Decision Support Systems 49 (2010) 110–119

Contents lists available at ScienceDirect

Decision Support Systems

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / d s s

Retail pricing decisions and product category competitive structure

Óscar González-Benito a,⁎, María Pilar Martínez-Ruiz b,1, Alejandro Mollá-Descals c,2
a
Dpto. Administración y Economía de la Empresa, Universidad de Salamanca, Campus Miguel de Unamuno, 37007 – Salamanca, Spain
b
Área de Comercialización e Investigación de Mercados, Universidad de Castilla-La Mancha, Avenida de los Alfares, 44, 16002 – Cuenca, Spain
c
Dpto. Comercialización e Investigación de Mercados, Universidad de Valencia, Avenida de los Naranjos, s/n, 46022 – Valencia, Spain

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

Article history: This study addresses the use of demand forecasting techniques by retailers to support their decision making.
Received 15 September 2008 Specifically, the authors propose a pricing decision support model for retailers to estimate optimal prices,
Received in revised form 11 January 2010 whose output depends on the configuration of a supporting measurement model. The measurement model is a
Accepted 24 January 2010
demand function that relates sales and prices within the category; optimal prices are those whose effects on
Available online 29 January 2010
demand and retail margins maximize the category's profitability. This investigation focuses particularly on the
Keywords:
role of competitive structure, such that the authors consider two types of price competition asymmetries for
Retail pricing decision model demand forecasting: those depending on the brand (differential price effects) and those dealing with demand
Demand forecasting for competing brands (cross-price effects). By explicitly modeling competitive asymmetries in the demand
Market share models function that underlies the decision support model, the authors assess implications for pricing decisions,
Category management sales, and profitability. The empirical application of the model to store-level, aggregated scanner data for
Competitive structure two frequently purchased categories reveals the impact of an asymmetric competitive structure on demand
Store-level scanner data forecasting and optimal pricing decisions. Furthermore, this article quantifies the costs of ignoring asymmetric
competitive interactions in retailers' decision making.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction A full understanding of any competitive structure requires the

The increasing power of retailers has become more apparent in
the form of greater autonomy in their final pricing decisions. These
autonomous retailer pricing decisions constitute a key element of mar-
keting channel performance that can determine the profits of manu-
facturers [18], especially if the decisions come from large-scale retailers
[20]. Therefore, it should come as no surprise that retailing research has
focused primarily on category management and the favorable con-
sequences of this management strategy compared with brand-centered
management [45]. Authors cite the key role of category management in
achieving more profitable retail pricing structures [3], yet no studies
investigate the extent to which these positive consequences might
depend on the implementation of the category management. Despite
the importance of assessing and understanding the competitive struc-
ture of product categories for successful category management, prior
research offers little empirical support. This article addresses this gap by
analyzing how a greater understanding of the competitive structure
within a product category may improve retail pricing decisions in the
context of category management approaches.
⁎ Corresponding author. Tel.: + 34 923 294 400 (3508); fax: + 34 923 294 715.
E-mail addresses: oscargb@usal.es (Ó. González-Benito),
MariaPilar.Martinez@uclm.es (M.P. Martínez-Ruiz), Alejandro.Molla@uv.es
(A. Mollá-Descals).
1
Tel.: + 34 902 204 100; fax: + 34 902 204 130.
2
Tel.: + 34 963 828 334; fax: + 34 963 828 333.
analysis of competition asymmetries [8]; we consider the role of two
types of price competition asymmetries. First, the impact of variations
in a brand's price may differ depending on the brand, in that the price
changes of different brands likely affect demand with greater or lesser
intensity. Second, the impact of variations in a brand's price may differ
across the various levels of demand for competing brands. The higher
the substitutability of two brands, the greater the impact of their price
changes. If retailers explicitly consider these asymmetries in their
category pricing decision making, they may make more precise pre-
dictions about market responses to their pricing decisions and thereby
improve their profitability.
We propose a pricing decision-making model, based on aggregat-
ed scanner data in the context of frequently purchased categories.
In this model, the optimal prices are those whose effects on demand
and retail margins maximize the category's profitability; the output
therefore depends on the configuration of a supporting measurement
model, in which the demand function relates sales and prices within
the category. Our proposed model explicitly details competitive
asymmetries in the demand function, which indicates their implica-
tions for pricing decisions and thus for sales and profitability.
Various authors have considered the problem of pricing decision
optimization for retailers [10,21,25,29,31,39,42]. In Table 1, we sum-
marize the key contributions of these studies for our investigation.
Although all these approaches recognize that the optimal prices for
a product category maximize its expected profits, few consider
competitive asymmetries explicitly when they specify the relationship

0167-9236/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.dss.2010.01.009
 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119
Table 1
Previous research contributions regarding the category pricing optimization decision.
Research Retail context
Reibstein and Gatignon [31] Grocery retailing
Vilcassim and Chintagunta [42] Grocery retailing
Kim, Blattberg, and Rossi [21] Grocery retailing
Tellis and Zufryden [39] Grocery retailing
Montgomery [29] Grocery retailing
Chintagunta [10] Grocery retailing
Mantrala et al. [25] Automotive aftermarket retailing
Notes: Each line corresponds to a brand and represents the effects of price changes across competing brands.
between prices and sales. Rather, the effects they find result from
different assumptions about consumer behavior. For example, some
studies incorporate latent heterogeneity among consumers [10,21,42],
whereas others incorporate latent heterogeneity among stores
[25], though in neither case can they isolate the role of competitive
asymmetries from other price optimization determinants.
Two studies formalize each brand's demand within a product cat-
egory as a function of all brand prices to capture competitive asym-
metries [29,31], but this procedure is not robust [42], because the model
can produce unrealistic solutions, such as infinite or negative prices. To
resolve this issue, we incorporate brand demand decomposition into
category demand and market share. In other words, the price effect
consists of purchase decision and brand choice decision effects; the
latter reflect the competitive interaction among brands. Using market
share models that possess logical consistency adds robustness to the
demand function and enables more explicit modeling of the asymmetric
price effects [12].
This article's contribution is twofold. First, we investigate the role
of competitive structure in successful retail category management,
especially the extent to which understanding competitive structures
within a product category improves the profitability of the whole
category as a result of improved pricing decisions. Our proposed model
decomposes the price optimization problem of retailers into its
relevant components (i.e., decision model, overall category demand
model, and within-category market share model) and offers a way to
understand the effects of the competitive structure (i.e., competitive
asymmetries between brands). Second, our empirical application
shows that the proposed decision support model can operate with the
input of readily available, store-level, aggregated scanner data; we also
provide a numerical example in which the consideration of compet-
itive asymmetries changes the retailer's optimal decisions. Although
our approach is based on assumptions subject to some limitations,
it improves researchers' and practitioners' ability to make pricing
decisions within a product category by accounting for competitive
asymmetries. Our goal is to demonstrate that incorporating asymme-
try in cross-price effects alters a retailer's pricing decision.
2. Modeling proposal
Our proposed model attempts to determine optimal prices for every
brand included in a frequently purchased product category in a retail
store. The proposal employs scanner data aggregated to the store-level
and assumes that the retailer has information about sales, prices, and
other marketing variables for every brand within the product category
for a specified sample of time periods.
Of the three possible aggregation levels for scanner data—
household, store, and market [17]—most studies focus on the store-
level, because of the limited availability of disaggregated scanner data
from point-of-sale systems located in retail stores. Information about
the purchase history of specific households and consumers is available
exclusively for those clients that the retailer can identify, such as
through loyalty cards or credit accounts. Using such information can
lead to bias though, such as when consumers use several cards or
apply them only in specific purchase occasions (e.g., large purchases).
111
Data (level of aggregation) Explicit asymmetric effects?
Store-level aggregated data. Two stores. Yes
Household-level data No
Household-level data No
Household-level data No
Store-level aggregated data. One store Yes
Store-level aggregated data. Several stores No
Store-level aggregated data. Several stores. No
Furthermore, the loyalty card consumer group may not represent the
store's clients overall. Therefore, an analytical methodology that aims
to be replicable for any store at any time and whose estimations and
recommendations involve all potential customers of the store requires
the use of scanner data aggregated to the store-level.
Our proposed model contains two key elements. First, the decision
model provides an objective function, whose optimization determines
the optimal prices, and the function depends on the market response
to marketing stimuli, such as price. Second, this proposal represents a
formalization of the measurement model that captures the relationship.
2.1. Decision model: objective function
A store's objective is to maximize the profit generated by the
product category [21,31,39,42]. Profit Bt in period t thus equals the
sum of the gross profit generated by each brand N that constitutes the
category, minus the fixed costs CFt for that category. The profit of brand
i equals the sold units Qit multiplied by the unitary margin, which is
the difference between the unitary price Pit and the unitary cost Cit. In
short,
 
N
Bt = ∑i = 1 ðPit −Cit Þ⋅Qit −CFt ð1Þ
The quantity that each brand sells depends on the pricing policy
adopted for the category. Although the model focuses on this variable,
the quantity sold also can represent a function of other marketing
variables managed within a category (e.g., promotions through feature
advertising and store displays). The quantity sold depends on other
contextual circumstances Et (e.g., seasonal consumer habits). Mode-
lers therefore must understand sales as an endogenous variable, ex-
plained as follows:
 
N
Qit = f fPit gi = 1 ; Et ð2Þ
Optimal prices maximize the expected profit, as in Eq. (1). These
prices do not depend on the fixed costs assumed by the retailer, so
modelers can ignore this factor from an operational point of view.
Although the retailing sector plays an increasingly important role in
setting prices, the manufacturer's restrictions on the retailer's brand
price modifications still influence the optimization problem; some
national brands impose particularly significant limitations on the
retail prices of their products.
In either case, our proposal constitutes a simplified interpretation
of the problem, with three primary limitations: First, it assumes a
fixed cost structure, but the unitary cost for a brand may depend on
sales volume, which can be negotiated with suppliers [39]. Similarly,
sales category volume may condition the fixed costs, at least partially.
Therefore, we assume that the manufacturer is a strategic player
with regard to maximization and that the wholesale price may be
restrictive. This assumption is not problematic, because this research
attempts to show that incorporating asymmetry in cross-price effects
can change the retailer's pricing decision. Second, though the analy-
sis does not include other product categories, complementary and
112 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119
substitutable relationships often exist across categories [9,33,43].
Certain product categories also may have objectives other than profit
maximization, such as generating customer flow to support other
categories, enhancing store loyalty, or defining the store image [14].
The optimal decision then maximizes profits across all product cate-
gories only after the consideration of all complementary and sub-
stitutability effects. Third, this approach does not integrate the
marketing actions of competing stores, even though the assortment,
prices, and promotions offered by competing stores likely affect cate-
gory sales [7,26,38,43]. Thus, the decision-making process may rep-
resent a search for equilibrium in the context of game theory.
These limitations relate closely to three important circumstances
that determine the pricing decisions of retailers [10], namely, negotia-
tions and agreements with suppliers, brand and category objectives, and
competition among retailers.
2.2. Measurement models: market response
After establishing the optimization problem, we must specify the
demand function in Eq. (2), which represents the measurement model.
Some researchers suggest splitting brand sales into total category sales
Qt and brand market share πit to obtain
Qit = πit ⋅Qt
This approach assumes that market share depends on the mar-
keting decisions for each brand, as well as the resulting competitive
interaction, whereas the total category sales absorb the effect of the
contextual circumstances. In this case, both variables appear endog-
enous, assuming that
   
N N
πit = f fPit gi = 1 and Qt = f fPit gi = 1 ; Et
2.2.1. Market share model
Models with logical consistency can explain market share [12,13].
This methodological framework requires coherence across estimated
market shares, which must be positive and add to 1; it also assumes
that the market share of each brand is equivalent to the attraction
Ait that the brand generates compared with the attraction of other
brands. That is,
Ait
πit =
∑Nj= 1 Ajt
To understand the attraction concept, consider the marketing
effort for a corresponding brand. Assuming a logit multinomial (MNL)
conditional configuration, attraction is:
Ait = exp ðαi + β⋅Pit Þ
where αi is an intrinsic attraction constant associated with each brand,
and β is a parameter that relates to the price effect. Logit models from a
disaggregated perspective can explain consumer choice behavior, such
that market share represents the choice probability. The theory of
random utility also helps clarify these models [24,27].
However, a classic explanatory configuration assumes symmetrical
competition between brands, such that the price effect of one brand on
another always remains the same, regardless of the brands considered
[22]. The β parameter, which indicates the degree of competition or
substitutability between brands, represents this constant effect. In
turn, the classic MNL configuration assumes two types of competitive
symmetries.
First, the effect of brand i on brand j is the same as the effect of
brand j on brand i (i.e., i → j = j → i). Such symmetry also marks a
model in which absolute cross-price effects, defined as the change in
the market share points of a target brand when the price of the
competing brand changes by 1% of the product category price, are
symmetric in a classic logit model [35]. However, this assumption
appears inadequate in certain competitive contexts. For example,
with regard to the relationships between the price and quality tiers
of brands and their price variations, promoting a high-priced brand
sometimes has a greater effect on the demand for low-priced brands
than vice versa [1,2,6,34–36]. The characteristics of some brands also
relate to category promotional price elasticities [44]. Incorporating a
specific parameter βi for each brand i offers a solution that assumes
different effects for each brand [12]. The differential effects extend the
proposed configuration in Eq. (6) as follows:
Ait = exp ðαi + βi ⋅Pit Þ ð7Þ
Second, this approach assumes that the effect of brand i on brand j
is the same as the effect of brand i on brand k or any other competing
brand (i.e., i → j = i → k). This property remains in effect even in a
model with differential effects, as outlined in Eq. (7). Therefore, the
cross-elasticity of the market share of brand j with respect to the price
of brand i must be independent of the former—a property called the
independence of irrelevant alternatives. This assumption is too re-
strictive in certain competitive contexts, such as when greater com-
ð3Þ petition occurs between brands that are closer in their price or quality
[32,35,36].
To solve this inconvenience, Cooper and Nakanishi [12] propose
an extended attraction model, in which the price of brand j has a
different effect βij on each competing brand i. This model of cross-
effects extends the configuration proposed in Eq. (7) as follows:
 
N
ð4Þ Ait = exp αi + ∑ j=1 βij ⋅Pjt ð8Þ
j≠i
The explicit representation of the asymmetric competitive effects
of price emerges through a sequential extension of the classic choice
model, that is, the differential effects model on the first level, and the
cross-effects model on the second.
2.2.2. Model of overall demand
To explain overall category demand, we propose an exponential
configuration:
 
12 6 δ
ð5Þ Qt = exp γ + ∑m = 1 ξm ⋅Emt + ∑d = 1 ζd ⋅Edt ⋅PCt ð9Þ
where Emt and Edt are dummy variables that indicate each month
m and each day d, respectively; ξm and ζ d are parameters for these
seasonal effects, respectively; and γ is the demand constant. To
identify this model, we must fix one parameter for the month effect
ð6Þ and one parameter for the day effect. Furthermore, PCt represents the
category price, or the sum of prices across brands weighted by their
observed market shares, calculated as
N
PCt = ∑i = 1 πit ⋅Pit ð10Þ
and δ is the price effect parameter. The price effect of each brand in
the category is proportional to the market share of those brands,
because the price variations of leader brands should have a greater
impact than those of less frequently purchased brands. The exponen-
tial configuration of the model also requires that the effect of the
category price be proportional to the sales in every season.
3. Empirical analysis
To exemplify the explanatory possibilities of our proposed model
and explore the role of a complex competitive structure with regard to
 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119 113

the optimization of category prices, our empirical application ad- Table 3

dresses the context of frequently purchased products. Brands in the yoghurt category: descriptive statistics.

Salesa
3.1. Study scenario and data
Mean S.D. Maximum Minimum Total Market
share
The data consist of receipts registered by a supermarket over the
Chamburcy 115.7 219.7 2226 0 34841 0.24
course of one year, which list 301 daily observations of prices and sales Sveltesse 76.4 97.4 1088 0 22996 0.16
and thus capture intra-week variance. However, no information about Danone 183.5 169.5 1168 0 55225 0.38
other promotional variables (e.g., feature advertising and displays) Yoplait 12.7 29.6 248 0 3828 0.03
that the grocery uses frequently is available. Two product categories Clesa 92.4 102.1 554 0 27812 0.19

represent the primary focus of our analysis: caffeinated ground coffee Priceb
in 250-gram packages and yoghurt in 125-gram packages. We provide
Mean S.D. Maximum Minimum
the statistics for the two categories in Tables 2 and 3.
Caffeinated ground coffee is a nonperishable category that we Chamburcy 27.4 2.8 31 21
Sveltesse 33.8 5.2 43 29
selected because it offers daily observations and was subject to frequent Danone 29.6 3.0 34 23
temporary price discounts during the study period. In addition, previous Yoplait 26.6 1.3 28 22
marketing studies have often analyzed this category [4,15,16,30]. The Clesa 20.7 2.0 24 19
yoghurt category also offers daily observations and frequent temporary a
Sales in units.
price discounts, but it represents a perishable product with a short b
Price in pesetas; 1 euro = 166.386 pesetas.
expiration date, which distinguishes it from the coffee category.
Thus, our experiment includes two product categories with dif-
ferent degrees of perishability. Prior marketing research demonstrates where git is the observed market share for brand i in period t. The
that promotional effects can cause significant differences across cat- values add to N for each period, that is, the number of observed
egories, depending on certain unpredictable phenomena, such as shares. In Tables 4 and 5, we summarize the estimation results for
stockpiling induced by promotions [28], and in this sense, the two both categories and outline the results of our test of the contribution
product categories also differ in their purchasing trends, such that of each explanatory configuration. The estimation procedure relies on
consumers buy yoghurt more frequently. This increased incidence GAUSS, with the OPTMUN (Newton–Raphson algorithm) optimiza-
can lead to more complex competitive structures, because substitut- tion routine.
ability and complementarity effects combine in frequently purchased The explanatory configuration for all models is significant. More-
categories [19]. over, incorporating asymmetric price effects has a significant impact
on both the differential effects model and the cross-effects model. The
3.2. Measurement models: estimation improvement in the model appears stronger for the yoghurt than for
the coffee category, which implies that the competitive structure is
As a first approximation of the decision problem, we consider three more complex in the former.
specifications to estimate the market share model (Eqs. 6–8). The The parameter associated with price in the classic model is neg-
estimation relies on an adaptation of the maximum likelihood method ative, as expected a priori. That is, a price increase reduces brand
for probabilistic models, and the parameter estimates maximize the attractiveness and market share. The price parameters with differen-
following likelihood function: tial effects are also negative, though they differ across brands. We
note that the price effects tend to be stronger for those brands with
N git ⋅N lower prices and market shares. The correlations between the price
L = ∏t ∏i = 1 πit ð11Þ
parameters and mean prices across the brands are 0.662 (coffee) and
0.207 (yoghurt); analogously, the correlations between the price
parameters and market shares are 0.645 (coffee) and 0.478 (yoghurt).
Finally, as we expected, all significant price parameters in the
Table 2
model with cross-effects are positive. This finding indicates that a price
Brands in the coffee category: descriptive statistics.
increase for a particular brand also increases the attractiveness of
Salesa competing brands, compared with the more expensive brand, causing
Mean S.D. Maximum Minimum Total Market their market shares to rise.
share A very useful methodological framework provides a means to
154 4.7 6.2 44 0 1411 0.10 interpret the competitive asymmetries between brands within a
Bonka 17.0 16.6 94 0 5120 0.35 product category [37]. Although analyzing specific competitive
Marcilla 14.6 15.7 91 0 4392 0.30 patterns in the coffee and yoghurt categories is beyond the scope of
Saimaza 5.2 9.5 76 0 1561 0.11
this article, we consider the relationship between price competition
Soley 4.6 7.3 57 0 1386 0.10
Bahia 2.2 3.4 26 0 646 0.04 and the price tiers of brands to be an interesting question. For each
brand, we therefore compute the correlations between the cross-price
Priceb parameters and the absolute mean price differences across competing
Mean S.D. Maximum Minimum brands. The average correlations within the categories are − 0.167
154 184.5 6.1 189 159 (coffee) and 0.133 (yoghurt). The result for the coffee category thus is
Bonka 206.4 13.7 235 185 consistent with prior research that finds higher substitution effects
Marcilla 217.6 21.0 259 189 between brands with more similar price levels. In marketing literature,
Saimaza 225.6 18.4 249 189
this phenomenon is called the neighborhood cross-price effect; price
Soley 204.7 11.1 235 187
Bahia 188.9 7.5 195 157 competition between brands grows as the similarity in their prices
increases [35]. The result for the yoghurt category is not consistent
Notes: Each line corresponds to a brand and represents the effects of price changes
across competing brands.
with that competitive pattern though. Furthermore, the observed
a
Sales in units. correlations are not high. It thus appears that competitive asymmetries
b
Price in pesetas; 1 euro = 166.386 pesetas. mainly arise from the unique features of the brand's own strategies [8].
114 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119

Table 4 3.3. Decision model: simulation

Market share model for ground coffee: estimation results.

Intrinsic attractiveness constants If we assume that the estimated measurement models provide the
market response, we can define optimal prices as those that maximize
Classic model Differential effects model Cross-effects model
the profit equation (Eq. 1). Without information about fixed or
154 0.55*** 0.96 6.31**
unit costs, our analysis again relies on certain assumptions; however,
Bonka 2.85*** 1.07 4.24***
Marcilla 3.16*** − 2.57 − 3.77*** because the optimization problem does not depend on fixed costs,
Saimaza 2.36*** 1.08 − 3.46 our estimation does not include any assumed values for them, and
Soley 1.53*** − 0.96 − 1.01 the estimated profit ignores such costs. In addition, a common pricing
Bahiaa 0 0 0 policy applies a similar margin (whether absolute or relative) to all
Price effectsb products, so we assume the unit cost for each product is proportional
to the average price, as listed in Tables 2 and 3. The costs for the coffee
154 Bonka Marcilla Saimaza Soley Bahia
brands equal 80% of the average shelf price, and those of the yoghurt
Classic model brands are 75% of their shelf price.
154 − 0.05*** 0c 0c 0c 0c 0c
Bonka 0c − 0.05*** 0c 0c 0c 0c
The estimation results obtained when we accept these assumptions
Marcilla 0c 0c − 0.05*** 0c 0c 0c appear in Tables 8 and 9 for the coffee and yoghurt brands, respectively.
Saimaza 0c 0c 0c − 0.05*** 0c 0c Our estimation, which does not assume any manufacturer restriction on
Soley 0c 0c 0c 0c − 0.05*** 0c the retailer prices, uses direct programming in GAUSS with the OPTMUN
Bahia 0c 0c 0c 0c 0c − 0.05***
(Newton–Raphson algorithm) optimization routines. The information
Differential effects model comprises optimal prices in both absolute and relative terms, that is, the
154 − 0.07*** 0c 0c 0c 0c 0c unitary and percentage margins, as well as the estimated effects of these
Bonka 0c − 0.05*** 0c 0c 0c 0c prices on sales and profits. We quantify the results for both categories
Marcilla 0c 0c − 0.03*** 0c 0c 0c and for each particular brand, again in both absolute and relative terms
Saimaza 0c 0c 0c − 0.06*** 0c 0c
Soley 0c 0c 0c 0c − 0.05*** 0c
(shares). The optimal prices do not depend on the seasonal period,
Bahia 0c 0c 0c 0c 0c − 0.06*** though sales and profits reflect the varying influence of different months
of the year and the days of the week.
Cross-effects model Classic analysis based on the symmetrical competition model in-
154 0c 0.07*** 0.07*** 0.08*** 0.07*** 0.07***
dicates that optimal pricing decisions are not homogeneous across
Bonka 0.04*** 0c 0.06*** 0.05*** 0.06*** 0.06***
Marcilla 0.03**** 0.03*** 0c 0.05*** 0.05*** 0.04*** brands; margins in both unitary and percentage terms differ across
Saimaza 0.05*** 0.06*** 0.06*** 0c 0.07*** 0.05*** brands. We therefore revise our proposed model to account for three
Soley 0.10*** 0.07*** 0.08*** 0.07*** 0c 0.07*** underlying brand-specific factors that may affect optimal prices: (1)
Bahia 0.06*** 0.05*** 0.06*** 0.08*** 0.06*** 0c observed market share, which affects total demand through the
Goodness of fitd category price; (2) intrinsic brand attraction, which influences the
estimated market share; and (3) unitary cost, which affects the profit
Classic Differential Cross-effects
model effects model model
margin. The percentage margins tend to be higher for brands with
lower prices. In addition, we find that the correlations between the
Likelihood ratio test 676.00*** 705.71*** 746.40***
percentage margins and prices across brands are − 0.941 (coffee) and
Pseudo R2 (ρ2) 0.12 0.13 0.13
−0.929 (yoghurt).
Comparison To model competitive asymmetries, we also need to consider
Likelihood ratio test 29.71*** variations in the estimated results of the optimal prices, sales, and
Likelihood ratio test 40.70**
profits. Explicitly modeling the individual price effects of each brand
***p < 0.01; ** p < 0.05; *p < 0.10. suggests different optimal prices, as does the explicit modeling of the
a
Reference brand, with a null constant of intrinsic attractiveness. different price effects across competing brands. The clearest variations
b
Effect of the price of the brand in this row on the attraction of the brand in this
column.
occur in the yoghurt category, consistent with its higher competitive
c
Parameters set to 0 in the explanatory configuration. asymmetry. Our analysis of both product categories indicates that no
d
Comparison with the trivial model specified only for the intrinsic attractiveness clear trends mark either the evolution of prices or estimated sales.
constants.
3.4. Economic implications of the asymmetric effects
The estimation of the total demand model in Eq. (9), after a log-
transformation of the function, provides a linear form. This estimation These results reveal competitive asymmetries between brands, so
is based on a multiple regression analysis carried out with SPSS, and we next attempt to quantify the impact of ignoring such effects, with
in Tables 6 and 7, we provide the estimation results for coffee and regard to optimal pricing decisions. Though a sequential configuration
yoghurt categories. of the market share model, we undertake an economic analysis in which
The explanatory configuration of the model is significant in both we compare the performance of each explanatory configuration.
categories, though the model adjustment is much greater for coffee The profit estimations for the classic, differential effects, and cross-
than for yoghurt. The price category effect also is significant and effects models in Tables 8 and 9 demonstrate that the differential
negative, which indicates that a price increase prompts an increase in effects model outperforms the classic model, and the cross-effects
the category price but a reduction in total demand. Significant seasonal model outperforms the differential effects model. The more extended
effects also appear in both categories; though they vary slightly the model, the better is its prediction of the market response. There-
between categories, higher demand for coffee and yoghurt consis- fore, we can estimate expected profits more precisely if we assume
tently occurs during the first months of the year and the last days of that the effect of prices on market shares behaves in accordance with
the week. the most complete model. Following this line of reasoning, we
An alternative definition of category price, as we introduced in evaluate the costs according to the cross-effects model and calculate
Eq. (10), that ignores weights (market shares) and simply includes the estimated sales and profits for the optimal decisions that derive
brand prices, produces a poorer fit of the models for both product from it. Furthermore, we consider the demand function of the model
categories. with differential effects and quantify the consequences of a decision
 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119 115

Table 5
Market share model for yoghurt: estimation results.

Intrinsic attractiveness constants

Classic model Differential effects model Cross-effects model

Chamburcy 1.10*** 6.19*** 8.04***

Sveltesse 2.02*** − 0.64 8.74***
Danone 2.36*** 3.50*** 10.55***
Yoplait − 0.81*** 8.52*** 21.14***
Clesaa 0 0 0

Price effectsb

Classic model

Chamburcy Sveltesse Danone Yoplait Clesa

Chamburcy − 0.19*** 0c 0c 0c 0c
Sveltesse 0c − 0.19*** 0c 0c 0c
Danone 0c 0c − 0.19*** 0c 0c
Yoplait 0c 0c 0c − 0.19*** 0c
Clesa 0c 0c 0c 0c − 0.19***

Differential effects model

154 Bonka Marcilla Saimaza Soley Bahia

Chamburcy Sveltesse Danone Yoplait Clesa

Chamburcy − 0.34*** 0c 0c 0c 0c
Sveltesse 0c − 0.07*** 0c 0c 0c
Danone 0c 0c − 0.19*** 0c 0c
Yoplait 0c 0c 0c − 0.51*** 0c
Clesa 0c 0c 0c 0c − 0.13***

Cross-effects model

Chamburcy Sveltesse Danone Yoplait Clesa

Chamburcy 0c 0.29*** 0.28*** 0.36*** 0.41***

Sveltesse 0.10*** 0c 0.06*** 0.02 0.03
Danone 0.28*** 0.11*** 0c 0.18** 0.27***
Yoplait 0.67*** 0.57*** 0.59*** 0c 0.67***
Clesa 0.01 0.14** 0.13*** − 0.06 0c

Goodness of fitd

Classic model Differential effects model Cross-effects model

Likelihood ratio test 341.56*** 455.76*** 527.62***

Pseudo R2 (ρ2) 0.08 0.11 0.12

Comparison
Likelihood ratio test 114.20***
Likelihood ratio test 71.86***

***p < 0.01; **p < 0.05; *p < 0.10.

a
Reference brand, with a null constant of intrinsic attractiveness.
b
Effect of the price of the brand in this row on the attraction of the brand in this column.
c
Parameters set to 0 in the explicative configuration of the model.
d
Comparison with the trivial model specified only for the intrinsic attractiveness constants.

based solely on the symmetric classical model. In Tables 10 and 11, we
present the results of these analyses, along with the costs calculated
for the day of the week and the month that correlate with the lowest
expected demand.
Our comparison of the expected results for the cross-effects
models (Tables 8 and 9) with our subsequent estimations (Tables 10
and 11) clarifies the impact of simplifying the competitive structure
representation. The greatest consequences pertain to the difference in
the predicted results when demand actually mirrors the predications
of the cross-effects model. For the coffee category, the losses in the
classic model equal 5.15 pesetas per day (827.74–822.60), whereas
for the yoghurt category, they reach 44.58 pesetas daily (1536.79–
1492.18). If we aggregate these losses for the whole year and take the
seasonal effects into account, we calculate a total annual loss of
3,827.35 pesetas (approximately 23€) in the coffee category and
27,716.06 pesetas (167€) in the yoghurt category. The economic
consequences are much more severe for the perishable category than
for the nonperishable category, consistent with the higher compet-
itive asymmetry we detect in that category.
In summary, because retailers typically carry several product lines
for a wide variety of manufacturers, they are interested in product
category sales, not just sales of specific items. An assessment of category
performance requires the consideration of competitive effects, such that
retailers need to measure the impact of each brand price on the sales of
competing brands, which can constitute a highly complex competitive
interaction. Our findings show that the economic impact of assuming
a simplified, symmetric competitive structure seems minimal if we
consider only one category, but in the context of the multiple product
categories that even a small retail store carries, the sales losses as-
sociated with a poor understanding of the competitive structure can be
severe and detrimental to its long-term economic viability.
4. Conclusions
To contribute to existing knowledge about the factors that underlie
category management success in retail stores, we investigate the role
of the category's competitive structure and propose a decision model
that can indicate optimal category prices, according to scanner data
116 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119

Table 6 Table 8
Total demand model for ground coffee: estimation results. Optimal pricing decisions in the coffee category.

Constant 35.69*** 154 Bonka Marcilla Saimaza Soley Bahia

Category price
Assumption
Category price − 6.13***
Unitary cost 147.58 165.10 174.08 180.46 163.72 151.11
Seasonal effects
Month of the year
Classic model
January 0.89***
February 0.72*** Optimal price decision
March 0.95*** Price 178.20 197.72 207.19 212.78 196.98 182.34
April 0.55*** Unitary margin 30.61 32.62 33.11 32.31 33.26 31.23
May 0.53*** Percentage margin 20.74 19.76 19.02 17.91 20.32 20.67
June 0.31***
July 0.40*** Results
August 0.22** Total salesa 25.34
September 0.11 Salesa 2.25 8.96 7.87 2.70 2.49 1.07
Octobera 0 Market share 8.87 35.35 31.07 10.66 9.81 4.23
November 0.10 Total profita 825.14
December 0.09 Profita 68.83 292.19 260.65 87.30 82.70 33.47
Day of the week Profit share 8.34 35.41 31.59 10.58 10.02 4.06
Mondayb 0
Tuesday 0.04 Differential Effects Model
Wednesday 0.02
Optimal price decision
Thursday 0.33***
Price 179.14 198.06 206.52 212.87 197.06 182.91
Friday 0.75***
Unitary margin 31.55 32.96 32.44 32.41 33.34 31.80
Saturday 1.07***
Percentage margin 21.38 19.96 18.63 17.96 20.37 21.04
Goodness of fitd
R2 0.64
Results
ANOVA 29,371***
Total salesa 25.29
***p < 0.01; **p < 0.05; *p < 0.10. Sales a 2.30 9.10 7.62 2.70 2.49 1.08
a
Reference month with a null parameter. Market share 9.08 35.97 30.13 10.69 9.86 4.28
b
Reference day with a null parameter. Total profita 824.83
Profit a 72.45 299.89 247.25 87.62 83.15 34.47
Profit share 8.78 36.36 29.98 10.62 10.08 4.18

aggregated to the store-level. Optimal prices maximize expected profit

Cross-Effects Model
through their effects on demand and margins. Moreover, by explicitly
Optimal price decision
modeling differential and cross-price effects on brand demand, we can
Price 180.26 198.33 203.44 208.86 194.42 179.22
assess the role of an asymmetric competitive structure for optimal Unitary margin 32.68 33.22 29.36 28.40 30.70 28.10
decisions, as well as quantify the economic consequences of ignoring Percentage margin 22.14 20.12 16.87 15.74 18.75 18.60
this asymmetry. The cross-effects model that we propose includes
various price effects between pairs of brands separately, which makes Results
Total salesa 26.56
Sales a 2.81 10.24 7.51 2.41 2.26 1.33
Market share 10.59 38.55 28.26 9.06 8.52 5.02
Total profit a 827.74
Table 7
Profit a 91.93 340.20 220.38 68.32 69.47 37.44
Total demand model for yoghurt: estimation results.
Profit share 11.11 41.10 26.62 8.25 8.39 4.52
Constant 16.80*** a
Calculated assuming the minimum demand seasonal period (October/Monday).
Category price
Category price − 3.44***
Seasonal effects
Month of the year
January 0.74*** it more sophisticated than any existing explanatory configuration,
February 0.57** such as those that distinguish price effects in terms of national and
March 0.64***
April 0.46**
store brands, low- and high-priced brands, and so forth.
May 0.36** The empirical application of our model to two frequently purchased
June 0.19 product categories, one perishable and one nonperishable, confirms the
July 0.08 analytic possibilities of our model. The robust and aggregated nature
August 0.29*
of this proposal makes it easily applicable by any retail establishment
Septembera 0
October 0.03 at any time. The empirical application also clarifies the consequences
November 0.11 for the retailer if it makes its pricing decisions based solely on a simpli-
December 0.15 fied interpretation of the category's competitive structure. Specifically,
Day of the week greater competitive asymmetry will lead to higher cost estimates if the
Monday 0.21*
retailer ignores the competitive structure.
Tuesday 0.22**
Wednesdayb 0 Scanner data can provide the information needed to optimize
Thursday 0.18 marketing policies in a particular product category, though their
Friday 0.55*** ability to do so depends on the accuracy of the market response
Saturday 0.93***
estimation. The key lies in the measurement models that support the
Goodness of fit
R2 0.30 decision model. Explicit modeling of asymmetric competitive effects
ANOVA 7.03*** involves variations in both the optimal prices and the expected
***p < 0.01; **p < 0.05; *p < 0.10.
results, so ignoring such variance across the categories that constitute
a
Reference month with a null parameter. a store's retail assortment, especially with regard to frequently pur-
b
Reference day with a null parameter. chased products, will mean lost profits.
 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119 117

Table 9 Table 10
Optimal pricing decisions in the yoghurt category. Economic analysis for the simplified interpretation of the competitive interaction
among coffee brands.
Chamburcy Sveltesse Danone Yoplait Clesa
154 Bonka Marcilla Saimaza Soley Bahia
Assumption
Unitary cost 20.51 25.37 22.23 19.97 15.53 Assumption
Unitary cost 147.58 165.10 174.08 180.46 163.72 151.11
Classic model
Differential effects model
Optimal price decision
Price 28.31 34.25 31.27 28.91 23.56 Results based on optimal decision with the classic effects model
Unitary margin 7.80 8.88 9.04 8.94 8.04 Total salesa 25.34
Percentage margin 38.03 35.00 40.66 44.75 51.77 Salesa 2.43 9.20 7.40 2.70 2.48 1.11
Market share 9.57 36.32 29.22 19.66 9.81 4.41
Results Total profita 824.50
Total salesa 174.5 Profita 74.26 300.22 245.15 87.26 82.75 34.86
Salesa 36.56 29.42 73.03 4.87 30.32 Profit share 9.01 36.41 29.73 10.58 10.04 4.23
Market share 20.99 16.89 41.92 2.79 17.40
Total profita 1493.84 Cross-effects model
Profita 285.23 261.29 660.14 43.51 243.67
Results based on optimal decision with the classic effects model
Share profit 19.09 17.49 44.19 2.91 16.31
Total salesa 25.34
Salesa 3.20 10.14 6.76 2.11 2.06 1.08
Differential effects model
Market share 12.62 40.01 26.67 8.31 8.13 4.25
a
Optimal price decision Total profit 822.60
a
Price 28.45 37.07 31.89 28.15 24.80 Profit 97.86 330.70 223.79 68.06 68.55 33.64
Unitary margin 7.94 11.70 9.66 8.18 9.28 Profit share 11.90 40.20 27.21 8.27 8.33 4.09
Percentage margin 38.69 46.10 43.45 40.96 59.74
Results based on optimal decision with the differential effects model
Results Total salesa 25.30
Total salesa 156.47 Salesa 3.03 10.02 7.02 2.12 2.04 1.06
Salesa 25.91 31.73 65.11 2.52 31.20 Market share 11.98 39.63 27.76 8.39 8.08 4.17
a
Market share 16.56 20.28 41.61 1.61 19.94 Total profit 824.28
Total profita 1515.65 Profita 95.59 330.42 227.79 68.78 68.14 33.57
Profita 205.61 371.07 628.92 20.65 289.40 Profit share 11.60 40.09 27.63 8.34 8.27 4.07
Profit share 13.57 24.48 41.49 1.36 19.09 a
Calculated assuming the minimum demand seasonal period (October/Monday).
Cross-effects model

Optimal price decision effects. An assessment of how the explicit consideration of specific
Price 27.57 36.73 31.81 25.81 26.12 cross-effects may improve decision making is beyond of the scope of
Unitary margin 7.06 11.35 9.58 5.83 10.60
Percentage margin 34.41 44.75 43.11 29.21 68.27
our analysis, though an approach similar to the one we propose herein

Results
Total salesa 158.28
Sales a 22.60 33.96 72.74 2.65 26.33 Table 11
Market share 14.28 21.46 45.96 1.67 16.64 Economic analysis for the simplified interpretation of the competitive interaction
Total profita 1536.79 among yoghurt brands.
Profit a 159.50 385.63 697.10 15.45 279.11
Chamburcy Sveltesse Danone Yoplait Clesa
Share profit 10.38 25.09 45.36 1.01 18.16
a Assumption
Calculated assuming the minimum demand seasonal period (September/Wednesday).
Unitary cost 20.51 25.37 22.23 19.97 15.53

Differential effects model

Results based on optimal decision with the classic effects model

This study also confirms the importance of assessing categories Total salesa 174.20
to support retail category management. Retailers must attend to the Salesa 26.62 37.97 71.88 1.69 36.05
competitive structures of their product categories to determine their Market share 15.28 21.79 41.26 0.97 20.70
optimal prices and thus their expected profit. The same marketing Total profita 1499.37
Profita 207.66 337.18 649.68 15.08 289.76
actions applied to different brands may elicit different responses from
Profit share 13.85 22.49 43.33 1.01 19.33
consumers, which is highly significant for determining category-specific
price structures. Similarly, rivalry between brands differs across the Cross-effects model
various pairs of brands in a category, but it affects the optimal price Results based on optimal decision with the classic effects model
structure. Retailers therefore must recognize and understand the com- Total salesa 174.20
plex competitive structure that results from the balance of substitution Salesa 24.33 32.61 69.56 0.71 46.99
and complementary relationships among brands within a category. Our Market share 13.97 18.72 39.93 0.41 26.98
Total profita 1492.18
findings suggest retailers should rely on their own scanner data, input a
Profit 189.80 289.57 628.76 6.33 377.72
into sophisticated forecasting techniques and decision support models, Profit share 12.72 19.41 42.14 0.42 25.31
to obtain optimal solutions. Because optimal category prices depend
on many factors beyond the asymmetric competitive effects between Results based on optimal decision with the differential effects model
Total salesa 156.47
brands, including market shares, intrinsic attractiveness, and unitary
Salesa 23.45 28.31 65.45 0.82 38.44
costs, we assert that no clear patterns for optimal prices exist. In other Market share 14.99 18.09 41.83 0.53 24.56
words, any inflexible rule for fixing prices can be misleading. Total profit a
1512.66
a
Competitive asymmetries may alter the optimal pricing decision, Profit 186.12 331.14 632.17 6.74 356.49
so explicit modeling can improve profitability. However, we consider Profit share 12.30 21.89 41.79 0.45 23.57

only two kinds of asymmetries, namely, differential effects and cross- a

Calculated assuming the minimum demand seasonal period (September/Wednesday).
118 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119
could straightforwardly assess the economic impact of ignoring spe-
cific asymmetric competition effects (e.g., high versus low market
share brands; private versus store brands).
This proposal in turn could lead to an analytical management tool
that is easy to integrate and use and effectively supports retail man-
agement decisions. Such a tool would offer valuable benefits for
retailers. To design and implement the analytical tool for their specific
situation, the retailers need to gather scanner data, along with support
information that should be available from the retail manager (e.g.,
displays and feature advertising). Furthermore, they would need
the support and cooperation of their retail distributors, which may
require them to emphasize the positive implications of this tool,
such as the advantages and potential benefits of its applications, to
managers. The basic information input therefore consists of the retail
database and managerial judgments supplied by distributors. The
implementation and calibration of this extended tool with actual data
is key to confirming its validity, robustness, and reliability.
With its various assumptions and limitations, our proposed model
remains incomplete, but its development and empirical application
indicate specific areas that require more sophisticated modeling. In
particular, we note the need to connect marketing decisions to the cost
structure, especially agreements with suppliers (i.e., the manufacturer
is a strategic player in pricing decisions); the need to consider different
categories, including the complementary and substitutability effects
across them, simultaneously; and the need to calculate the response
of competing stores and the potential equilibrium state that may lead
to an optimal decision.
Furthermore, our empirical results must be interpreted with
caution due to our data limitations. Because we ignore other marketing
mix variables, our estimations may be biased. Another source of bias
in the estimation stems from the potential problems with price
endogeneity, which we do not properly address. Several researchers
have noted the need to consider endogeneity in firm marketing
decisions in explanatory models of market response (e.g., [5]), and
since then, some research contributions have focused on incorporating
the decision rules that govern firms' actions [11,23].
Another limitation pertains to the static nature of our proposed
model. Practitioners understand any marketing decisions about a
product category from a dynamic perspective; the lagged effects of
prices, as well as other marketing variables, should prevent inter-
pretations of prices as representative of time-isolated decisions. For
example, Van Heerde and colleagues [40,41]recognize that the growth
in sales caused by price promotions results from increased consump-
tion and brand switching, as well as from consumers' anticipation of
future purchases. Thus, any maximization of profits in a current period
might come at the cost of future profits. Researchers should extend
our investigation to determine the sequence of optimal decisions for a
product category over an extended period of time.
Acknowledgments
The authors thank the Spanish Ministry of Education and Science for
its financial support of the research reported in this paper (Research
Project SEJ ECON SEJ2005-06105). The authors also thank two
anonymous reviewers for their useful comments and suggestions on
previous versions of this article.
References
[1] G.M. Allenby, A unified approach to identifying, estimating and testing demand
structures with aggregate scanner data, Marketing Science 8 (3) (1989) 265–280
{AU: I did not find this source in the text but added it to a citation on p. 10; is that
placement okay? It will be much easier to add this source in the text than to
renumber all the references, so if not, please add it where appropriate}.
[2] G.M. Allenby, P.E. Rossi, Quality perceptions and asymmetric switching between
brands, Marketing Science 10 (1971) 185–204.
[3] S. Basuroy, M.K. Mantrala, R.G. Walters, The impact of category management on
retailer prices and performance: theory and evidence, Journal of Marketing 65 (4)
(2001) 16–32.
[4] D.R. Bell, J. Chiang, V. Padamanabhan, The decomposition of promotional response:
an empirical generalization, Marketing Science 18 (4) (1999) 504.
[5] D. Besanko, S. Gupta, D. Jain, Logit demand estimation under competitive pricing
behaviour: an equilibrium framework, Management Science 44 (11) (1998)
1533–1547.
[6] B. Bronnenberg, L. Wathieu, Asymmetric promotion effects and brand positioning,
Marketing Science 15 (4) (1996) 379–394.
[7] R.E. Bucklin, J.M. Lattin, A model of product category competition among grocery
retailers, Journal of Retailing 68 (3) (1992) 271–293.
[8] G.S. Carpenter, G.S.; L. G. Cooper; D. M. Hanssens and D. F. Midgley. Modeling
asymmetric competition, Marketing Science 7 (4) (1988) 393–412.
[9] Y. Chen, J.D. Hess, R.T. Wilcox, Z.J. Zhang, Accounting profits versus marketing
profits: a relevant metric for category management, Marketing Science 18 (3)
(1999) 208–229.
[10] P.K. Chintagunta, Investigating category pricing behavior at a retail chain, Journal
of Marketing Research 39 (2) (2002) 141–154.
[11] P.K. Chintagunta, V. Kadiyali, N.J. Vilcassim, Endogeneity and simultaneity in
competitive pricing and advertising: a logit demand analysis, Journal of Business
79 (6) (2006) 2761–2787.
[12] L.G. Cooper, Competitive maps: the structure underlying asymmetric cross
elasticities, Management Science 34 (3) (1988) 707–723.
[13] L.G. Cooper, Market-share models, in: J. Eliashberg, G.L. Lilien (Eds.), Marketing,
Handbooks in Operations Research and Management Science, vol. 5, Elsevier
Science Publishers B.V, North Holland, 1993, pp. 259–314, Amsterdam, 8.
[14] S.K. Dhar, S.J. Hoch, Affective category management depends on the role of the
category, Journal of Retailing 77 (2) (2001) 165–184.
[15] K.E. Fish, J.D. Johnson, R.E. Dorsey, J.G. Blodgett, Using an artificial neural network
trained with a genetic algorithm to model brand share, Journal of Business Research
57 (2004) 79–85.
[16] S. Gupta, Impact of sales promotion on when, what and how much to buy, Journal
of Marketing Research 25 (4) (1988) 342–355.
[17] S. Gupta, P. Chintagunta, A. Kaul, D.R. Wittink, Do household scanner data provide
representative inferences from brand choices: a comparison with store data,
Journal of Marketing Research 33 (4) (1996) 383–398 8.
[18] H. Haghighat, H. Seifi, A.R. Kian, The role of market pricing mechanism under
imperfect competition, Decision Support Systems 45 (2) (2008) 267–277 May.
[19] B.A. Harlam, L.M. Lodish, Modeling consumers' choices of multiple items, Journal
of Marketing Research 32 (4) (1995) 404–418.
[20] N. Hyun, J. Kyu, Knowledge assisted dynamic pricing for large-scale retailers, Decision
Support Systems 28 (4) (2000) 347–363.
[21] B.-D. Kim, R.C. Blattberg, P.E. Rossi, Modeling the distribution of price sensitivity
and implications for optimal retail pricing, Journal of Business and Economic
Statistics 13 (3) (1995) 291–303.
[22] L. Krishnamurthi, S.R. Raj, K. Sivakumar, Unique inter-brand effects of price on
brand choice, Journal of Business Research 34 (1995) 47–56.
[23] P. Manchanda, P.E. Rossi, P.K. Chintagunta, Response modeling with nonrandom
marketing-mix variables, Journal of Marketing Research 41 (4) (2004) 467–478.
[24] C. Manski, The structure of random utility models, Theory and Decision 8 (1977)
229–254.
[25] M.K. Mantrala, P.B. Seetharaman, R. Kaul, S. Gopalakrishna, A. Stam, Optimal
pricing strategies for an automotive aftermarket retailer, Journal of Marketing
Research 43 (4) (2006) 588–604.
[26] B.W. Marion, Competition in grocery retailing: the impact of new strategic group
on price increases, Review of Industrial Organization 13 (4) (1998) 281–299.
[27] D. McFadden, Conditional logit analysis of qualitative choice behavior, in: P. Zarembka
(Ed.), Frontiers in Econometrics, Academic Press, New York, 1974, pp. 105–142.
[28] C. Mela, K. Jedidi, D. Bowman, The long term impact of promotions on consumer
stockpiling behavior, Journal of Marketing Research 35 (2) (1998) 250–262.
[29] A.L. Montgomery, Creating micro-marketing pricing strategies using supermarket
scanner data, Marketing Science 16 (4) (1997) 315–337.
[30] C. Narasimhan, S.A. Neslin, S.K. Sen, Promotional elasticities and category character-
istics, Journal of Marketing 60 (1996) 17.
[31] D.J. Reibstein, H. Gatignon, Optimal product line pricing: the influence of elasticities
and cross-elasticities, Journal of Marketing Research 21 (3) (1984) 259–267.
[32] G.J. Russell, A model of latent symmetry in cross-price elasticities, Marketing
Letters 3 (1992) 157–169.
[33] S. Sayman, J.S. Raju, Investigating cross-category effects of store brands, Review of
Industrial Organization 24 (2) (2004) 129–141.
[34] R. Sethuraman, A meta-analysis of national brand and store brand cross-promotional
price elasticities, Marketing Letters 6 (4) (1995) 275–286.
[35] R. Sethuraman, V. Srinivasan, D. Kim, Asymmetric and neighborhood cross-price
effects: some empirical generalizations, Marketing Science 18 (1) (1999) 23–41.
[36] K. Sivakumar, Price-tier competition: distinguishing between intertier competi-
tion and intratier competition, Journal of Business Research 56 (2003) 947–959.
[37] K. Sivakumar, Manifestation and measurement of asymmetric brand competition,
Journal of Business Research 57 (2004) 813–820.
[38] K. Sudhir, D. Talukdar, Does store brand patronage improve store patronage?
Review of Industrial Organization 24 (2) (2004) 143–160.
[39] G.J. Tellis, F.S. Zufryden, Tackling the retailer decision maze: which brands to
discount, how much, when and why? Marketing Science 14 (3) (1995) 271–299.
[40] H.J. Van Heerde, P.S.H. Leeflang, D.R. Wittink, The estimation of pre- and postpromo-
tion dips with store-level scanner data, Journal of Marketing Research 37 (3) (2000)
383–395.
 Ó. González-Benito et al. / Decision Support Systems 49 (2010) 110–119
[41] H.J. Van Heerde, P.S.H. Leeflang, D.R. Wittink, Is 75% of sales promotion bump due
to brand switching? No, only 33% is, Journal of Marketing Research 40 (4) (2003)
481–491.
[42] N.J. Vilcassim, P.K. Chintagunta, Investigating retailer product category pricing
from household scanner panel data, Journal of Retailing 71 (2) (1995) 103–128.
[43] R.G. Walters, Assessing the impact of retail price promotions on product
substitution, complementary purchase, and interstore sales displacement, Journal
of Marketing 55 (2) (1991) 17–28.
[44] R.G. Walters, W. Bommer, Measuring the impact of product and promotion-
related factors on product category price elasticities, Journal of Business Research
36 (1996) 203–216.
[45] M.J. Zenor, The profit benefits of category management, Journal of Marketing Research
31 (2) (1994) 202–213.
Óscar González-Benito is Full Professor at the Área of Comercialización e Investigación de
Mercados of the University of Salamanca (Spain). He has participated in different
Conferences and Seminars worldwide as a presenter, reviewer and chairman. He has
written several articles in different international journals with a high standing impact (e.g.
Journal of Retailing, British Journal of Management, Journal of the Operational Research
Society, Marketing Letters, etc.). His main research lines are retailing, distribution and
operations management. Member of the European Marketing Academy (EMAC).
119
María Pilar Martínez-Ruiz is Associate Professor at the Department of Marketing of the
University of Castilla-La Mancha (Spain). She has participated in different Conferences
and Seminars worldwide and has written several articles in different high standing
international journals (e.g., The International Journal of Market Research, Journal of the
Operational Research Society, European Journal of Marketing, etc.). Her main research
lines are retailing, marketing communications, sales promotions and product and
services innovation. Member of the following associations: Academy of Marketing
Science (AMS), European Marketing Academy (EMAC) and The European Association
for Education and Research in Commercial Distribution (EAERCD).
Alejandro Mollá-Descals is Full Professor at the Department of Comercialización e
Investigación de Mercados of the University of Valencia (Spain). He has participated in
different Conferences and Seminars worldwide and has written several articles in different
international journals with a high standing impact (e.g. Journal of the Operational Research
Society, International Journal of Market Research, etc.). His main research lines are retailing,
distribution and consumer research. Member of the European Marketing Academy (EMAC).