess the generalisability of implicit packaging prices. Overall, the largest price dif-
Scanner data ferences were found to relate to region of origin followed by label design type, grape variety, label colour
Revealed versus stated preferences and label information. The packaging elements bottle closure and bottle form could only explain low
price variance shares. The importance of packaging elements in explaining market price differences
increased when taking into account unit sales as an indicator of consumer acceptance. Packaging vari-
ables were more important for imported rather than for domestic products, where grape variety and label
information accounted for larger market price differences. Congruent price premiums for a number of
packaging elements across domestic and imported products suggest a generalisable packaging based
market price differentiation, which was also found to largely agree with previous experimental findings.
Implications for food consumer research and food product marketing are discussed.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction This study employs the hedonic pricing methodology, partition-


ing observed product market price into implicit price premiums
The effect of extrinsic product characteristics such as label for individual product characteristics. Wine, a product highly
information and packaging on consumers’ experimentally mea- differentiated in its intrinsic and extrinsic characteristics, was cho-
sured food preferences is well documented. It is so far unknown sen as a suitable food product for this study. The ability of different
to which degree these stated preferences for food packaging are extrinsic product characteristics to generate product expectations
also reflected in market price differences and to what extent exper- has been particularly well researched for wine (Mueller & Szolnoki,
imental results and observed market price differentials agree with 2010). Also, wine is a highly differentiated product with several
each other. Insights into the ability of packaging characteristics to thousand different products existing in the market. This abundance
explain market price differentials would allow food product devel- of different product characteristic combinations provides sufficient
opers to tailor product appearance to conform to other products in data observations for the hedonic price analysis. Nevertheless, the
a certain price category. Knowledge about changes of implicit price methodology presented here is applicable across food products in
premiums observable for packaging elements over time might pro- general.
vide food marketers with strategic insights into upcoming or A unique scanner data set tracking actual market transactions,
receding product design trends. product availability and market prices for more than one thousand
different red wines over one year in two US markets was used to
address the research questions. The data set covers all commer-
cially relevant price tiers and includes information on price
⇑ Corresponding author at: MAPP Centre for Research on Customer Relations in discounts and promotions, thereby truly reflecting those prices
the Food Sector, Aarhus University, Haslegaardsvej 10, 8210 Aarhus V, Denmark.
consumers face in store.
Tel.: +45 8716 5157; fax: +45 8615 3988.
E-mail addresses: SiMu@asb.dk, simone.mueller@marketingscience.info
This paper has four main objectives. First we examine the extent
(S. Mueller Loose). to which product price differences for red wine can be related to

0950-3293/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodqual.2012.02.009
172 S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182

product packaging characteristics, providing a measure for packag- the most important choice drivers followed by brand, label colour,
ing related market price differentiation. In a second step we analyse region of origin and bottle form, but only few origins and brands
if products with high consumer demand differ more systematically were included in the experiment. Overall, chateau and graphic
in implicit price premiums for packaging attributes than niche labels were more preferred than traditional and minimalistic la-
products. Therefore the hedonic pricing analysis is extended by tak- bels, and white and yellow labels had a higher utility than orange
ing consumers’ product acceptance in the form of unit sales into ac- and grey colours, but preferences strongly differed across five con-
count as product weights. Third, we assess if domestic and sumer segments. Studying young wine consumers from Western
imported products differ in their price variation related to packag- Australia, Jarvis, Mueller, and Chiong (2010) found wine label sym-
ing attributes. Differences in wine labelling traditions and consum- bols and headlines to effect wine choice more strongly than grape
ers’ lower ability to meaningfully interpret foreign label variety and region of origin. The chateau label style was preferred
information would suggest that packaging plays a larger role in dif- over animal labels or labels without any symbol. Set in a Chinese
ferentiating imported products. Finally, we relate estimated impli- cultural context, Convey (2011) identified price and country of ori-
cit market prices for packaging characteristics to results from gin as stronger choice drivers for red wine than label colour and la-
previous experimental packaging research. Thereby we assess the bel style. Chinese consumers preferred burgundy red, black and
congruency of packaging related findings from stated and revealed white labels over other colours such as blue, green, yellow and or-
preference approaches. ange. Chateau labels were uniformly preferred over clean labels
and chateau labels with highlighted accents received highest
1.1. Impact of product packaging on stated preferences acceptance.
Concurring elements can be identified across these studies, such
The effect of extrinsic product attributes on consumers’ sensory as a higher preference for chateau labels over labels with animals
product evaluation in laboratory experiments has been extensively or no graphical element. Findings for colour preferences seem to
documented in the literature (Combris, Bazoche, Giraud-Héraud, & be culture dependent. All choice based studies agreed in the exis-
Issanchou, 2009; Lange, Issanchou, & Combris, 2000; Lange, Martin, tence of several consumer segments, which differed in both the
Chabanet, Combris, & Issanchou, 2002; Stefani, Donato, & Cavicchi, particular preferred design and colour as well as in their choice
2006). This research stream, which is based on consumer stated sensitivity to packaging in general (Mueller & Szolnoki, 2010).
preferences about liking, purchase intent or willingness to pay, When analysing market price differences related to packaging
concurs that product expectations can be generated from cues such attributes, it should therefore be taken into account that these
as product information, labelling and packaging (Deliza & MacFie, experimentally observed preference differences could counteract
1996). Thereby consumers unconsciously use symbolic informa- or cancel each other out on the aggregated market level. In such
tion connoted by packaging features to draw inferences about a case, price differentials estimated from aggregated market data
the product content (Becker, van Rompay, Schifferstein, & Galetzka, would underestimate individual consumer preferences for packag-
2011). ing attributes.
Consumer preferences for product design and packaging colour
have been intensively studied in marketing research. Orth and 1.2. Analysis of implicit market prices
Malkewitz (2008) identified five design prototypes, which generate
different product impressions and affect how consumers perceive A considerable number of studies applied hedonic price analysis
product traits such as product quality, prestige and value for to estimate implicit prices for product attributes. They are based on
money. The effect of colour on consumer preferences is based on the hypothesis that any product represents a bundle of character-
conditioning and associative learning processes (Crozier, 1999; istics that define its overall value (Lancaster, 1966). Hedonic pric-
Grossman & Wisenblit, 1999), but was found to strongly depend ing analysis originated in agricultural economics (Waugh, 1928)
on the product and the cultural context (Aslam, 2006; Wei, Ou, and later was mainly used in environmental economics to infer
Luo, & Hutchings, 2012). Accordingly, cross-cultural differences the monetary value of non-directly traded goods. Thereby market
are to be expected, when examining consumer preferences for prices of bundled goods, such as house prices, are analysed to sep-
product design and packaging colour (Malhotra, Agarwal, & Peter- arate the implicit value for non-market goods like environmental
son, 1996). characteristics. For an overview see Bockstael and MacConnell
A number of studies across different cultural contexts have (2007) and Nerlove (1995). The theoretical foundation for the
examined consumers’ preferences for wine packaging cues. Boud- application of hedonic pricing analysis to market goods was laid
reaux and Palmer (2007) measured the effect of wine label image, by Rosen (1974), positing that goods are valued for their utility-
label colour and label layout on purchase intent and associated generating attributes. Under the assumption of perfect competi-
product personality for US west coast consumers. The label image tion this theory suggests that competitive implicit prices reflect
had the strongest effect and wine related images such as grape or consumers’ utility of these attributes when making a purchase
chateaux graphics received the highest valuation, while unusual decision. However, if competition is not perfect or the market is
animals were least preferred. Warm colours (red, orange) and neu- not in equilibrium, then these implicit prices can also be influenced
tral colours (white, black) had a positive effect on purchase intent. by producers’ or intermediaries’ pricing decisions, but the impor-
Also using US west coast respondents, Orth and Malkewitz (2008) tance of this assumption only received very limited attention in
examined associations consumers have with different wine pack- most hedonic pricing studies (Nerlove, 1995).
aging design prototypes. Natural and delicate wine designs were Following Rosen’s approach, a large number of studies esti-
perceived to be of higher quality, massive and contrasting designs mated implicit prices for wine product characteristics such as re-
were most strongly associated with being inexpensive and natural gion and country of origin (Nerlove, 1995; Oczkowski, 1994;
designs were related with higher value for money than nondescript Panzone & Simões, 2009), grape variety (Schamel & Anderson,
designs. 2003; Steiner, 2004), producer characteristics (Ling & Lockshin,
While the previous two studies focused on consumers’ associa- 2003), wine critic scores (Costanigro, McCluskey, & Mittelhammer,
tions with different packaging types, several studies examined the 2007; Landon & Smith, 1997; Oczkowski, 2001; Schamel & Ander-
relative effect of wine packaging elements on stated choice relative son, 2003; Steiner, 2004), availability of label information in a wine
to other wine attributes. For Australian red wine consumers, Muel- magazine (Costanigro et al., 2007) or intrinsic characteristics oper-
ler, Lockshin, and Louviere (2010) identified price and label style as ationalised by expert sensory evaluations (Brentari, Levaggi, &
S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182 173

Zuccolotto, 2011; Cardebat & Figuet, 2004; Combris, Lecocq, & Vis- accessible to previous hedonic pricing studies that mainly used
ser, 1997, 2000). To our best knowledge, no previous study has one point in time wine guide information instead of true market
studied implicit prices for food packaging attributes. transaction data.
The majority of hedonic price analysis studies in food research It is common to estimate implicit prices separately for sub-mar-
assumed that implicit prices exclusively reflect consumer prefer- kets to analyse if they differ in their pricing structure. Schamel and
ences not taking into account the identification issues between de- Anderson (2003) estimated different models for products from
mand and supply functions from market price–quantity Australia and New Zealand, Costanigro et al. (2007) examined dif-
observations already pointed out by Working (1927). Contrary to ferences in the implicit pricing structure between four different red
common practice, Nerlove (1995) emphasises that coefficients wine price segments and Brentari et al. (2011) separately analysed
estimated in hedonic regression reflect both supply and demand implicit prices for different retail formats. Several factors lead us to
considerations, producer costs and consumer preferences. expect the pricing structure between domestic US wines and im-
From the previous studies only Nerlove (1995) seems to have ported wines to differ. Particularly wines from France, Italy and
assessed the degree to which the assumptions of perfect compe- Spain, the ‘old wine world’, have traditionally marketed and la-
tition and market equilibrium were satisfied. Very few or no belled their products with a strong regional focus. Their wine laws,
barriers to market entry and exit is one of the requirements for such as the AOC or AOP, still prescribe the use of varietal blends
perfect competition to be prevalent. For the US wine market, and do not allow the indication of grape varieties on the front label.
the state specific three-tier alcohol trade system, characterised This is contrary the ‘new wine world’, such as the US, where grape
by a complex arrangement of licensed importers, wholesalers varietals play a dominant role in product labelling and marketing
and retailers, poses a major entry barrier for any wine producer, (Goodman, 2009). Accordingly we would expect that grape variety
particularly for import wines. Also, high food retailer concentra- has a stronger impact on price differentiation for domestic US
tion suggests that intermediaries can exert influence on market wines compared to imported wines. A second reason relates to
prices, incongruent with the assumption of perfect competition. product cue predictability, consumers’ reduced ability to interpret
In front of this background, we cannot assume that there is per- and utilise label information of imported wines they are less famil-
fect competition in the US wine market. Accordingly, implicit iar with. From the expected lower price separation for grape vari-
prices may not only reflect consumer preferences but are very ety and front label information we expect packaging to play a
likely also influenced by producer and retailer decisions. Hence, relatively larger role for price differentiation for imported wines.
we do not claim that the implicit prices we estimate are a perfect
and unbiased indicator of consumer preferences but can also 1.3. Research questions
reflect producer cost factors.
Contrary to consumer experiments, hedonic price analysis is This study contributes to the existing body of research by
unable to establish causality. It could therefore be argued that con- addressing the following four research questions:
sumer experiments result in more valid measures for consumer
preferences. However, it has to be considered that laboratory (1) To what degree can market price differences for red wine be
experiments often face different kinds of problems with incentive accounted for by product differences in packaging character-
compatibility and social demand bias, particularly when it comes istics? The answer to this question indicates the extent to
to the measurement of price and willingness to pay (Norwood & which wine offered at different price levels systematically
Lusk, 2011). Also, to manage complexity laboratory experiments differs in its packaging characteristics, providing insights
can only examine few product attributes and attribute levels; they for product development and pricing decisions.
are rarely able to reflect the full variance of products available in (2) How are implicit prices for packaging affected, when con-
the market. This limitation to few experimental factors can lead sumers’ revealed product preferences are taken into account
to an overestimation of their true effect size (Gao & Schroeder, in the form of unit sales? An increased importance of pack-
2009). Accordingly, stated preferences in consumer experiments aging would indicate that products with high consumer
and revealed preferences from market analysis can be seen as demand are more systematically differentiated in their
two imperfect measurements of the same underlying consumer packaging characteristics across different price levels.
preferences, which should be triangulated, contrasted and com- (3) How strongly do domestic and imported products differ in
pared to establish validity and to identify more generalisable pat- their implicit prices and the degree of price variation related
terns (Mueller, Osidacz, Francis, & Lockshin, 2010). to packaging attributes? Congruent packaging price premi-
Existing hedonic pricing studies for food and wine considered ums would indicate rather generalisable findings across
each product with the same weight when estimating implicit products from different origins.
prices. Thereby niche products with low sales and highly de- (4) To what extent do implicit market prices for packaging char-
manded products are given the same importance. Complementing acteristics concur with experimentally measured wine pack-
the traditional approach we suggest to also utilise information con- aging preferences? Congruency between stated and revealed
sumers have provided about their product preferences in the form preferences would indicate that market price premiums
of unit sales data, indicating how many units of a product have reflect consumer preferences to a larger degree than produc-
been purchased by consumers over a certain time period. Weight- tion costs and pricing decisions by market intermediaries.
ing products by their market share or unit sales in the hedonic
pricing model, an approach previously taken in land economics 2. Description of the data set
(see for instance Parsons, 1990), allows us to reflect consumer pref-
erences across the total of all wine purchases, not just across all This study utilised a unique retail measurement data set from
products bought at least once. This approach mitigates the effect the AC Nielsen ScantrackÒ, tracking red wine retail prices, unit
of outlier and niche products, which exist in the market because sales and store availability across 149 retailer chains and 18 liquor
the importer and wholesaler managed to acquire a retailer listing, retailers for two major US metropolitan markets, Chicago (IL) and
but which are rarely purchased by consumers. By weighting prod- Tampa (FL). The data set covers transactions over thirteen four
ucts by unit sales, stronger emphasis is given to those product week periods from 29 July 2007 to 26 July 2008, representative
characteristics which are highly demanded by consumers. This of wine sales in these two markets over one full year. The data is
weighting approach requires sales information, which was not based on retailers reporting sales (checkout counter scanner data)
174 S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182

and actual product prices to AC Nielsen, which aggregates those for 2.4. Origin and grape variety variables
each single product over a four week period.
Our data base reflects actual market transaction at true market Based on the product information in the data set and the down-
prices. It thereby is distinguished from most previous hedonic loaded label images all products were classified by country and re-
wine price studies, which were mainly limited to recommended gion of origin. The 1168 products originated from 14 different
retail prices in wine reviews or wine guide databases. These prices countries and 104 regions of origin. Small origin groups only con-
often differ from later market prices, one reason being that price taining very few products were condensed into larger categories
promotions and price discounts are not reflected. Also, wine guides and a total of 27 country specific regions were included in the anal-
are mainly limited to premium and superpremium products, ysis as effects coded nested categorical variables (Table 2).
neglecting products in lower price segments, which represent the From a total of 31 identified red wine grape varieties nine had
dominant market sales share. This study is limited to red wine. sufficient observations to enter the model as categorical variables.
Red and white wine differ in their average prices, grape varieties, Agreeing with our argumentation in Section 1.2, imported wines
region of origins and ageing requirements, commanding separate were found to have a considerably higher share of products con-
analysis (Thrane, 2004). sisting of other or blends of different grape varieties (see Table 2).

2.1. Market and data selection 2.5. Front label information variables

Scanner data from two major metropolitan areas in Illinois and While the variables for grape varieties and regions of origin de-
Florida were selected, which represent the 2nd and 3rd highest per scribed in Section 2.4 relate to objective constituting product char-
capita wine consumption in the US, after California. Contrary to the acteristics retrieved from the data base or from the producer, we
US west coast, where domestic wine consumption dominates, the separately analysed the price impact of providing this information
two selected markets have a considerable import wine market to consumers on the front label of the product. It can be argued
share of about 40 percent. Our data covered the Chicago (9.1 mil- that information stated on the front label is more easily accessible
lion inhabitants) and Tampa (5.4 million) metropolitan areas, to consumers than information provided on the back label or
jointly representing 4.8% of the US population. elsewhere.
Retail data for Tampa covered red wine sales in food retail and Six dummy variables were used to measure price differences re-
liquor stores. The Chicago data additionally contained drug stores, lated to the existence (yes/no) of front label information for grape
but these represented only 3% of the total wine sales volume. The variety, country of origin, region of origin, medals, specific produc-
wine sales share of the food retail and liquor store channels were of tion information (organic, reserve, estate bottled, single vineyard,
similar size (47% food retail and 53% liquor stores in Tampa and estate grown, old wine) and other information (e.g. on history of
49% and 48% in Chicago). winery). There were no sufficient observations available in our
The raw data sets included 2449 individual products for Chicago data set to separately estimate price premiums for individual pro-
and 2539 products for Tampa. To avoid a price confound by vol- duction information, such as organic, as differentiated by Costani-
ume, in a first step all products different from 750 mL were ex- gro et al. (2007).
cluded, leaving about 75% of products. In a second step, wines
which were sold in both markets were selected, in order to avoid 2.6. Packaging variables
modelling products with only local relevance. This left 1168 red
wines in the data set for further analysis. We first aimed to categorise the products according to the five
design types developed by Orth and Malkewitz (2008), which were
2.2. Wine identification developed for wines on the US west coast, but it turned out that
particularly imported wines did not seem to fit under any of the
In the raw data sets each wine was characterised by the brand, available categories. It was therefore required to develop new label
country of origin and wine name on the label, which either con- design type categories. To categorise products according to wine
tained the grape variety or region of origin. Based on this product label design styles a random selection of 500 of the product photo-
identification photographs showing the front label and the total graphs were printed in colour. We asked eight regular wine con-
product were accessed for all 1168 wines through the internet or sumers from South Australia, who differed in gender, age and
requested from producers. Information in the raw scanner data wine involvement, to sort the 500 wine photographs into groups
base on region of origin and grape variety was validated against they perceived to be distinct wine label categories.
online and producer information and supplemented if possible. Although respondents differed in the number of categories they
created, whereat more wine interested consumers not surprisingly
2.3. Price, units sold, store availability showed a stronger differentiation, we observed a large accordance
in the general type of categories identified. In a consensus meeting
We calculated a unit weighted average price from promotional respondents agreed on eight wine label design categories for which
and non-promotional prices over the total one year time period one example label is provided in Table 3. The nondescript and del-
across markets and sales channels that represents our dependent icate label types are identical to the same design types in Orth and
variable. The total number of units sold was calculated by adding Malkewitz. The animal graphic and artwork graphic label types were
units across all sales channels and both markets across the full perceived as two distinct subcategories of the ‘contrasting’ design
year. An availability index was provided in the raw data set for type in Orth and Malkwitz. Respondents also expressed the need
each product by sales channel, indicating the percentage of stores to subcategorize the ‘natural’ category in Orth and Malkewitz
at which the product was listed. We calculated an aggregated and two types of clean and chateau labels were discriminated.
availability indicator for each product by a sales volume weighted Within both of the clean and chateau categories a difference was
average across markets and sales channels. Descriptive summary made by respondents between the standard and the highlighted
statistics in Table 1 show that our data set covers a very wide range versions, where the latter was characterised by the use of red
of the commercially relevant red wine market spectrum. It in- and golden accents and label highlights. Respondents did not iden-
cludes all relevant price tiers, wines with wide and limited store tify any products for the ‘massive’ design type in Orth and Malke-
availability and more and less frequently selling wines. witz, most likely because we had excluded products with more
S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182 175

Table 1
Descriptive summary of continuous variables for n = 1168 red wines (August 2007–July 2008), Chicago (IL) and Tampa (FL) metropolitan areas.

Median Mean Stdev Min Max


Price $11.32 $14.76 $11.42 $3.09 $157.43
Store availability [0;100] 32.7 34.8 22.2 0.2 95.9
Units sold 5,084 16,537 30,620 28 303,420
Sales volume $67,211 $165,686 $268,723 $407 $2,438,080

Note: Sum of total sales value in data set was $90.8 million.

than 750 mL from the data set. The development of our label de- Although consumers’ sensory liking was found to be affected by
sign style categories was completed in late 2008 and it can be wine ratings (Siegrist & Cousin, 2009), several reasons motivated
interpreted as a confirmation of our approach that a product cate- us to not include wine critics’ scores in the hedonic pricing model.
gory specific market research company later independently devel- First, the majority of wines sold through food retail and liquor
oped a relatively similar wine label typology, also consisting of stores are not rated by any wine critic. For a subsample of 210
eight label types (WineIntelligence, 2011). products from our data set we identified less than 10% of wines
The colour of wine labels was first coded in detail and then later of one vintage to be rated by any of seven different wine critics.
summarised into seven categories. Thereof two were the non-col- This large amount of missing data poses severe estimation prob-
our groups white and black and a middle category contained cream lems (Kamakura, Basuroy, & Boatwright, 2006). Second, there are
and grey colours. Two colour categories for labels with one domi- several sensory studies which found a disconnect between con-
nant colour or a mix of several colours were developed. Similar to sumers’ blind product preferences and expert evaluations (Delgado
the approach by Boudreaux and Palmer (2007) colours were fur- & Guinard, 2011; Goldstein et al., 2008; Schiefer & Fischer, 2008),
ther partitioned into a ‘warm’ subcategory with psychologically suggesting that critics’ ratings are only very limitedly informative
activating colours (red, orange, yellow) and a ‘cold’ subcategory of consumers’ sensory preferences. Last, there is a potential model
(blue, green, violet). specification endogeneity issue for those critics who do not assess
Three different bottle shapes, straight, rounded and other were wines blindly, as price was found to exert a very strong subliminal
identified. In the wine industry the straight and round shapes are effect on human product evaluation (Plassmann, O’Doherty, Shiv, &
also called ‘Bordeaux’ and ‘Burgundy’ style, but we refrained from Rangel, 2008). For informed product assessments it can be sug-
using these regional names for bottle shape to avoid any potential gested that the impact of price on wine critics’ ratings is potentially
confusion with the identical region of origin variables. The am- larger than the impact ratings have on market price (Thrane, 2004).
phora-shaped bottle form with wider shoulders and a straight nar- This unresolved causality in the theoretical relationship between
rowing body is the most prominent examples of several ‘other’ price and wine ratings should also be taken into account when
bottle shapes. interpreting marginal price effects for critics’ scores reported in
Products were categorised into cork and screw cap categories previous hedonic pricing studies.
according to the wine closure they had on the product photo. Sim-
ilar to a consumer at the time of the decision in store, from the out-
3. The model and specification tests
side we were unable to discriminate natural from synthetic cork
closures. A low effect of natural versus synthetic cork could be ex-
According to conventional hedonic models (Rosen, 1974), the
pected anyway because US consumers’ valuation was found to only
price of a product is a linear function of its characteristics for which
weakly differ between both cork types but a negative effect was
implicit prices are predicted. For this study the suggested relation-
expected for screw caps (Marin & Durham, 2007). We also cannot
ship between price, market availability and product characteristics
exclude the possibility that producers provide different closure
can formally be written as:
versions of their products to different retailers.
After all categories for label design style, label colour, bottle Price ¼ f ðavailability; grape variety; origin; front label information;
shape and bottle closure were determined, objective coding books label style; label colour; bottle form; closureÞ þ e ð1Þ
were developed showing several examples defining each label var-
iable category. Using these coding books two independent coders The model assumes that price is affected by a product’s avail-
categorised each of the 1168 products in the data set. Both coders ability, its grape variety (eight variables), its origin (26 nested vari-
agreed in their coding of the label categories for 88.4% of all prod- ables for country and region), its packaging, and its front label
ucts. An agreement was reached jointly for those cases where both information. Each product’s packaging is defined by its label style
coders differed in their packaging categorisation. (seven variables), label colour (six variables), bottle form (two vari-
ables) and closure, including a disturbance or error term e. Six bin-
ary variables were used to quantify the presence or absence of
2.7. Variables not included in the model front label information: country of origin, region of origin, grape
variety, additional specific production information and other infor-
Information on the alcohol content (Brentari et al., 2011; Ner- mation. Availability entered the model as a linear and quadratic
love, 1995) was not considered in our model because wines of dif- term and was mean centred to avoid collinearity issues. All cate-
ferent vintages and potentially different alcohol levels were sold gorical variables were effects coded, allowing us to calculate an im-
under the same product name over the course of one year and can- plicit price for the reference category that is not confounded with
not be separately identified in the data set. the constant, which in our model represents the average price
Because the products analysed in our study were derived from (grand mean).
market representative retail scanner data and not retrieved from Following our discussion in Section 1.2, we extend common
selective product reviews or wine guides, comparable standardised practice in hedonic pricing research for food products by also con-
information on the chemical composition or sensory descriptive sidering consumer preferences in form of unit sales when estimat-
evaluation through one uniform panel (Brentari et al., 2011; Com- ing implicit prices. When weighting products according their unit
bris et al., 1997; Nerlove, 1995) was not available to us. sales, price differentials are estimated across the total sum of
176 S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182

Table 2
Sample composition for categorical variables in percent.

Unweighted – share of products Weighted – share of units sold


Total sample Domestic US Import Total sample
n 1168 630 538 19.3 million
Grape variety
Cabernet blends 2.9 2.1 3.9 2.3
Cabernet Sauvignon 23.0 31.0 13.8 29.0
Malbec 2.7 0.2 5.6 1.3
Merlot 17.1 22.5 10.8 21.8
Pinot Noir 12.5 17.6 6.5 11.1
Shiraz 11.6 6.3 17.8 12.5
Tempranillo 1.5 0.0 3.3 0.7
Zinfandel 6.8 12.5 0.0 5.0
Others 21.8 7.8 38.3 16.2
Region of origin
US_Alexander Valley 2.1 3.8 1.2
US_California 17.0 31.5 30.0
US_Central Valley 3.9 7.1 7.9
US_Napa 9.8 18.3 3.9
US_Sonoma 6.6 12.2 5.3
US_Washington 2.6 4.8 4.9
US_Oregon 1.5 2.9 0.4
US_other 10.5 19.5 6.2
Argentina 3.9 8.6 1.6
Australia_SEA 6.4 13.8 17.0
Australia_other 5.8 12.6 3.0
Chile_Central Valley 1.0 2.2 1.0
Chile_Colchagua 1.6 3.5 0.5
Chile_Maipo 0.9 2.0 0.6
Chile_other 1.3 2.8 0.9
France_Beaujolais 1.0 2.2 1.8
France_Bordeaux 0.8 1.7 0.7
France_Rhône 0.8 1.7 0.3
France_other 2.3 5.0 1.4
Italy_Chianti 3.5 7.6 3.4
Italy_Toscana 1.5 3.2 0.8
Italy_Veneto 2.5 5.4 1.3
Italy_other 3.1 6.7 1.9
South Africa 2.1 4.5 0.7
Spain_Rioja 1.4 3.0 0.9
Spain_other 4.5 9.7 2.2
Other origins 1.7 3.7 0.3
Front label information
Country of origin 31.2 0.8 66.7 29.9
Region of origin 83.0 89.8 74.9 81.8
Grape variety 82.8 94.4 69.1 88.6
Medal 1.6 1.0 2.4 1.0
Production method 1.5 10.5 12.6 8.0
Other information 8.0 0.3 16.9 6.6
Label type
Clean 7.7 7.1 8.4 8.0
Clean_highlight 27.1 24.6 29.9 27.5
Chateau 10.3 10.0 10.6 6.7
Chateau_highlight 12.2 14.6 9.5 17.8
Delicate elegant 6.3 6.0 6.7 7.0
Animal graphic 9.2 8.4 10.0 12.8
Artwork graphic 23.5 25.4 21.4 18.9
Nondescript 3.7 3.8 3.5 1.4
Label colour
Cream, grey 17.4 20.2 14.1 15.6
Black 6.6 6.7 6.5 3.8
Unicolour warm 9.9 7.5 12.8 13.8
Unicolour cold 6.4 6.2 6.7 7.7
Multicolour warm 4.9 5.1 4.6 3.8
Multicolour cold 12.6 12.9 12.3 13.1
White 42.2 41.6 42.9 42.3
Bottle form
Straight 67.0 60.3 74.9 73.8
Rounded 13.4 18.4 9.7 10.1
Other 19.6 21.3 15.4 16.1
Bottle closure
Screw cap 8.5 5.1 12.5 7.8
Cork 91.5 94.9 87.5 92.2
S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182 177

Table 3
Label design category examples.

Label style Example Label style Example


Clean Delicate elegant

Clean highlight Animal graphic

Chateau Artwork graphic

Chateau highlight Nondescript

Note: further label examples can be obtained from the authors.

products sold in the two markets under consideration, not only specification is the most common model form (Brentari et al.,
across products offered on the shelf. As expected, we observed a 2011; Cardebat & Figuet, 2004; Combris et al., 1997; Nerlove,
negative correlation between price and unit sales, which was only 1995; Oczkowski, 2001; Panzone & Simões, 2009; Schamel &
of moderate size (r = 0.215, p < 0.001) and did not introduce any Anderson, 2003; Steiner, 2004). An examination of the correlation
collinearity issues. Weighting was performed by means of fre- matrices for the coefficient estimates suggested that no serious de-
quency weights in STATA 11.0. gree of multicollinearity as dependency between price and error
The theoretical model described in Eq. (1) limits the type of term was present in the data.
explanatory variables, but it does not restrict the functional form The equation that was estimated according the log-linear spec-
to be estimated. When using OLS regression, the functional form ification for the pooled model is given as:
of hedonic price models is empirically determined by applying a
number of tests, identifying the model specification which least X
8
2
violates the assumptions of linear regression. In the empirical liter- ln P i ¼ const þ aAvailmc þ bAvailmc þ cj Grapej
j¼1
ature of hedonic pricing, a variety of different functional forms
have been explored and specifications from the ladder of power X
26 X
6 X
7

(including p2, p1.5, linear p, log linear ln p, p0.5, Box–Cox, p0.5, þ dj Regionj þ ej Infoj þ fj Labelj
j¼1 j¼1 j¼1
p1, p1.5 and p2) are usually examined.
We performed a Ramsey RESET test to all those price transfor- X
6 X
2
þ g j Colourj þ hj Bottlej þ kClos þ ei ð2Þ
mations to choose the functional form for the unweighted and
j¼1 j¼1
weighted model. For the unweighted model the inverse quadratic
(p0.5) and log linear (ln p) specifications performed similarly well The domestic model was identical to Eq. (2) with the exception
(RESET test of 0.68 and 1.28) but for the weighted model the log lin- that it contained only six parameters cj for grape varieties grown in
ear model strongly outperformed any other specification (RESET the US and seven parameters dj for domestic US regions. Similarly,
test of 1.69). To allow a consistent comparison of estimates across the import model contained seven parameters cj for grape variety
the unweighted and weighted models we decided to use the log lin- and 18 parameters dj for non-US regions or origin.
ear specification across both models. This choice concurs with the Because of the log-linear transformation, each coefficient esti-
majority of hedonic wine pricing studies, for which the log-linear mate a–k represents the percentage price premium or discount
178 S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182

from the average price related to a product attribute. The price im- thereby gives more weight to highly demanded products. Because
pact of overall product packing is assessed based on the partial price and units sold are weakly correlated (see Section 3) the aver-
contribution of the packaging variables to the model’s explained age price and price dispersion reduces for all weighted models
variance, their relative effect size. Thereby the model is estimated (bottom of Table 5). This indicates that very highly priced and
with and without the variables for label type, label colour, bottle lowly priced niche products are sold less frequently and are there-
shape and closure, whereas the observed difference in explained fore deemphasised in their impact on implicit price estimates.
variance represents the contribution of packaging variables (Eq. How does weighting by unit sales affect the ability of the hedo-
(3)). nic price model to explain price differences? The explained vari-
ance (Adj. R2) increases through weighting by between 3.5% in
Contribution to model fit ¼ R2 total model
the domestic model, 5.8% in the pooled model to 15.1% in the im-
 R2 model without packaging variables ð3Þ port model (Table 4). This increase in explained variance indicates
that prices of more highly demanded products can be better ex-
The attribute importance of individual characteristics Aj is asse-
plained by their product attributes compared to niche products.
sed through the range (max–min) of estimated price effects rela-
What is the effect of taking unit sales into account for the effect
tive to the sum of estimated price ranges across all attributes
size of packaging characteristics? Across all models the contribu-
(Eq. (4)):
tion of packaging variables to the explained variance is higher for
maxðestimateÞ  minðestimateÞ the weighted compared to the unweighted model (Table 4). This
Aj Aj
Attribute importances Aj ¼ Pn : ð4Þ increase is highest for the domestic model (5.7%) and smaller for
i¼1 maxðestimateÞ  minðestimateÞ imported wines (1.7%). A similar result is observed for packaging
Aj Aj
attributes’ importance as share of price range in Table 6, which
slightly increases across all models and where the effect is largest
4. Results for domestic products, increasing from 29.2% to 36.4% joined attri-
bute importance. This indicates that products with higher demand
4.1. Importance of packaging in explaining price variation are more strongly differentiated in their product packaging across
different price tiers.
The effect size of packaging characteristics in explaining market Stronger price discrimination by packaging attributes can also
price differences is assessed by their contribution to explained var- be observed when comparing the estimated percentage price pre-
iance (Eq. (3)). According to results in Table 4, packaging variables miums and discounts across the unweighted and weighted model
overall contribute about 3.5% to explained variance (Adj. R2) of the (Table 5). For the pooled model, the number of significant price ef-
pooled unweighted model, which is a rather modest improvement. fects reduces for grape variety, region, bottle form and closure,
The importance of each attribute was calculated according Eq. (4) while more label design types become statistically significant.
based on estimated implicit price premiums from Table 5. Impor- For domestic products we observed an increase in the number of
tance weights in percent and attribute ranks are provided in Table 6. significant price effects for label type, label colour and bottle form,
Across all products region of origin had the highest importance, when taking unit sales into account, while the price discount for
representing 54% of the sum of price ranges in the unweighted screw caps became insignificant. The increase of attribute impor-
pooled model. These strong estimated price effects for region of ori- tance for packaging attributes mainly came at the cost of region
gin concur with similar findings in previous hedonic price studies of origin and front label information, for which the share of esti-
(Combris et al., 1997; Schamel & Anderson, 2003). In the light of mated price differences decreased across all weighted models. This
imperfect competition these price differences can be related to dif- suggests that there are lower regional and information related
ferences in consumers’ perceived regional reputation as well as to price differences for highly demanded products but that niche
differences on the supply side, such as production costs, market products with lower sales are more strongly price differentiated
power or differences in market access. Packaging attributes jointly by region of origin and front label information.
account for 27.9% of the sum of estimated implicit price differences,
thereof label design type and label colour were more important than 4.3. Differences in the pricing structure between domestic and
closure and bottle form. In the pooled unweighted model label type imported products
was the second most important attribute in explaining price differ-
ences followed by label information and grape variety. A number of differences appear when analysing domestic and
imported wines separately. First it can be observed that domestic
4.2. Effect of weighting products by unit sales wines on average sell at higher prices than imported wines
($16.96 versus $12.18) and that this difference is slightly reduced
Weighting products by unit sales in the hedonic price model when taking unit weights into account ($11.23 versus $8.68), also
estimates implicit prices across all 1.93 million units sold and see Table 5. This price difference is also reflected in the sales value

Table 4
Effect size – contribution of packaging variables to explained variance of models (in percent).

All variables Without packaging Contribution packaging


Pooled model (n = 1168) Unweighted R2 57.07 53.02 4.05
Unweighted Adj. R2 54.74 51.27 3.47
Weighted Adj. R2/R2 60.56 54.15 6.41
Domestic US model (n = 630) Unweighted R2 62.44 57.61 4.83
Unweighted Adj. R2 59.89 56.07 3.82
Weighted Adj. R2/R2 63.40 53.89 9.51
Import model (n = 538) Unweighted R2 48.33 39.93 8.40
Unweighted Adj. R2 42.91 36.00 6.91
Weighted Adj. R2/R2 57.98 49.34 8.64

Notes: For frequency weighted model R2 is identical to Adj. R2 at four decimal digits.
S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182 179

Table 5
Hedonic pricing model estimates (percent price premium or discount).

Dependent variable ln (price) Pooled model Domestic US model Import model


Un-weighted Coef. Weighted Coef. Un-weighted Coef. Weighted Coef. Un-weighted Coef. Weighted Coef.
Availability
Linear term 0.005⁄⁄⁄ 0.010⁄⁄⁄ 0.006⁄⁄ 0.015⁄⁄⁄ 0.003 0.002
Quadratic term 8.77+,⁄⁄⁄ 9.45+,⁄⁄⁄ 9.22+,⁄⁄⁄ 13.9+,⁄⁄⁄ 7.20+,⁄⁄ 3.76+
Grape varietals
Cabernet Sauvig. 0.018 0.003 0.038 0.006 0.005 0.069
Cabernet blends 0.028 0.001 0.148 0.191⁄⁄ 0.050 0.051
Malbec 0.130 0.176 0.098 0.215
Merlot 0.101⁄⁄⁄ 0.032 0.170⁄⁄⁄ 0.058 0.062 0.021
Pinot Noir 0.009 0.021 0.018 0.223⁄⁄⁄ 0.102 0.019
Shiraz 0.126⁄⁄⁄ 0.051 0.293⁄⁄⁄ 0.314⁄⁄⁄ 0.029 0.025
Tempranillo 0.090 0.128 0.119 0.221⁄⁄
Zinfandel 0.030 0.056 0.110⁄ 0.021
Others 0.002 0.053 0.089 0.057 0.031 0.140⁄⁄
Region of origin
US_Alexander Val. 0.667⁄⁄⁄ 0.539⁄⁄⁄ 0.272⁄⁄⁄ 0.262⁄⁄⁄
US_California 0.146⁄⁄⁄ 0.186⁄⁄⁄ 0.518⁄⁄⁄ 0.435⁄⁄⁄
US_Central Valley 0.146⁄⁄ 0.111⁄⁄ 0.209⁄⁄⁄ 0.132⁄⁄⁄
US_Napa 0.847⁄⁄⁄ 0.664⁄⁄⁄ 0.434⁄⁄⁄ 0.371⁄⁄⁄
US_Sonoma 0.423⁄⁄⁄ 0.333⁄⁄⁄ 0.025 0.058
US_Washington 0.196⁄⁄⁄ 0.096 0.226⁄⁄⁄ 0.384⁄⁄⁄
US_Oregon 0.647⁄⁄⁄ 0.573⁄⁄⁄ 0.255⁄⁄⁄ 0.304⁄⁄⁄
US_other 0.363⁄⁄⁄ 0.218⁄⁄⁄ 0.032 0.043
Argentina 0.227⁄⁄ 0.101 0.018 0.296⁄⁄
Australia_SEA 0.206⁄⁄⁄ 0.163⁄⁄⁄ 0.007 0.012
Australia_other 0.312⁄⁄⁄ 0.213⁄⁄⁄ 0.394⁄⁄⁄ 0.307⁄⁄⁄
Chile_Central Val. 0.338⁄⁄⁄ 0.282⁄⁄⁄ 0.127 0.131
Chile_Colchagua 0.158⁄ 0.126 0.011 0.013
Chile_Maipo 0.019 0.029 0.209⁄ 0.178⁄
Chile_other 0.231⁄⁄ 0.452⁄⁄⁄ 0.098 0.320⁄⁄
France_Beaujolais 0.763⁄⁄⁄ 0.380⁄⁄⁄ 0.639⁄⁄⁄ 0.354⁄⁄⁄
France_Bordeaux 0.152 0.277⁄⁄⁄ 0.071 0.126
France_Rhône 0.402⁄⁄⁄ 0.058 0.335⁄⁄ 0.009
France_other 0.167⁄⁄ 0.027 0.053 0.066
Italy_Chianti 0.294⁄⁄⁄ 0.166⁄ 0.111 0.102
Italy_Toscana 0.015 0.102 0.192⁄⁄ 0.000
Italy_Veneto 0.071 0.019 0.295⁄⁄⁄ 0.180⁄⁄⁄
Italy_other 0.083 0.314⁄⁄ 0.252⁄⁄⁄ 0.161⁄
South Africa 0.195⁄⁄ 0.037 0.039 0.032
Spain_Rioja 0.315⁄⁄⁄ 0.008 0.153 0.124
Spain_other 0.274⁄⁄⁄ 0.210⁄⁄⁄ 0.119⁄ 0.167⁄⁄
Other origins 0.077 0.087 0.262 0.162
Front label information
Country origin 0.042⁄⁄ 0.030 0.333⁄⁄⁄ 0.202 0.029 0.022
Region of origin 0.040⁄⁄ 0.043⁄⁄ 0.066⁄⁄ 0.057⁄ 0.034 0.030
Grape variety 0.067⁄⁄⁄ 0.007 0.133⁄⁄⁄ 0.032 0.016 0.006
Medal 0.063⁄ 0.072⁄⁄ 0.023 0.079⁄ 0.009 0.034
Production method 0.097⁄⁄⁄ 0.112⁄⁄⁄ 0.100⁄⁄⁄ 0.139⁄⁄⁄ 0.104⁄⁄⁄ 0.100⁄⁄⁄
Other information 0.210⁄⁄⁄ 0.135⁄⁄⁄ 0.014 0.086 0.231⁄⁄⁄ 0.136⁄⁄⁄
Label type
Nondescript 0.220 0.211 0.289 0.255 0.107 0.151
Clean 0.070 0.192⁄⁄⁄ 0.042 0.246⁄⁄⁄ 0.128⁄⁄ 0.070
Clean_highlight 0.095⁄⁄⁄ 0.147⁄⁄⁄ 0.126⁄⁄⁄ 0.170⁄⁄⁄ 0.070⁄ 0.117⁄⁄⁄
Chateau 0.022 0.085 0.074 0.154⁄ 0.018 0.030
Chateau_highlight 0.087⁄⁄⁄ 0.129⁄⁄⁄ 0.110⁄⁄ 0.187⁄⁄⁄ 0.065 0.034
Delicate elegant 0.001 0.004 0.041 0.019 0.040 0.026
Animal graphic 0.085⁄⁄ 0.100⁄⁄ 0.070 0.120⁄ 0.078 0.071
Artwork graphic 0.044 0.108⁄⁄ 0.141⁄⁄⁄ 0.160⁄⁄ 0.036 0.005
Label colour
White 0.018 0.032 0.007 0.045 0.024 0.010
Cream, grey 0.093⁄⁄⁄ 0.064 0.037 0.014 0.189⁄⁄⁄ 0.216⁄⁄⁄
Black 0.155⁄⁄⁄ 0.123⁄⁄⁄ 0.154⁄⁄⁄ 0.174⁄⁄ 0.170⁄⁄⁄ 0.103⁄
Unicolour warm 0.095⁄⁄⁄ 0.130⁄⁄⁄ 0.070 0.129⁄⁄ 0.142⁄⁄⁄ 0.163⁄⁄⁄
Unicolour cold 0.102⁄⁄ 0.133⁄⁄⁄ 0.111⁄⁄ 0.143⁄⁄ 0.189⁄⁄⁄ 0.226⁄⁄⁄
Milticolour warm 0.002 0.096⁄⁄ 0.020 0.086⁄ 0.072 0.132⁄
Multicolour cold 0.067⁄⁄ 0.051 0.003 0.048 0.124⁄⁄⁄ 0.071
Bottle form
Straight 0.074⁄⁄⁄ 0.029 0.039 0.095⁄⁄⁄ 0.114⁄⁄⁄ 0.024
Rounded 0.014 0.026 0.016 0.041 0.047 0.049
Other bottle form 0.088⁄⁄⁄ 0.003 0.023 0.135⁄⁄⁄ 0.161⁄⁄⁄ 0.025

(continued on next page)


180 S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182

Table 5 (continued)

Dependent variable ln (price) Pooled model Domestic US model Import model


Un-weighted Coef. Weighted Coef. Un-weighted Coef. Weighted Coef. Un-weighted Coef. Weighted Coef.

Closure
Screw cap 0.049⁄⁄ 0.030 0.084⁄⁄ 0.034 0.040 0.027
Cork 0.049 0.030 0.084 0.034 0.040 0.027
Constant 2.749⁄⁄⁄ 2.402⁄⁄⁄ 2.668⁄⁄⁄ 2.650⁄⁄⁄ 2.567⁄⁄⁄ 2.300⁄⁄⁄
n 1168 630 538
R2 0.571 0.606 0.624 0.634 0.483 0.580
Adj. R2 0.547 0.599 0.429
Average price $14.76 $10.21 $16.96 $11.23 $12.18 $8.68
Stdev. average price $11.42 $5.62 $13.05 $6.26 $8.46 $4.05
Aver. availability 34.8 62.3 35.2 62.6 34.4 61.8
Stdev. aver. availab. 22.2 21.5 22.6 20.8 21.8 22.6

Notes: +(105), ⁄⁄⁄


p < 0.01, ⁄⁄
p < 0.05, ⁄p < 0.10, sum of weights = 1.9312  107.

Table 6
Attribute importance as share of implicit price ranges in percent.

Pooled model Domestic model Import model


Unweighted Rank Weighted Rank Unweighted Rank Weighted Rank Unweighted Rank Weighted Rank
Region of origin 54.0 1 47.4 1 30.0 1 27.8 1 42.3 1 31.4 1
Label type 10.6 2 15.2 2 13.6 4 15.2 3 8.5 6 12.7 4
Grape variety 8.6 5 12.9 3 26.1 2 26.2 2 9.0 5 20.7 3
Label colour 8.6 4 10.9 4 8.4 5 10.9 4 15.5 2 21.0 2
Label info 9.5 3 8.8 5 14.7 3 9.7 5 10.1 4 8.1 5
Closure 3.3 7 2.6 6 5.3 6 2.3 7 3.3 7 2.6 7
Bottle form 5.4 6 2.3 7 2.0 7 7.9 6 11.3 3 3.5 6
Sum packaging attributes 27.9 31.0 29.2 36.4 38.5 39.8
Sum other attributes 72.1 69.0 70.8 63.6 61.5 60.2

Note: Packaging attributes are label type, label colour, closure, bottle form. Other attributes are region of origin, grape variety and label info.

share of domestic US wines (67.3%) relative to imported wines be related to the better model fit for domestic wines. Less available
(32.7%), which is higher than the domestic share of available prod- label information on grape varieties for imported wines can be re-
ucts (61.1%). We are unable to identify the specific reasons for this lated to their lower explained price variance. It is interesting to ob-
price difference, which could be related to demand effects such as serve though that grape variety strongly increases in importance
consumer ethnocentrism (Sharma, Shimp, & Shin, 1995) and repu- when taking unit sales into account, suggesting that imported vari-
tation effects (Combris et al., 1997), or supply side factors, such as etal wines sell above average. This suggestion is also strengthened
differences in market access and production costs, considering that by the fact that model fit increases most predominantly through
competition and therefore substitution are not perfect. weighting for imported wines. In the weighted model domestic
Second, prices of domestic products can consistently be better and imported wines concur in comparable price premiums for pro-
explained by hedonic price models than the prices of imported duction method related and other front label information, but
wines (higher explained variance in Table 4). There are a number other information on the winery seems to be more valued for im-
of potential reasons for this, such as the closer geographical prox- ported wines.
imity allowing consumers to more easily discriminate between Finally, can we identify consistent price effects for packaging
familiar domestic regions and producers to better understand elements across domestic and imported products, which would
and adjust to local consumers’ preferences. In contrast, only some suggest more generalisable patterns? For all packaging attributes
of the imported wines were produced specifically for the US mar- all significant price effects consistently point in the same direction
ket but instead broadly target international and home market con- across domestic and imported models, suggesting congruent pric-
sumer preferences. ing patterns for domestic and imported wines.
Third, domestic and imported products differ considerably in
the relative importance attributes have in explaining price differ- 4.4. Congruency of implicit market prices with experimental findings
ences. Overall, packaging characteristics account for a larger share
of price range for imported products, but this difference reduces When comparing estimated implicit price effects to experimen-
when taking unit sales into account (Table 6). Contrasting to tal findings discussed in Section 1.1, differences in the cultural con-
domestic US wines, label colour has a higher importance for price text have to be taken into account. For the label design types in
differences in imported wines than label design type. This could be Table 5 we consistently observed high price premiums for nonde-
related to some traditional region specific European label styles, script labels followed by clean labels without highlights. This agrees
which in the past showed less variation than new world wines with US consumers’ perceptions of high quality reported by Orth
(Celhay & Passebois, 2011). There are fewer attribute importance and Malkewitz (2008). For the pooled and domestic models we ob-
differences between domestic and imported wines for other pack- served a positive but insignificant respectively marginally signifi-
aging characteristics. cant positive price premium for chateau labels, concurring with
Grape variety accounts for a quarter of the implicit price range US findings by Boudreaux and Palmer (2007). It is interesting to ob-
of domestic US wines but only for a tenth of that for imported serve that both highlighted versions of the clean and chateau style
wines. This agrees with our suggestion in Section 1.2 and can also labels, characterised by red and golden accents, were significantly
S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182 181

lower priced in our study, contrasting to findings for Chinese con- The fact that overall about 30% of wine price differences were
sumers by Convey (2011). Similarly, clean labels generally received related to packaging characteristics suggests a moderate overall
price premiums over chateau labels in our study, opposite to Chi- potential for packaging related marketing activities. But packaging
nese and Australian consumer preferences (Convey, 2011; Jarvis related price differences were generally higher for highly de-
et al., 2010; Mueller, Lockshin, et al., 2010). Delicate elegant design manded wine, suggesting that wine packaging can drive sales vol-
styles, associated by consumers in Orth and Malkewitz (2008) with umes to a larger degree than price. Marketers seem to be bound by
higher quality, did not result in significant price premiums in our price differences relating to intrinsic product characteristics region
study. We found a consistent price discount for animal graphic label of origin and grape variety, which are most likely determined by
styles across all models, agreeing with experimental findings by production cost factors.
Boudreaux and Palmer (2007), Jarvis et al. (2010) and associations
of inexpensiveness for contrasting designs in Orth and Malkewitz
4.6. Limitations and future research
(2008).
For label colour we observed consistent positive price premi-
Analysing market price differences does not allow conclusions
ums for black, followed by cream/grey colours, which were partic-
about causality. Although we identified a large degree of congru-
ularly high priced for imported products (Table 5). Uni-coloured
ence with experimentally measured consumer preferences, we
labels generally received a significant price discount, while there
cannot exclude that price differences are at least partially caused
was a tendency for warm multi-coloured labels to receive a modest
by producer and retailer decisions and market influence. Future re-
price premium. Our findings partially agree with Boudreaux and
search should further examine the underlying psychological pro-
Palmer (2007), who reported a positive impact on purchase intent
cesses related to these label colour and label style preferences
of US consumers by neutral (white, black) and warm colours. There
(Becker et al., 2011), providing indication of their potential stabil-
is also similarity to findings by Convey (2011) with the exception
ity. It seems particularly interesting to examine the degree to
of burgundy red, which seems to be a culturally determined highly
which these preferences are learned or psycho-physically hard-
preferred colour for Chinese consumers. Concordance of our study
wired. This would also provide valuable insights into the degree
with findings by Mueller, Lockshin et al. (2010) is mixed, who re-
to which preferences for different extrinsic cues are malleable or
ported white and yellow to be preferred over orange and grey by
permanent.
Australian consumers.
The stability of packaging related market price differences over
For bottle form our findings suggested a significant positive
time requires more research. It is unknown to what extend packag-
price premium for non-standard bottles in the weighted domestic
ing trends and design fashions, such as the critter-label wave for
and the unweighted import models. Mueller, Lockshin et al. (2010)
wine earlier this century, impact price differentials over time. We
only tested straight and round shaped bottles and their low re-
would also expect price premiums to be eroded if the majority of
ported attribute importance agrees with our findings (Table 6).
producers start to adopt these packaging elements. Potentially,
Cork closure only has a significant price premium in unweighted
price discounts for red and golden highlighted labels observed in
models (Table 5) and has a low importance (Table 6), partially dis-
this study can be explained by an erosion effect, when producers
agreeing with the stronger experimental closure influence experi-
started to increasingly use them. Long term studies would allow
mentally measured for US consumers (Marin & Durham, 2007).
an answer to this question.
The overall comparison indicates a large degree of congruence
Even though this study observed strong agreement in packaging
between estimated implicit prices and previous experimental find-
price differences for domestic and international products, further
ings for US consumers. Preference measurements in different cul-
research is required to address product related and cross-cultural
tural contexts show less agreement. Those studies measuring the
differences in the relative pricing importance and price premiums
relative importance of attributes (Jarvis et al., 2010; Mueller, Lock-
for different packaging elements. The research approach of this
shin et al., 2010) tended to overestimate the relative importance of
study to triangulate findings from stated and revealed preferences
packaging characteristics compared to our findings, most likely be-
seems to be promising for other food products and markets.
cause they only considered few origins and grape varieties in their
experiments, thereby under representing market variance.
5. Conclusion
4.5. Practical implications
This study used market transaction data to assess price premi-
Packaging characteristics of wine were found to be related to ums and discounts related to food packaging characteristics for red
market price differences and this differentiation followed similar wine. Significant price differences observed for food label designs
patterns across wines of different origins. Observed packaging and label colours suggest that products differ in their packaging
price premiums and discounts to a large degree agreed with find- characteristics across different price levels. This packaging price ef-
ings from previous experimental research in the US, suggesting fect was more pronounced for highly demanded than for niche
that the price differences are at least partially driven by consumer products. Imported wine, for which consumers have to rely on few-
preferences. Wine marketers can take advantage of this knowledge er intrinsic diagnostic cues, were more strongly price differentiated
by matching the packaging to the targeted price tier or by using by packaging elements than domestic products. The congruency
packaging to create consumer expectations of higher product va- we observed between packaging characteristics with price premi-
lue. This would suggest that for the US market higher priced wines ums across domestic and imported products suggest pricing pat-
should be furnished with nondescript and clean label styles in neu- terns that are fairly generalisable. When comparing implicit
tral colours. Other bottle forms and cork closures were also to market price premiums to previous experimentally measured sta-
some degree related to higher price levels. Wine labels using art- ted packaging preferences we observed some agreement within
work or animal graphics as well as contrasting highlights are rec- the same country but differences across divergent cultures. Results
ommended for lower price tiers. Providing production method or suggest that experimentally measured preferences for food pack-
other information on the front label was observed to be related aging are able to provide a first valid indication for later market
to price premiums. Although providing grape variety information price differences. The effect size of experimental findings seems
did not relate to market price, higher sales were observed for im- to be overestimated for product attributes with high market
ported wines of the main grape varieties. variance.
182 S. Mueller Loose, G. Szolnoki / Food Quality and Preference 25 (2012) 171–182

Acknowledgements Lange, C., Issanchou, S., & Combris, P. (2000). Expected versus experienced quality:
Trade-off with price. Food and Quality Preference, 11, 289–297.
Lange, C., Martin, C., Chabanet, C., Combris, P., & Issanchou, S. (2002). Impact of the
We thank AC Nielsen, particularly Danny Brag and Tom Greene information provided to consumers on their willingness to pay for Champagne:
from the Beverage Alcohol Group, for providing us access to the Comparison with hedonic scores. Food and Quality Preference, 13, 597–608.
Ling, B. H., & Lockshin, L. (2003). Components of wine prices for Australian wine:
scanner data sets. The authors acknowledge project funding from
How winery reputation, wine quality, region, vintage, and winery size
Australia’s grapegrowers and winemakers through their invest- contribute to the price of varietal wines. Australasian Marketing Journal (AMJ),
ment body the Grape and Wine Research Development Corpora- 11(3), 19–32.
Malhotra, N. K., Agarwal, J., & Peterson, M. (1996). Methodological issues in cross-
tion with matching funding from the Australian Government
cultural marketing research: A state-of-the-art review. International Marketing
(GWRDC project USA 06/01). Gergely Szolnoki’s research visits Review, 13(5), 7–43.
were financially supported by the Geisenheim Research Center Marin, A. B., & Durham, C. A. (2007). Effects of wine bottle closure type on consumer
and the School of Marketing at the University of South Australia. purchase intent and price expectation. American Journal of Enology and
Viticulture, 58(2), 192–201.
Mueller, S., Lockshin, L., & Louviere, J. (2010). What you see may not be what you
References get: Asking consumers what matters may not reflect what they choose.
Marketing Letters, 21(4), 335–350.
Aslam, M. M. (2006). Are you selling the right colour? A cross-cultural review of Mueller, S., Osidacz, P., Francis, I. L., & Lockshin, L. (2010). Combining discrete choice
colour as a marketing cue. Journal of Marketing Communications, 12(1), 15–30. and informed sensory testing in a two-stage process: Can it predict wine market
Becker, L., van Rompay, T. J. L., Schifferstein, H. N. J., & Galetzka, M. (2011). Tough share? Food Quality and Preference, 21(7), 741–754.
package, strong taste: The influence of packaging design on taste impressions Mueller, S., & Szolnoki, G. (2010). The relative influence of packaging, labelling,
and product evaluations. Food Quality and Preference, 22(1), 17–23. branding and sensory attributes on liking and purchase intent: Consumers
Bockstael, N. E., & MacConnell, K. E. (2007). Environmental valuation with revealed differ in their responsiveness. Food Quality and Preference, 21(7), 774–783.
preferences: A theoretical guide to empirical models. Netherland: Springer. Nerlove, M. (1995). Hedonic price functions and the measurement of preferences:
Boudreaux, C. A., & Palmer, S. (2007). A charming little Cabernet: Effects of wine The case of Swedish wine consumers. European Economic Review, 39(9),
label design on purchase intent and brand personality. International Journal of 1697–1716.
Wine Business Research, 19(3), 170–186. Norwood, F. B., & Lusk, J. L. (2011). Social desirability bias in real, hypothetical, and
Brentari, E., Levaggi, R., & Zuccolotto, P. (2011). Pricing strategies for Italian red inferred valuation experiments. American Journal of Agricultural Economics,
wine. Food Quality and Preference, 22(8), 725–732. 93(2), 528–534.
Cardebat, J.-M., & Figuet, J.-M. (2004). What explains Bordeaux wine prices? Applied Oczkowski, E. (1994). A hedonic price function for Australian premium table wine.
Economics Letters, 11(5), 293–296. Australian Journal of Agricultural Economics, 38, 93–110.
Celhay, F., & Passebois, J. (2011). Wine labelling: Is it time to break with tradition? A Oczkowski, E. (2001). Hedonic wine price functions and measurement error. The
study of the moderating role of perceived risk. International Journal of Wine Economic Record, 77(239), 374–382.
Business Research, 23(4), 318–337. Orth, U. R., & Malkewitz, K. (2008). Holistic package design and consumer brand
Combris, P., Bazoche, P., Giraud-Héraud, E., & Issanchou, S. (2009). Food choices: impressions. Journal of Marketing, 72(3), 64–81.
What do we learn from combining sensory and economic experiments? Food Panzone, L. A., & Simões, O. M. (2009). The importance of regional and local origin in
Quality and Preference, 20(8), 550–557. the choice of wine: Hedonic models of Portuguese wines in Portugal. Journal of
Combris, P., Lecocq, S., & Visser, M. (1997). Estimation of a hedonic price equation Wine Research, 20(1), 27–44.
for Bordeaux wine: Does quality matter? Economic Journal, 107, 390–402. Parsons, G. R. (1990). Hedonic prices and public goods: An argument for weighting
Combris, P., Lecocq, S., & Visser, M. (2000). Estimation of a hedonic price equation locational attributes by in hedonic regressions by lot size. Journal of Urban
for Burgundy wine. Applied Economics, 32(8), 961–967. Economics, 27, 308–321.
Convey, M. (2011). The impact of wine packaging on Chinese consumer choice. MBA Plassmann, H., O’Doherty, J., Shiv, B., & Rangel, A. (2008). Marketing actions can
thesis. Bordeaux School of Business (BEM). modulate neural representations of experienced pleasantness. Proceedings of the
Costanigro, M., McCluskey, J. J., & Mittelhammer, R. C. (2007). Segmenting the wine National Academy of Sciences, 105(3), 1050–1054.
market based on price. Hedonic regression when different prices mean different Rosen, S. (1974). Hedonic Prices and Implicit Markets: Product differentiation in
products. Journal of Agricultural Economics, 58(3), 454–466. pure competition. Journal of Political Economy, 82(1), 34–55.
Crozier, W. R. (1999). The meanings of colour: Preferences among hues. Pigment and Schamel, G., & Anderson, K. (2003). Wine quality and varietal, regional and winery
Resin Technology, 28(1), 6–14. reputations: Hedonic prices for Australia and New Zealand. Economic Record,
Delgado, C., & Guinard, J.-X. (2011). How do consumer hedonic ratings for extra 79(246), 357–369.
virgin olive oil relate to quality ratings by experts and descriptive analysis Schiefer, J., & Fischer, C. (2008). The gap between wine expert ratings and consumer
ratings? Food Quality and Preference, 22(2), 213–225. preferences: Measures, determinants and marketing implications. International
Deliza, R., & MacFie, H. J. H. (1996). The generation of sensory expectation by Journal of Wine Business Research, 20(4), 335–351.
external cues and its effect on sensory perception and hedonic ratings: A Sharma, S., Shimp, T. A., & Shin, J. (1995). Consumer ethnocentrism: A test of
review. Journal of Sensory Studies, 11(2), 103–128. antecedents and moderators. Journal of the Academy of Marketing Science, 23(1),
Gao, Z., & Schroeder, T. C. (2009). Effects of label information on consumer 26–37.
willingness-to-pay for food attributes. American Journal of Agricultural Siegrist, M., & Cousin, M. E. (2009). Expectations influence sensory experience in a
Economics, 91(3), 795–809. wine tasting. Appetite, 52(3), 762–765.
Goldstein, R., Almenberg, J., Dreber, A., Emerson, J. W., Herschkowitsch, A., & Katz, J. Stefani, G., Donato, R., & Cavicchi, A. (2006). Consumer expectations, liking and
(2008). Do more expensive wines taste better? Evidence from a large sample of willingness to pay for specialty foods: Do sensory characteristics tell the whole
blind tastings. Journal of Wine Economics, 3(1), 1–9. story? Food Quality and Preference, 17(1–2), 53–62.
Goodman (2009). An international comparison of retail consumer wine choice. Steiner, B. (2004). French wines on the decline? Econometric evidence from Britain.
International Journal of Wine Business Research, 21(1), 41–49. Journal of Agricultural Economics, 55(2), 267–288.
Grossman, R. P., & Wisenblit, J. Z. (1999). What we know about consumers’ color Thrane, C. (2004). In defence of the price hedonic model in wine research. Journal of
choices. Journal of Marketing Practice: Applied Marketing Science, 5(3), 78–88. Wine Research, 15(2), 123–134.
Jarvis, W., Mueller, S., & Chiong, K. (2010). A latent analysis of images and words in Waugh, F. V. (1928). Quality factors influencing vegetable prices. Journal of Farm
wine choice. Australasian Marketing Journal (AMJ), 18(3), 138–144. Economics, 10(2), 185–196.
Kamakura, W. A., Basuroy, S., & Boatwright, P. (2006). Is silence golden? An inquiry WineIntelligence. (2011). The secrets behind winning label designs. London: London
into the meaning of silence in professional product evaluations. Quantitative International Wine Fair (LIWF).
Marketing and Economics, 4(2), 119–141. Wei, S.-T., Ou, L.-C., Luo, M. R., & Hutchings, J. B. (2012). Optimisation of food
Lancaster, K. J. (1966). A new approach to consumer theory. Journal of Political expectations using product colour and appearance. Food Quality and Preference,
Economy, 74, 132–157. 23(1), 49–62.
Landon, S., & Smith, C. E. (1997). The use of quality and reputation indicators by Working, E. J. (1927). What do statistical ‘‘demand curves’’ show? The Quarterly
consumers: The case of Bordeaux wine. Journal of Consumer Policy, 20(3), Journal of Economics, 41(2), 212.
289–323.

S-ar putea să vă placă și