This article was downloaded by: [Laurentian University]On: 09 October 2014, At: 23:15Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Applied EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/raec20
Determinants of price elasticities for private labels andnational brands of cheeseMin-Hsin Huang a , Eugene Jones b & David E. Hahn ba Department of Asia-Pacific Industrial and Business Management , National University ofKaohsiung , 700, Kaohsiung University Road, Kaohsiung 811, Taiwanb Environmental and Development Economics, Department of Agricultural , The Ohio StateUniversity , 2120 Fyffe Road Columbus, Ohio 43210-1067Published online: 30 Oct 2009.
To cite this article: Min-Hsin Huang , Eugene Jones & David E. Hahn (2007) Determinants of price elasticities for privatelabels and national brands of cheese, Applied Economics, 39:5, 553-563, DOI: 10.1080/00036840500439069
To link to this article: http://dx.doi.org/10.1080/00036840500439069
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Applied Economics, 2007, 39, 553–563
Determinants of price elasticities for
private labels and national brands
of cheese
Min-Hsin Huanga, Eugene Jonesb,* and David E. Hahnb
aDepartment of Asia-Pacific Industrial and Business Management, National
University of Kaohsiung, 700, Kaohsiung University Road, Kaohsiung 811,
TaiwanbEnvironmental and Development Economics, Department of Agricultural,
The Ohio State University, 2120 Fyffe Road Columbus, Ohio 43210-1067
An Almost Ideal Demand System model is developed and used to estimate
price elasticities for US cheese sold at retail. Growing consumption of
cheese coupled with fierce competition between private labels and national
brands serves as motivating factors for this study. Per capita consumption
of cheese grew by 75% during 1980–2004 and private labels captured a
rising share of this growth. Private labels today account for 35% of market
share; national brands, for the remaining 65%. Kraft accounts for 45% of
national brands, but price increases for Kraft brands led to a sizeable price
gap between its brands and private labels. This gap helped to stimulate
growth of private labels. Marketing managers seek to capitalize on both
growing cheese sales and price gaps for brands. Relevant information for
marketing managers is consumer sensitivity to price changes. This study
uses 69 weeks of scanner data, with consumers segmented by income levels
to derive price elasticities for both lower-and higher-income consumers.
Results show lower-income consumers to be more price sensitive. If large
price gaps are maintained, the results suggest continued growth of private
labels. Yet, meta-analyses for this study suggest that Kraft could lower the
price gap and regain market share.
I. Introduction
An issue of paramount importance to the long-turn
growth and profitability of the US dairy industry is
the growing influence of private labels and/or store
brands of dairy products. These products have long
dominated fluid milk sales and, over the past two
decades, have gained prominence in the cheese, butter
and ice cream categories. This study focuses on the
dairy product category with the largest economic
value: cheese. Cheese overtook fluid milk as the
largest user of raw milk in the late 1990s and,
by 2001, it accounted for $19.56 billion in sales
(US Census Bureau, 2004).The retail cheese market is best characterized as an
amalgam of leading brands: store brands account for
35% of total market share; and national brands
account for the remaining 65%, but Kraft alone
*Corresponding author. E-mail: [email protected]
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online � 2007 Taylor & Francis 553http://www.tandf.co.uk/journalsDOI: 10.1080/00036840500439069
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
accounts for 45% of this total (Cropp, 2001).
Together, store brands and Kraft account for 80%
of the retail cheese market. Observations on prices
show that store brands have lower prices and this is
generally attributed to lower manufacturing, adver-
tising and overhead costs. As distributors and
retailers seek to expand cheese sales, they are
naturally interested in consumers’ sensitivity to
price changes for store and national brands.An understanding of factors influencing consumer
sensitivity to prices is critical to profitable production
and marketing decisions. Marketing managers at the
retail level must make pricing decisions for promo-
tional and nonpromotional brands at fairly regular
intervals. Critical to these decisions is the knowledge
of price elasticities and the interrelationship of these
elasticities among brands. As managers seek to
maximize cheese revenue and profit, they must (1)
identify products that have inelastic or near inelastic
demands and can sustain price increases; (2) select
products that can best stimulate total category sales;
and (3) identify appropriate discount levels that will
maximize sales while simultaneously providing
desired levels of profit.The marketing literature is ripe with studies
addressing determinants of price elasticities, and
most of these studies address broad product cate-
gories with considerable focus on alternative func-
tional forms. Examples include a study by Hoch
et al. (1995) in which the authors used a log-linear
function to estimate store-level elasticities for 18
product categories, including dairy products.
Similarly, a study by Mulhern, et al. (1998) used a
negative exponential function to estimate price
elasticities for liquor purchases across 35 stores.
These studies, with considerable emphasis on
functional forms, were less concerned about the
development of a comprehensive framework that
incorporated economics fundamentals (Baltas, 2002).
Further, emphasis on broad product categories,
such as soft drinks, paper towels and toothpaste,
increased the probability of aggregation bias in the
estimation procedures. This article avoids such
aggregation bias by focusing on a single product:
cheese. Specifically, cheese is disaggregated into
product classes by brands and package sizes to
estimate price elasticities at a narrow and more
refined level.The purpose of this study is to estimate brand
demand elasticities by using plausible economic
theory within a consumer demand framework. This
framework, provided in the next section, incorporates
factors that determine price elasticities for store and
national brands of cheese. Subsequent sections of this
article provide the empirical models, data, results andconclusions/implications.
II. Theoretical Framework
With neoclassic economic theory as its foundation,applied demand analysis, as formulated in this study,focuses on the optimal allocation of consumerexpenditure among different products and services.A demand system is derived from the utilitymaximization problem and its parameters are esti-mated on the basis of observations on price andexpenditure. Assuming weak separability of prefer-ences, a brand demand system is defined as a set ofdemand equations that determine the utility-maximizing allocation of category expenditureamong competing brands (Baltas, 2002).
Applied economists have utilized several econo-metric models or functional forms for estimatingconsumer demand (Gould, 1997; Gould and Dang,2000; Maynard, 2000; Maynard and Lin, 1999).A major goal of this study is to derive demandelasticities for store and national brands of cheese byestimating theoretically plausible demand systems.Furthermore, demographic and marketing effects areknown to impact brand level demand and thus aflexible functional form that incorporates demo-graphic and marketing variables is utilized. Fromthis viewpoint, the Almost Ideal Demand System(AIDS) (Deaton and Muellbauer, 1980) providesmore efficient estimates than other common demandsystems, such as Linear Expenditure System (LES)(Stone, 1954), Rotterdam model (Barten, 1964; Theil,1965) and Translog model (Christensen et al., 1975).
Another factor suggesting the AIDS model is thatit satisfies the axioms of choice exactly, therebyallowing for testing and for imposing homogeneityand symmetry conditions. Further, this model per-mits some forms of aggregation and it is mathe-matically integrable. Such desirable theoreticalproperties and flexibility of the AIDS model facilitatethe incorporation of marketing and demographicvariables into the model. In particular, the ‘priceindependent generalized log’ (PIGLOG) class ofexpenditure functions in the AIDS model fulfills theconditions required for exact nonlinear aggregation.That is, the share equations and the expenditurefunction derived from the AIDS model can be seen ascoming from a single representative household. Thus,the parameters of a household’s expenditure functioncan be (approximately) recovered even though theshare equations are estimated using aggregate data.This advantage of the AIDS model is extremelyimportant when using store-level, supermarket
554 M.-H. Huang et al.
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
scanner data, because demand equations in this studyrepresent retail-level market demand.
Utilizing these principles, the economic form of theAIDS budget share demand function for a cheeseproduct category can be written as:
wi ¼ �i þXj
�ij log pj þ �i logx
P
� �ð1Þ
where wi is average expenditure share for a specificproduct class i (e.g. store brands of small packagesize); �i,�i, gij, are parameters of the system;
x ¼Xni¼1
piqi
is total expenditure in the cheese product category; pjrepresent the price of the jth product class; pi and qirepresent the price and quantity, respectively, of theith product class; and P is a price index defined as
logP ¼ �0 þXk
�k log pk þ1
2
Xk
Xj
�kj log pk log pj
ð2Þ
The parameter restrictions derived from utility theoryrequire that the following conditions be satisfied:
Xni¼1
�i ¼ 1;Xni¼1
�ij ¼ 0;Xni¼1
�i ¼ 0 ðadding-upÞ
ð3ÞXj
�ij ¼ 0 ðhomogeneityÞ ð4Þ
�ij ¼ �ji ðsymmetryÞ ð5Þ
It should be noted that within the same brand,small and large package sizes provide different levelsof utility to consumers. For example, Folkes et al.(1993) showed that large packages, as compared tosmall ones, encourage greater use because consumersare less concerned about running out of the product.The greater the supply of a product (e.g. largepackage), the lower the transaction (replacement)costs for using the product and greater the volumepeople are willing to use (Lynn 1992). Wansink (1996)further indicated that packaging influences purchasebehaviour and usage behaviour. Building on thesetheoretical arguments, small and large package sizesare incorporated into a brand demand system.
III. Estimation Procedure
The estimation procedure involves a two-stagemodelling process. First, the empirical AIDS
model is derived and used to estimate price
elasticities. Second, a meta-analysis procedure that
uses price elasticities as starting data points is used
to estimate the impact of independent factors on
these elasticities. That is, estimated price elasticities
from the AIDS model for both store and national
brands of cheese are regressed on a set of
independent factors.
The empirical AIDS model
As currently expressed, Equation 1 is void of
marketing variables. Given the influence of market-
ing activity on consumer shopping behaviour, it is
natural to extend the AIDS model to incorporate
these marketing variables. This study employs the
linear demographic translating method to incorpo-
rate marketing variables as discussed in Pollak and
Wales (1978, 1980). That is, the intercept term, �i inEquation 1, is assumed to be a linear function
of marketing attributes such as price promotion,
customer counts, holidays and seasonal effects.
More specifically,
�i ¼ ��i þXj
�ijPRþ �iCCþ �iHDþX3k
�ikSE ð6Þ
where ��i is the intercept net of marketing effects; PR
is price promotion, representing the number of items
on price discount within a product class during a
given week; CC is customer count that is the number
of customers shopping each week and this variable is
specified to capture the effect of store traffic on
particular product sales; HD is a zero-one dummy
variable that captures the effect of calendar holidays;
and SE is a seasonal effect, expressed as quarterly
dummies to capture seasonal effects of cheese
purchase. Substituting Equation 6 into the AIDS
model presented in Equation 1, the empirical AIDS
model incorporating marketing variables used in this
study can be derived as
wi ¼��i þ
Xj
�ijPRþ �iCCþ�iHDþX3k
�ikSE
þXj
�ij logpjþ�i
� logx��0�X
��i þ
Xj
�ijPRþ �iCCþ�iHD
þX3k
�ikSE
�� logpk�
1
2
Xk
Xj
�kj logpk logpj
!
ð7Þ
Determinants of price elasticities 555
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
where w denotes expenditure share, p represents price,
x is total expenditure in the cheese product category;
�,�, �, �, �, �, �, are model parameters to be estimated.Theoretical restrictions placed on the parameters
can be summarized as
Xni¼1
��i ¼ 1;
Xni¼1
�ij ¼ 0;Xni¼1
�i ¼ 0
Xni¼1
�ij ¼ 0;Xni¼1
�i ¼ 0
Xni¼1
�i ¼ 0;Xni¼1
�ik ¼ 0 ðadding-upÞ
ð8Þ
Xj
�ij ¼ 0 ðhomogeneityÞ ð9Þ
�ij ¼ �ji ðsymmetryÞ ð10Þ
Once the parameters have been estimated, the price
elasticities can be calculated as follows (Green and
Alston, 1990):
eii ¼ �1þ�ij � �ið�i þ
Pj �ij log pjÞ
wi
� ðown-price elasticityÞ ð11Þ
eij ¼�ij � �ið�j þ
Pj �ij log pjÞ
wi
� ðcross-price elasticityÞ ð12Þ
The linear approximation AIDS (LA/AIDS) model
is also estimated to compare the results with those
of the original nonlinear AIDS model. The linear
approximation AIDS model involves the replacement
of logP with a simpler index as suggested by
Moschini (1995). That is, the simpler price index
can be represented as
lnP� ¼Xni¼1
w0i lnðpitÞ ð13Þ
where w0i is the expenditure share of good i in the base
period. Consequently, the empirical LA/AIDS model
used in this study can be represented as
wi ¼ ��i þXj
�ijPRþ �iCCþ �iHD
þX3k
�ikSEþXj
�ij log pj þ �i logx
P�
� �ð14Þ
and the price elasticities can be calculated as follows
eii ¼ �1þ�iiwi
� �� �i ðown-price elasticityÞ ð15Þ
eij ¼�ijwi
� ��
�i
wi
� �wj ðcross-price elasticityÞ ð16Þ
The literature supports the hypothesis that higher-and lower-income consumers exhibit different shop-ping behaviour and sensitivity to price changes (e.g.Jones and Mustiful, 1996; Mulhern et al., 1998).Anticipating different behaviour across products andincome groups, cheese is classified into five productcategories based on product forms: shredded, sliced,chunk, snack and miscellaneous. Given five productcategories and two income groups, a total of 20demand systems are estimated for the two AIDSmodels (Equations 7 and 14).
IV. Data Discussion and Issues
The data used in this study are store-level scannerdata provided by a national supermarket chain in theColumbus, Ohio Metropolitan Area (COMA). Thedata represent weekly observations and they includeUniversal Product Codes (UPCs), prices, packagesizes and dollar and quantity sales. These datarepresent 69 weeks of consumer purchases, coveringthe period 31 December 2000 to 21 April 2002.
Six stores are included in the data set, and thesestores represent two broadly defined store groups:higher-and lower-income stores. These groups areidentified from socioeconomic information providedby the chain for all residents within a 3-mile radius ofeach store. The lower-income stores (1, 2 and 3) arelocated in areas that have large proportions of lower-income shoppers, while the higher-income stores (4, 5and 6) are located in areas that have large propor-tions of higher-income shoppers. The three lower-income stores are within the inner city of Columbusand the three higher-income stores are located inColumbus suburbs. As shown in Table 1, an averageof 4.2% of the residents in higher-income areas hashousehold incomes less than $10 000 annually. At theopposite end of the income spectrum, an averageof 34.1% of residents in higher-income areas haveannual household incomes exceeding $75 000.In contrast, 12% of lower-income residents havehousehold incomes below $10 000 and just 11.8%have incomes above $75 000.
Differences between the two groups exist not onlyfor income, but also for education and race. Asshown in Table 1, only 10% of the prospectiveshoppers in areas identified as lower-income stores(stores 1, 2 and 3) are college graduates, as comparedto 38% of prospective shoppers in higher-incomeareas (stores 4, 5 and 6). The lower-income areas arealso characterized by populations that are moreheterogeneous than the homogeneous populationsof the higher-income areas.
556 M.-H. Huang et al.
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
Stationarity is an important issue when usingtime series data in econometric analysis. Thetraditional test for stationarity of time-series datais called the unit-root test. In the current study, theAugmented Dickey–Fuller (ADF) tests are con-ducted to identify whether each individual time-series variable in the AIDS model is stationary ornonstationary. The results of ADF test indicatethat absolute values of the estimated t-statistics arelarger than corresponding asymptotic critical valuesfor most variables in the AIDS model, with someexceptions. In other words, the time-series variablesused in the present study are stationary, and onecan reject the null hypothesis that the data containa unit root.
V. Results of AIDS Models
A total of 20 demand systems are estimated for thenonlinear original AIDS as well as for the linearapproximation AIDS models. These 20 demandsystems led to a total of 84 equations and 964parameters, and therefore these numerous para-meters are not reported in this article. Derivedelasticities provide interesting perspectives onconsumer behaviour, and therefore own-price elas-ticities from both the linear and nonlinear AIDSmodels for store and national brands of cheese aresummarized in Table 2. These are Marshallian(uncompensated) elasticities and they are shown by
store location, product category and package size.Hicksian or compensated elasticities are generallysmaller (absolute value) than uncompensated elasti-cities and that pattern is observed for the estimatesof this study.
As economic theory would predict, shoppersresiding in lower-income areas are shown to bemore price sensitive than those residing in higher-income areas. Shoppers of the lower-income storesare shown to be more price sensitive for all products,except store brands of sliced cheese, national brandsof sliced cheese (large package size) and nationalbrands of chunk cheese (small package size).Of course, it must be recognized that the lower-andhigher-income stores of this study are not patronizedsolely by lower- or higher-income shoppers. Yet,income differences are sharp enough to warrant thedescribed classification. Indeed 24 of the 32 own-priceelasticities for each of the AIDS models in Table 2suggest that the income classification is appropriateand meaningful.
Since cheese purchases represent a major expendi-ture and national brands of cheese have higher pricesthan store brands, economic theory would predicthigher price sensitivity for these brands on the part oflower-income consumers. Yet, given the large andoften sustained advertising campaigns for nationalbrands, theory would also predict lower pricesensitivity for national brands relative to storebrands. Indeed one of the objectives of brandadvertising is to diminish consumers’ price sensitivityfor national brands relative to store and/or other
Table 1. Household demographic data for six stores (by percentage)
Lower-income stores Higher-income stores
Demographic informationa Store 1 Store 2 Store 3 Average Store 4 Store 5 Store 6 Average
Household incomeUnder $10 000 13.8 12.9 9.3 12.0 3.8 5.0 3.8 4.2$10 000–$49 999 57.6 58.3 54.1 56.7 32.8 41.8 37.7 37.4$50 000–$74 999 18.5 18.2 22.4 19.7 27.4 20.9 24.6 24.3$75 000–$99 999 6.5 6.3 8.4 7.1 17.5 12.1 15.3 15.0$100 000þ 3.8 4.3 5.9 4.7 18.8 20.2 18.2 19.1
RaceWhite 59.2 83.6 85.7 76.2 95.4 92.4 93.1 93.6Black 38.6 14.4 12.1 21.7 2.3 3.2 5.0 3.5Others 2.1 2.0 1.8 2.0 2.6 4.6 1.9 3.0
EducationGrade school 7.3 10.0 11.1 9.5 4.1 2.0 2.5 2.9Some high school 21.3 25.4 25.8 24.2 11.6 5.0 8.6 8.4High school gradate 33.5 36.7 37.6 35.9 28.2 16.2 27.0 23.8Some college 24.3 19.2 17.8 20.4 26.2 26.6 28.2 27.0College graduate 13.8 8.8 7.5 10.0 29.9 50.6 33.5 38.0
Source: Spectra, 2001.Note: aNumbers may not add to 100.0 because of rounding.
Determinants of price elasticities 557
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
competing brands. Advertising expenditures wereunavailable for this study, so an advertising proxyof price promotion (observed price reductions) wasused. Price promotions are registered immediately inconsumer purchases and therefore these capturedeffects are short-term and may be larger than those ofadvertising expenditures that seek to generate positiveand long-term product differentiation. Stated differ-ently, unlike the short-term elasticities estimated inthis study, long-term price elasticities would likelycapture the effects of advertising expenditures andshow lower-income shoppers to be more pricesensitive than higher-income shoppers for all productclasses. Further, relative to store brands, they wouldlikely show national brands to be less price-sensitiveto a larger extent than what is shown in this study.
Although contrary to economic expectation, theempirical results of Table 2 show both lower- andhigher-income consumers to be more price sensitivefor national brands than for store brands. Thispattern is observed for five of the eight productclasses from each set of AIDS results. Again, withprice promotion being a proactive tool to moveproduct through the marketing channels, these
elasticities are likely to reflect the extent to whichconsumers have responded to price promotions fornational brands. Observations on price promotionsshow that promotion for branded products is mosteffective when the price gap between national andstore brands falls to consumers’ threshold purchaselevel. Indeed studies show that a promotion price fora national brand that is equal to that of a store brandleads to practically zero sales for the store brand(Wedel and Zhang, 2004). Such price promotionswould not continue indefinitely if media advertising isalso used to convey persuasive messages about thecharacteristics of national brands. Thus, the own-price elasticities of this study should be viewed asshort-term elasticities that show consumer reactionsto price promotions.
Although the results of Table 2 show that lower-income consumers have greater price sensitivity thanhigher-income ones, and that national brands havehigher elasticities than store brands, the results inTable 3 are intended to provide a comparison of allstore brands elasticities with those of national brands.The un-weighted values of both the AIDS and LA/AIDS models show that national brands not only
Table 2. The estimated own-price elasticities of store brands and national brands*
Lower-income stores Higher-income stores
Store brands National brands Store brands National brands
AIDS model
Shredded cheeseSmall sizesa �1.655 �3.453 �1.520 �2.704Large sizes �2.165 �1.865 �1.653 �1.573
Sliced cheeseSmall sizesb �1.935 �3.454 �1.988 �2.114Large sizes �1.778 �3.049 �2.943 �3.149
Chunk cheeseSmall sizesa �2.568 �1.235 �2.306 �2.275Large sizes �2.014 �2.331 �1.874 �2.232
Snack cheese �2.397 �1.530 �2.161 �0.538Miscellaneous cheese �1.350 �1.979 �0.526 �1.649
LA/AIDS model
Shredded cheeseSmall sizes �1.691 �3.407 �1.508 �2.686
Large sizes �2.170 �1.764 �1.601 �1.426Sliced cheese
Small sizes �1.851 �3.470 �1.901 �2.070
Large sizes �1.700 �3.011 �2.936 �3.137
Chunk cheeseSmall sizes �2.452 �1.187 �2.182 �2.695Large sizes �2.121 �2.341 �1.837 �2.326
Snack cheese �2.256 �1.262 �1.981 �0.363Miscellaneous cheese �1.477 �1.956 �0.553 �1.704
Notes: aPackage size at 8-oz or below; bPackage size at 12-oz or below.*All own-price elasticities are derived from statistically significant parameters as shown in Equation 11. Each elasticity isstatistically significant at the 0.10 level or better. Numbers in bold show national brands to have more elastic demands thanprivate labels.
558 M.-H. Huang et al.
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
have higher elasticities, but they have a wider range.As shown, own-price elasticities for national brandsrange from �0.36 to �3.47; those for store brandsrange from �0.53 to �2.94. These results suggest thatnational brands of cheese are quite responsive to pricechanges and that revenue maximization is bestachieved through price reductions.
The aforementioned elasticities shed some insighton the price competition struggle between Kraft andmakers of store brands of cheese. This struggle hasbeen addressed in several Wall Street Journal articlesas well as other publications. These sources include:‘Kraft Profit Misses Expectations, Hurt by Private-Label Brands’ (17 July 2003); ‘Kraft Loses 2 TopExecutives Amid Private-Label Struggle’ (10 July2003); ‘Food for Thought: Why Kraft Is StillFrowning’ (16 April 2003); ‘Kraft’s Private-LabelLesson’ in Reveries Magazine (Hoyt, 2004); and‘Commodity Prices Strain Marketing’ in AdvertisingAge (Thompson, 2004). Although much of theliterature (e.g. Bushman, 1993; Richardson et al.,1994) support the premise that store brands are lowerin quality than national brands, the price responsesof consumers shown in Table 3 call this premise intoquestion. This question is especially relevant whenconsidered against the evidence of large marketinginvestments to build brand loyalty. Brand loyaltycommits consumers to paying premium pricesfor national brands, and therefore this loyalty leadsto lower price sensitivity for national brands.Store brands are generally hypothesized to havefairly elastic price responses and therefore theresults of Table 3 do not support the theoreticalarguments of consumer demand. Of course, itmust be recognized that 71% of supermarketshoppers now consider private-labels to be of equalor better quality than national brands (Hoyt, 2004).Further, marketing analysts agree that Kraft allowedthe price gap between its brands of cheese
and private-label brands to grow too wide(Thompson, 2004).
VI. A Meta-Analysis
Meta-analysis is a regression technique that allowsfor comparison of similar, but not necessarilyidentical quantity estimates. These estimates aremost often derived from different empirical modelsand meta-analysis provides a way for synthesizing
and interpreting these research results (Farely andLehmann, 1986). Meta-analysis has been developedand widely used in the context of the social sciencessuch as marketing, economics and psychology andrefers to the statistical analysis of empirical research
results (Stanley, 2001). Meta-analysis has also beenestablished to synthesize empirical research results bymeans of an analysis of the variation in estimateddemand elasticities. For example, Tellis (1988)described a meta-analysis of econometric studies
that estimated the elasticity of selective sales ormarket share to price. Dalhuisen et al. (2003)presented a meta-analysis of variations in price andincome elasticities of residential water demand.
In the current study, meta-analysis is used tosynthesize the factors that determine the estimatedstore-level price elasticities for national brands and
store brands of cheese. More specifically, the meta-regression of price elasticity for store brands andnational brands, respectively can be written as:
Elasticity ¼ �0 þ �1 Shareþ �2 LowIncome
þ �3 ShredCheeseþ �4 SlicedCheese
þ �5 SnackCheeseþ �6 MiscelCheese
þ �7 SmallSizeþ �8 AIDSþ " ð17Þ
Table 3. The descriptive statistics of estimated own-price elasticities for store brands and national
brands
BrandsNumber ofobservations Mean SD
Minimumof magnitude
Maximumof magnitude
AIDSStore brands 16 �1.927 0.552 �0.526 �2.943National brands 16 �2.196 0.818 �0.538 �3.454
LA/AIDSStore brands 16 �1.889 0.519 �0.553 �2.936National brands 16 �2.175 0.876 �0.363 �3.470
POOLEDStore brands 32 �1.908 0.527 �0.526 �2.943National brands 32 �2.185 0.834 �0.363 �3.470
Determinants of price elasticities 559
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
where:
Elasticity is the absolute value of own-priceelasticity for a specific productbrand;
Share is the market share of the specificproduct brand;
LowIncome is 1 for lower-income stores; 0otherwise;
ShredCheese is 1 if shredded cheese; 0 otherwise;SlicedCheese is 1 if sliced cheese; 0 otherwise;SnackCheese is 1 if snack cheese; 0 otherwise;MiscelCheese is 1 if miscellaneous cheese; 0
otherwise;SmallSize is 1 if small package size; 0 otherwise;
AIDS is 1 if elasticity i is estimated bynonlinear AIDS model; 0 otherwise;and �’s are parameters to beestimated.
VII. Results of Meta-Analysis
The relatively large price elasticities shown inTables 2 and 3 together with the substantialvariability among these elasticities suggest a need toassess factors which are most influential in theirdetermination. This study estimates two meta-regres-sions. One set of regression involves the own-priceelasticities for store-brands of cheese; the otherinvolves own-price elasticities for national-brands ofcheese. The two-stage modelling process is utilized inthe current research. More specifically, in the firststage, the own-price elasticities for store brands andnational brands are estimated by AIDS models(original nonlinear AIDS and linear approximationAIDS). In the second stage, the estimated own-priceelasticities are regressed on the determinant factors.
This meta-analysis is used to reveal the factors thataffect store-level price elasticities of store brands andnational brands of cheese. For instance, �2 representsthe difference in price elasticity of store brands(national brands) cheese associated with a changefrom the higher-income group to the lower-incomegroup. A test of the null hypothesis that �2¼ 0provides a test of the hypothesis that there is nodifference between the price elasticity of store brands(national brands) of cheese associated with thehigher-income group and that associated with thelower-income group.
The meta-regression results for store brands areprovided in Table 4. The R2 and adjusted R2 are 62.3and 49.2%, respectively. The goodness of fit measuresshow reasonably good performance for the model.
It should be noted here that the dependent variable,
own-price elasticity, is expressed as an absolute value,
as the interest lies in determining the impact of
independent factors on the magnitude of change for
cheese price elasticities.Interpreting the estimated parameters, market
share has a negative impact on own-price elasticity
for store brands, but this coefficient is statistically
insignificant. This suggests that higher market shares
have not reduced the price elasticity for store brands.
In addition, lower-income stores, as compared to
higher-income ones, have larger own-price elasticities,
although statistically this difference is not significant.
Two parameter estimates that are statistically sig-
nificant are: shredded cheese and miscellaneous
cheese. Relative to the base of chunk cheese, these
two classes of cheese decrease the price elasticity of
store brand cheese.The results of meta-regression for national brands
are reported in Table 5. The R2 and adjusted R2 are
69.2 and 58.5%, respectively, which indicates that the
model explains a great deal of the variation in price
elasticities. Market share has a negative and statisti-
cally significant impact on the magnitude of price
elasticities for national brands. In other words, large
market shares convey market power and serve to
diminish consumers’ price sensitivity. With respect to
store location, lower-income stores have a higher
price elasticity for national brands, and this difference
is statistically significant with chunk cheese as the
base category. The results show that snack cheese
serves to decrease price elasticity for national brands.
In contrast, sliced and miscellaneous cheeses serve to
Table 4. Meta-analysis parameter estimates for
determinants of store brands price elasticities
VariableRegressioncoefficient SE Prob.> jtj
Intercept 2.447 0.321 0.000Share �1.190 1.014 0.253LowIncome 0.193 0.143 0.190HighIncome (base)ShredCheese �0.424 0.188 0.034SlicedCheese �0.132 0.204 0.523SnackCheese 0.295 0.303 0.341MiscelCheese �1.323 0.252 0.000ChunkCheese (base)SmallSize 0.005 0.179 0.977LargeSize (base)AIDS 0.038 0.133 0.775LA/AIDS (base)N 32Model F-statistic 4.75 0.002R-square 0.623Adjusted R-square 0.492
560 M.-H. Huang et al.
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
increase price elasticity for national brands. Relative
to package size, small package sizes have a
positive impact on the price elasticity for national
brands. Results in both Tables 4 and 5 show no
statistically significantly difference between the
LA/AIDS and AIDS model for store or national
brands of cheese.Results from the meta-regressions for both store
and national brands provide informative information
for marketing managers in the cheese industry. First,
as discussed in the empirical results section, lower-
income shoppers are more price-sensitive than higher-
income shoppers and these meta-analysis results
provide additional information for examining differ-
ences in their shopping behaviour. For national
brands of cheese, lower-income shoppers are shown
to be considerably more price sensitive, as measured
by a statistically significant coefficient of 0.352. That
is, relative to the base of higher-income stores, the
lower-income stores significantly increase the magni-
tude of the own-price elasticity for national brands.
Similar results are also revealed for store brands.
That is, relative to the base of higher-income stores,
lower-income stores are shown to increase the
magnitude of own-price elasticity, although the
coefficient is not statistically significant. Differences
in consumers’ price sensitivity for store locations
(e.g. lower-and higher-income) suggest that retail
managers could utilize micro-marketing strategies
with different prices across stores or retail market
areas. That is, these findings suggest that there are
possible advantages to having store- or area-specific
pricing. A retailer might be able to set prices in a
more profit-maximizing manner by matching
prices to customers’ price sensitivity within a given
store or location.A second piece of information for marketing
managers is derived from the observation that
brands with higher market shares have lower levels
of price sensitivity. Meta-regression results show
market share coefficients to be negative for both
national and store brands, although estimates for
store brands are not statistically significant. Yet, the
negative signs lend support to the economic principle
that market share leads to market power and there-
fore less elastic price responses. For national brand
manufacturers, these findings suggest that consumer
price sensitivity might be reduced by adopting an
explicit strategy of increasing market share. For
manufacturers of store brands, this research suggests
that store brands can be competitive products,
particularly after they build up store loyalty. In the
long run, store brands are likely to experience growth
and narrow the price gap between store and national
brands and capture a larger percentage of store
profits.A third piece of informative information relates to
the fact that price elasticities for both store and
national brands of cheese vary by product types.
Meta-regression results for national brands indicate
that, relative to the base of chunk cheeses, snack
cheese has a coefficient of �0.897 and this measure is
statistically significant. That is, holding the sales
of chunk cheese constant, an increase in the sales of
snack cheese will reduce the price-sensitivity of
national brands of cheese. Stated differently, con-
sumers are less sensitive to price changes for national
brands and the coefficient estimate for snack
cheese suggests a willingness to pay premium prices
for value-added products, such as snack cheeses.
A plausible explanation for this observation is that
consumers may associate higher-levels of product
quality with some value-added products such as
snack cheese, but may associate lower-levels of
product quality with other value-added products.
These research findings suggest that manufacturers of
national brands can implement premium price
policies for some product categories, but utilize
competitive price policies for other product cate-
gories. For store brands, the meta-regression results
show that, relative to the base of chunk cheeses,
shredded and miscellaneous cheese significantly
decrease the magnitude of own-price elasticities.
This suggests that store managers may wish to
focus much of their advertising, merchandising and
promotion efforts on increasing the sales of shredded
and miscellaneous cheese.
Table 5. Meta-analysis parameter estimates for determi-
nants of national brands price elasticities
VariableRegressioncoefficient SE Prob.> jtj
Intercept 2.971 0.485 0.000Share �10.568 3.767 0.010LowIncome 0.352 0.191 0.079HighIncome (base)ShredCheese �0.325 0.345 0.356SlicedCheese 1.518 0.358 0.000SnackCheese �0.897 0.370 0.024MiscelCheese 1.364 0.641 0.044ChunkCheese (base)SmallSize 1.115 0.390 0.009LargeSize (base)AIDS 0.020 0.190 0.916LA/AIDS (base)N 32Model F-statistic 6.46 0.000R-square 0.692Adjusted R-square 0.585
Determinants of price elasticities 561
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
VIII. Conclusions and Implications
Marketing managers are very interested in gainingan understanding of consumers’ sensitivity to pricechanges. The most prevalent measure of consumerprice sensitivity is the price elasticity of demand,which represents the percentage change in quantitysold for a given percentage change in price. Byutilizing retailer-supplied scanner data, this studydeveloped and estimated a theoretically plausiblemodel to provide price elasticities for specific cheeseproducts. These elasticities were then used in a meta-regression model to reveal factors that determine priceelasticities for store and national brands of cheese.The findings of this study provide a rich knowledgebase for retail store managers and manufacturersto use for maximizing sales and profits.
With the results showing lower-income consumersto be more price sensitive, cheese manufacturers anddistributors may wish to provide retailers a differentset of price incentives for lower-income stores thanthose provided for higher-income ones. That is, agiven price reduction in lower-income stores is likelyto lead to a greater quantity response than the sameprice reduction would generate in higher-incomestores. While the meta-analyses show that marketshare gains tend to decrease consumer price elasticity,the higher elasticities for national brands relative toprivate labels suggest that Kraft, in the process ofgaining larger market shares, may have increased theprice gap between the brands beyond consumers’tolerance level for brand loyalty. In essence, alowering of the price gap between private labels andKraft’s brands could diminish price elasticities fornational brands to the extent that Kraft’s brandsbecome far more competitive with private labels.With Kraft having such a large share of the cheesemarket, its response to competition from privatelabels will determine the ultimate growth andstructure of the cheese market. If Kraft lowers theprice gap between the national and private-labelbrands and adopts a long-run strategy of regainingmarket share through advertising and competitivepricing, growth of private labels is likely to bearrested. In contrast, if the company focuses onmaximizing its short-run profits by relying onpurchases from those loyal to its brands, privatelabels are likely to continue their encroachment onthe market shares of Kraft and other brands.
References
Baltas, G. (2002) An Applied analysis of brand demandstructure, Applied Economics, 34, 1171–5.
Barten, A. P. (1964) Consumer demand functions underconditions of almost additive preferences,Econometrica, 32 , 1–38.
Bushman, B. J. (1993) What’s in a name? The moderatingrole of public self consciousness on the relationbetween brand label and brand preference, Journal ofApplied Psychology, 78, 857–61.
Christensen, L. R., Jorgenson, D. W. and Lau, L. J. (1975)Transcendental logarithmic utility functions, AmericanEconomic Review, 65, 367–83.
Cropp, R. (2001) Retail distribution of dairy products.Department of Agricultural and Applied Economics,University of Wisconsin-Madison.
Dalhuisen, J. M., et al. (2003) Price and income elasticitiesof residential water demand: a meta-analysis, LandEconomics, 79, 292–308.
Deaton, A. and Muellbauer, J. (1980) An almost idealdemand system, American Economic Review, 70,312–26.
Farely, J. U. and Lehmann, D. R. (1986) Meta-Analysis inMarketing, Lexington Books, Lexington, MA.
Folkes, V., Martin, I. and Gupta, K. (1993) When to saywhen: effects of supply on usage, Journal of ConsumerResearch, 20, 467–77.
Gould, B. W. (1997) Consumer promotion and purchasetiming: the case of cheese, Applied Economics, 29,445–57.
Gould, B. W. and Dong, D. (2000) The decision of when tobuy a frequently purchased good: a multi-periodprobit model, Journal of Agricultural and ResourceEconomics, 25, 636–52.
Green, R. and Alston, J. M. (1990) Elasticities in AIDSmodels, American Journal of Agricultural Economics,72 , 442–45.
Hoch, S. J., et al. (1995) Determinants of store-level priceelasticity, Journal of Marketing Research, XXXII,17–29.
Hoyt, C. W. (2004) Kraft’s private-label lesson, ReveriesMagazine, February.
Jones, E. and Mustiful, B. W. (1996) Purchasingbehavior of higher-and lower-income shoppers: alook at breakfast cereals, Applied Economics, 28,131–7.
Lynn, M. (1992) Scarcity’s enhancement of desirability: therole of naive economic theories, Basic and AppliedSocial Psychology, 13, 67–78.
Maynard, L. J. (2000) Sources of irreversible consumerdemand in U.S. dairy products, in AmericanAgricultural Economics Association Annual Meetings,Tampa, Fl.
Maynard, L. J. and Liu, D. (1999) Fragility in dairyproduct demand analysis, in American AgriculturalEconomics Association Annual Meetings, Nashville,TN.
Moschini, G. (1995) Units of measurement and the stoneprice index in demand system estimation, AmericanJournal of Agricultural Economics, 77, 63–8.
Mulhern, F. J., Williams, J. D. and Leone, R. P. (1998)Variability of brand price elasticities across retailstores: ethnic, income, and brand determinants,Journal of Retailing, 74, 427–46.
Pollak, R. A. and Wales, T. J. (1978) Estimation ofcomplete demand system from household budgetdata: the linear and quadratic expenditure systems,American Economic Review, 68, 348–59.
562 M.-H. Huang et al.
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4
Pollak, R. A. and Wales, T. J. (1980) Comparison ofquadratic expenditure system and translog demandsystems with alternative specifications of demographiceffects, Econometrica, 48, 595–612.
Richardson, P. S., Dick, A. S. and Jain, A. K. (1994)Extrinsic and intrinsic cue effects on perceptionsof store brand quality, Journal of Marketing, 58,28–36.
Stanley, T. D. (2001) Wheat from chaff: meta-analysis asquantative literature review, Journal of EconomicPerspectives, 15, 131–50.
Stone, J. R. (1954) Linear expenditure systems and demandanalysis: an application to the pattern of Britishdemand, Economic Journal, 64, 511–27.
Tellis, G. J. (1988) The price elasticity of selective demand:a meta-analysis of econometric models of sales,Journal of Marketing Research, XXV, 331–41.
Theil, H. (1965) The information approach to demandanalysis, Econometrica, 33, 67–87.
Thompson, S. (2004) Commodity prices strain marketing,Advertising Age, 75, 4–5.
US Census Bureau, Statistical Abstract of the UnitedStates, US Government Printing Office, 2004.
Wansink, B. (1996) Can package size accelerate usagevolume?, Journal of Marketing, 60, 1–14.
Wedel, M. and Zhang, J. (2004) Analyzing brand competi-tion across subcategories, Journal of MarketingResearch, 41, 448–56.
Determinants of price elasticities 563
Dow
nloa
ded
by [
Lau
rent
ian
Uni
vers
ity]
at 2
3:15
09
Oct
ober
201
4