The Role of Strategic Pricing by Retailers in the Success of Store Brands

Sergio Meza *
Rotman School of Management
University of Toronto
Toronto, ON M5S 3E6, Canada
E-Mail: sergio.meza@rotman.utoronto.ca
Phone: 905-569-4962
Fax: 905-569 -4302

K. Sudhir
Yale School of Management
135 Prospect St, PO Box 208200
New Haven, CT 06520
Email: k.sudhir@yale.edu
Phone: 203-432-3289
Fax: 203-432-3003

December 2005

The work described in this paper is part of the first authors dissertation at New York University. We thank Joel
Steckel, Yuxin Chen, Peter Golder and Pinelopi Goldberg for their comments and suggestions on this research. We
thank the seminar participants at Boston University, HEC, IESE, New York University, Rutgers, Santa Clara
University, SUNY Buffalo, Universidade Catolica Portuguesa, University of Texas at Austin, University of Central
Florida, University of British Columbia, University of Miami, University of Toronto and University of Washington.
We also thank the participants at the Albert Haring Doctoral Consortium at Indiana University, the Cornell
University Pricing Conference and the 2003 Summer Institute in Competitive Strategy at UC Berkeley for their
comments.

The Role of Strategic Pricing by Retailers in the Success of Store Brands

Abstract

A number of papers have evaluated demand and cost based explanations for the rapid
growth and success of store brands. However there has been little empirical investigation of the
strategic role of the retailer in facilitating the success of store brands. In this paper, we examine
the pricing behavior of a retailer in the ready to eat (RTE) breakfast cereal category in
investigating whether and how a retailer strategically favors store brands.
Our key result is as follows: After introducing the store brand, the retailer disfavors
national brands that it imitates (and is therefore in greater competition) by charging higher
margins. In contrast, it treats the national brands that it did not imitate (and therefore in less
competition) more favorably by charging lower margins. Thus somewhat counter intuitively,
some national brands actually benefit from the store brand introduction.
However such strategic behavior happens only in market segments that are attractive to
the retailer in terms of market size and the opportunity to steal share from the leading brand.
There is no difference in the treatment of imitated and non-imitated brands in the non-attractive
segments. We show that a nave approach to infer retailer behavior without considering such
segment level differences can lead to wrong and misleading inferences about the retailers
strategic pricing. We discuss the managerial implications of our results for pricing and product
positioning of national brands.

1. Introduction
The Success of Store Brands
Store brands have enjoyed tremendous success over the last two decades in gaining
market shares at the expense of national brands. In an analysis of over 225 categories during the
period 1987 to 1994, Hoch et al (2000) found the average annual share of sales for store brands
increased by 1.12%, while the average shares for the top three national brands in each category
fell by 0.20%. Mathews (1996) reports that in 1996, private sales in food stores increased 6.3%
compared to 1.3% for national manufacturers.
This success of store brands has drawn the attention of both retailers and manufacturers.
Progressive Grocers Annual Reports for 1999 and 2000 state that Stress Private Labels was
rated as the first likely action to be taken by retailers during these years. Manufacturers have in
turn responded vigorously to store brand introductions. For example, Cotterill et all (2000) report
that in response to the success of store brands, Post and Nabisco recently cut their prices and
consequently increased their market share from 16% to over 20% while decreasing private label
shares. In response Kelloggs announced a 20% across the board price cut.
The growth of store brands has spawned an academic literature investigating the factors
that facilitate its success (Hoch and Banerji, 1993; Starzynsky 1993, Raju et al, 1995; Hoch,
1996; Narasimhan and Wilcox, 1997; Dhar and Hoch, 1997; Chintagunta et al, 2002, Cotterill et
al 2000; Hoch et al 2000, Sethuraman 2000). The factors that have been considered in previous
research may be classified into three groups: (1) those associated with consumer characteristics
(demographics) and consumer preferences (brand loyalty, price sensitivity) (2) those associated
with the inherent costs and benefits of store brands (e.g. low cost of store brand, quality
differential between store and national brands) and (3) those associated with competitive
conditions of the category (e.g. number of competing brands, advertising levels).
This stream of literature however does not consider the retailers role as a strategic player
that can facilitate the success of store brands. As Hoch et al (2000) point out, the sustained
growth of the store brand over the last ten years could be because of the fact that the retailer has
control over not only the marketing mix of the store brand but also the marketing mix variables
of its competing national brands. Several papers have acknowledged or modeled theoretically the
strategic role of the retailer in the success of store brands. They have focused on two areas: (1) A

number of papers (Hoch 1996, Hoch et al 2000, Sayman, et al 2001, Scott Morton and
Zettelmeyer 2001) argue that store brands have an inherent advantage in that the retailer can
flexibly choose the positioning of store brands conditional on the positioning of national brands.
(2) Other papers have discussed the retailers control over several marketing mix variables such
as price, shelf space position and promotion (Hoch and Banerji 1993, Raju et al 1995; Hoch
1996; Narasimhan and Wilcox 1998; Dhar and Hoch 1997). But these papers do not empirically
analyze whether and how a strategic retailer contributes to the success of the store brand by
exercising its power to set the retail marketing mix for both the store brand and competing
national brands.
Research Questions and Empirical Strategy
Our goal in this paper is to investigate the strategic role played by the retailer in enabling
the success of store brands. Specifically, we seek answers to the following questions: First, does
the retailer change its pricing rule for national brands after a store brand is introduced in order to
favor store brands? Second, if the retailer exercises its pricing power to favor store brands, do all
national brands suffer? Finally, if there is heterogeneity in how the retailer prices different
national brands after introducing the store brands, what are some of the market factors that
determine the differences in retailer pricing of national brands?
The introduction of a store brand affects demand facing national brands in potentially
complicated ways. Typically, the elasticities will increase for national brands, if the store brand
choose a similar positioning as the national brand, but it may fall if only price sensitive
customers systematically shift to the store brand. Chintagunta et al (2002) find significant
changes in elasticities. Such changes in demand elasticity due to the introduction of the store
brand will cause the retailer to change retail prices. Further, manufacturers may change their
wholesale prices to reflect the new market reality. The change in marginal costs for the retailer
(due to changes in wholesale prices) also causes the retailer to change retail prices. These
changes in retail prices are a natural outcome of the changed demand curves and marginal costs
faced by the retailer and not due to a strategic decision by the retailer to favor store brands.
Therefore, it is not possible to simply look at the differences in retail prices before and after store
brand introduction to see if the retailer is changing its pricing strategy towards manufacturers.
The empirical strategy needs to control for changes in demand and cost conditions before and

after the store brand introduction in order to identify whether the retailer strategically favors
store brands.
How we control for demand and cost conditions in the market is tied to the nature of the
data that we use to investigate the strategic behavior of the retailer. We now explain the choice of
data that we use for exploring the research questions discussed above. We use the Dominicks
data made available by the University of Chicago for our investigation, because Dominicks
introduced store brands in multiple categories during the period for which the data are available.
In order to empirically investigate the research questions about heterogeneity in strategic retailer
behavior as a function of market characteristics, one possible strategy would be to analyze data
on how retailer strategic behavior varied after the introduction of store brands in multiple product
categories with different market characteristics. However, a problem with this strategy is that the
players (manufacturers) in the different categories would be different and it would be hard to
isolate whether the differences in retailer strategic behavior is due to the inherent differences in
the relationships with the different manufacturers or due to differences in the market
characteristics. To avoid this problem, an alternative strategy would be to use a category where
store brands are introduced in multiple sub-categories with different characteristics (such as size
or concentration), but where the manufacturers in all of the subcategories are the same. Our
exploration of the Dominicks data revealed that only the breakfast cereal category satisfied this
criterion. Six store brands were introduced in different sub-categories of the breakfast cereal
category and the manufacturers in all of these categories were the same. Further ready to eat
breakfast cereal is one of the most important product categories for a retailer and there would be
significant incentives for the retailer to indulge in strategic behavior. We therefore use the
breakfast cereal category for this research. We discuss more details about the data in a later
section.
The Dominicks data set has information on wholesale prices, which allows us to easily
control for retailer cost differentials before and after store brand introduction. However,
controlling for demand effects is more difficult. We need to estimate a demand model that allows
us to flexibly characterize the demand facing the retailer before and after the store brand
introduction. From an estimation point of view, it is difficult to estimate a demand model for the
breakfast cereal category. This is because using just aggregate store level data, we have to
characterize cross-elasticities across a large number of brands. However, we use recent advances

in the econometrics literature (Berry 1994; Berry, Levinsohn and Pakes 1995; Nevo 2000) to
estimate a flexible random coefficients logit model accounting for both observed heterogeneity
(demographics) as well as unobserved heterogeneity in the preferences for product attributes. See
Dube et al. (2002) for a discussion of the flexibility of the random coefficients logit model when
using aggregate data.
Related Research
This paper is closely related to recent research in the New Empirical Industrial
Organization (NEIO) framework that has focused on the detailed analysis of the strategic
behavior of the retailer within a category. Kadiyali et al. (2000) investigates the nature of power
in manufacturer-retailer interactions in their analysis of the analgesics market. Sudhir (2001b)
tests for three alternative types of behavior by the retailer: constant markup, brand profit
maximization and category profit maximization. He finds support for category profit maximizing
behavior for both the peanut butter as well as yogurt categories. Berto-Villas Boas (2001)
extends the analysis in Sudhir (2001b) using a random coefficients logit model. Villas-Boas and
Zhao (2001) perform a similar analysis of the ketchup market using individual level data.
Recently, researchers have paid specific attention to the retailers pricing behavior after
introducing store brands. Chintagunta (2002) finds that the retailer deviates from category profit
maximizing behavior in order to favor the store brand in the analgesics category. Chintagunta et
al. (2002) find that after a store brand is introduced, retail margin for Quaker Oats (the only
major incumbent in the oatmeal category they analyze) increases indicating that the retailer
gained power. The results of the above studies suggest that national brand manufacturers suffer
when store brands are introduced, as retailers will use their power to set the marketing mix for
both their store brands and national brands in such a way as to favor the store brand.1 The above
two studies do not differentiate the strategic behavior of the retailer towards different national
brands. In contrast we investigate and document differences in behavior towards imitated and
non-imitated national brands and in different segments of the category. As we will see later, a
key finding in our paper is that not all brands suffer due to the store brand introduction. In fact,
the retailer may favor some national brands after the introduction of store brands.
1

Chintagunta et al. (2002) also analyze the pasta category where they find that one of the national brands raises its

wholesale prices, but they attribute it to demand advantages gained by this manufacturer due to expansion of the
product line at the same time the store brand was introduced.

The rest of the paper is structured as follows: Section 2 develops the expectations about
retailer behavior. Section 3 introduces the model and Section 4 discusses the estimation
methodology. Section 6 describes the data. In section 6 we present the results and its managerial
implications. Section 7 concludes.

2. Expectations about Retailer Behavior

A retailer has strong incentives to ensure the success of its store brands and therefore
strategically take actions to support the store brands. Chintagunta (2002) finds that a retailer
supports store brands by reducing the prices of store brands. In contrast, we suggest that the
retailer could not only reduce the prices of store brands, but also the raise prices of national
brands in order to favor store brands. However, the retailer may incur some costs when they
choose to favor the store brands. Therefore, the retailer needs to tradeoff the benefits from
favoring store brands to potential costs. We discuss these tradeoffs below.
Tradeoffs Facing the Retailer When Introducing Store Brands
We begin with the potential benefits of supporting store brands. First, store brands tend to
typically provide greater margins for the retailer (Hoch and Banerji 1993, Sayman et al 2000,
Narasimhan and Wilcox 1998, Ailawadi & Harlam 2002, Pauwels and Srinivasan, 2002),
because the wholesale prices for the store brands tend to be smaller than for national brands
(Narasimhan and Wilcox, 1998, Raju, Sethuraman, and Dhar, 1995, McMaster 1987). A strong
store brand also improves the retailers bargaining power with national manufacturers (Giblen
1993, Chintagunta et al 2002 and Kadiyali et al 2000). The improved bargaining power is due to:
(1) a high cross price sensitivity between the store brand and the national brands (Sayman, et al
2000) or (2) the existence of a considerable group of consumers willing to switch to the store
brand (Narasimhan and Wilcox, 1998). The improved bargaining power can translate into: better
trade deals (Giblen 1993), deeper and more frequent trade deals (Lal 1990), lower wholesale
prices (Narasimhan and Wilcox 1998, Sayman, et al 2001) and higher percent margins in
national brands (Ailawadi & Harlam 2002) for the retailer.
Corstjens and Lal (2000) provide a store loyalty based explanation as to why a retailer
might wish to have strong store brands. According to their model, a retailer would provide strong
incentives to favor the store brand initially after its introduction. Given a high quality store
brand, a reasonable fraction of consumers who try the store brand will become loyal to the store

brand and the resulting store differentiation will improve the long run profitability of the retailer.
Thus a retailer wishing to favor the store brand might use its power to set retail prices to raise the
relative prices of national brands.
We now consider the potential costs of favoring store brands: A retailer who changes the
relative prices of national brands from the short-run profit maximizing price loses profits in the
short-run in the hope of potential higher long-run benefits from a strong store brand. However
this higher profit cannot be taken for granted. Second, disfavoring national brands by raising
prices, can lead to short-run losses if consumers switch to competing stores in order to purchase
their favorite national brands at lower prices. Finally national brand manufacturers could retaliate
by substantially lowering advertising and promotional support for the retailer (Hoch and Banerji
1993). The retailer, who needs the advertising and promotional support of manufacturers to grow
the overall demand for the product category in the long run, will tradeoff this cost with gains
obtained by favoring the store brand and disfavoring the national brand.2
In choosing an appropriate strategy of favoring certain national brands and disfavoring
others, in order to support the store brands, the positioning of the store brand needs to be
considered.
Store Brand Positioning
When introducing store brands, retailers may use either a differentiation strategy or an
imitation strategy in positioning the store brands. Examples of a high quality differentiation
strategy where retailers introduce high quality differentiated brands that differentiate them from
the national brands include Presidents Choice from Loblaws in Canada, World Classics
from Topco, and Sams Choice from Wal-Mart. Alternatively, the retailer may differentiate by
offering a white-label generic or a low quality store brand (e.g., A&Ps Savings Plus line)
targeted to low quality oriented customers (Hoch 1996).

Even though the Robinson Patman act precludes manufacturers from discriminating between retailers, informal

conversations with sales people of consumer goods manufacturers suggests that manufacturers overcome this
problem by offering a menu of trade promotion deals to manufacturers, but given any particular retailers
requirements, business practices and cost structure they will choose the deal that the manufacturers want them to
choose.

The more common strategy however is an imitation strategy, where a retailer introduces a
store brand as a me-too product relative to a popular national brand. (Hoch et al 2000, Hoch
1996, Schmalensee 1978, Scott Morton and Zettelmeyer 2001). This strategy accounts for more
than 50% of the store brand introductions in the grocery industry (Scott Morton, and Zettelmeyer
2001). In our dataset, the retailer Dominicks used the imitation strategy: the national brands it
imitated are sales leaders in their respective market segments and among the largest brands in the
overall cereal market: Cheerios (#1 in sales), Frosted Flakes (# 2 in sales), Rice Krispies (#3 in
sales), Corn Flakes (#5 in sales), Raisin Bran (#6 in sales), and Froot Loops (#10 in sales). Our
focus will therefore be on retailer behavior after they have adopted the imitation strategy.3
Retailer behavior towards Imitated and Non-imitated brands
The retailer introducing a store brand can create room for the store brand in the market
in several ways: (1) it can provide the best location in the aisles (in the center of the aisle or
besides the best selling brand in the category (Scott-Morton and Zettelmeyer 2001); (2) it can
reduce the promotional activity of national brands and (3) It can decrease the price of the store
brand (Chintagunta, 2002) or (4) it can raise the relative prices of national brands with respect to
the store brand. In this paper, we will focus only on how the retailer uses prices to favor the store
brand.
When retailers imitate national brands with their store brands, the cross price elasticity of
the imitated national brands with respect to the store brands should be very high. However
imitated national brands pull away much more sales from the corresponding store brand with a
price cut rather than vice versa. Given this asymmetry, an effective strategy would be to
disfavor (by means of increasing their prices) the national brands that are being imitated by the
3

We use the names of the store brands to identify which national brand it imitated. In some cases, Dominicks used

the same names as the national brands i.e., Corn Flakes, Raisin Bran and Frosted Flakes. For others, it used
different but suggestive names indicating the national brands that it imitates. We treat Crispy Rice as an imitator
of Rice Krispies. Fruit Rings as an imitator of Froot Loops and Tasteeos as an imitator of Cheerios. In our
empirical analysis, we verify whether our interpretation of these as imitation brands is appropriate. We indeed find
that the cross-elasticity between the store brand and the appropriate imitated brand is much larger than with respect
to other brands (see Table 6. Further, imitated national brands pull away much more sales from the corresponding
store brand with a price cut rather than vice versa. This asymmetry in cross-elasticity is consistent with Blattberg
and Wisnewski (1989) and Sethuraman (1995).

store brands. This idea is consistent with Sayman, Hoch and Raju (2001) who show that when
introducing a store brand, retailers need to target the leading brand in order to maximize profits.
We therefore expect the retailer to disfavor national brands that are being imitated, relative to the
pre-store brand introduction period.
Also as we discussed earlier, the retailer should trade off the benefits from favoring the
store brand with the costs. The long-term losses that can happen due to retaliation from national
brand manufacturers who may withdraw promotional and advertising support, which are
essential to the development of the category itself (Hoch and Banerjee 1993). Such support helps
the whole category because it builds awareness and drives traffic to the store. Loss of such
support can thus hurt the growth of the category at the retailer. Also it would hurt the retailer
relative to other retailers who keep the manufacturer support. In resolving this tradeoff, the
retailer has two options: (1) it may choose not to disfavor the brand that it imitates and continue
to treat it as before the store brand introduction; (2) it may disfavor the store brand, but favor
other national brands that do not directly compete with the store brands. Thus the national brand
manufacturers may forgive the retailer for intruding on the imitated brand because of the
benefits to the other national brands. However, given that the imitated brand tends to be the most
popular brand in the category, the effectiveness of the second strategy in avoiding retaliation
may not be very high. Manufacturers may not tolerate even a small loss to their popular brands.
We therefore treat the issue of whether the retailer will disfavor national brands that are being
imitated or favor national brands that are not being imitated as an empirical question.
Retailer pricing behavior towards brands on promotion (featured/displayed)
As we have argued previously, it is possible that in order to facilitate the success of the
store brands, the imitated national brands could be disfavored, less often promoted through
feature and display advertising. Further, even when promoted they may not be offered as large a
discount that accompanied such promotions prior to the store brand introductions. However, as
before, if the threat of manufacturer retaliation is large, we may expect that the retailer may not
disfavor the national brand. However, the retailer may assuage the losses of the manufacturers by
favoring their non-imitated brands (while still disfavoring the imitated brands to support store
brands) in order to reduce the threat of retaliation. Also, if one of the costs that a retailer expects
to incur by supporting store brands is the loss of trade promotions, the retailers may provide

greater support for trade promoted brands to encourage manufacturers to continue trade
promotions. Our analysis will help to shed light on which of these strategies the retailer will
adopt.
Differences in Retailer Behavior in Different Market Segments
In the previous sub-sections, we discuss the motivation that a retailer has in behaving
strategically and disfavoring certain brands, while favoring others. However, will retailer
behavior be different across different segments or sub-categories of products? We believe that
the retailer is likely to engage in strategic behavior in segments or sub-categories of products
where the strategic behavior can lead to large payoffs. For example, the retailer may be induced
to behave strategically in larger segments due to the larger payoffs involved. Further, when
pursuing an imitation strategy, a market with higher concentration (or more specifically, where
the leader has a very high share) is potentially more attractive so that it might be relatively easy
to steal share from such a national brand. Also, a retailer might not also find it attractive to price
strategically in segments that are not very price sensitive. We therefore treat large segments,
highly concentrated segments and relatively price-sensitive segments as attractive segments for
the retailer to engage in strategic behavior. We expect retailer behavior to change with respect to
national brands in the attractive segments. In contrast, in unattractive segments, one may not
expect the retailer to behave very strategically, because the gains from such strategic behavior
may be minimal. We therefore expect that there will not be changes in the retailer behavior after
the introduction of the store brands in unattractive segments.

3. The Model
As we discussed in the introduction, our goal is to understand how retailers strategically
change the prices of national brands in order to favor the store brand. This requires that we
control for demand and cost effects when analyzing retailer pricing behavior. Accordingly, we
present a model of retailer pricing conditional on an estimated demand model and wholesale
prices. We explain the details of the demand and retailer pricing models below.
Demand Model
Given the large number of products in the breakfast cereal category (over 40), we cannot
model demand using simple specifications such as linear or loglinear models since the number of
9

parameters necessary to model cross price effects will be enormous (see discussion in BLP 1995;
Sudhir, 2001a). We use a logit model where utility for a product is modeled as a function of
attributes and heterogeneity in consumer preferences for attributes is modeled using random
coefficients. The specicfication is both flexible and parsimonious and is ideal for our purposes.
Since different stores cater to different demographics and the intrinsic preferences for products
and price sensitivity are a function of demographics, we allow the preferences for product
attributes (distribution of the random coefficients associated with attributes) to be a function of
the empirical distribution of customers in the stores trading area. Past research has shown
clearly that demographic characteristics of the area served by the store can be associated with the
success of the store brand (Hoch and Banerji, 1993; Raju et al (1995)). This modeling framework
allowing the random coefficients distribution to be a function of demographics for demand is
similar to Nevo (2001). We thus allow for observable heterogeneity as well as unobservable
heterogeneity into our specification of demand.
Since our data is observed at the level of each store, we specify a demand model at the
store level. We observe the data for each store s = 1,..., S of the chain for t = 1,..., T periods of
time. The conditional indirect utility of consumer i for brand j at store s at period t is then given
by:
uijst = x j i* i* p jst + j + ijst

(1)

where xj is the k-vector of observable characteristics, pjst is the price of j at store s at time t, j is
the chain-level mean of brand specific valuation of j, ijst is a mean zero error term, and ( i* i* )
are k+1 individual-specific coefficients.
We also allow the consumer not to choose any of the J brands; i.e., we treat this nonpurchase option as the outside good whose average utility across individuals is normalized to
zero. Thus we can allow consumers to choose out of the category, if they find the prices are too
high.
Define i* = ( i* i* ) as a k+1 column vector containing the individual-specific

coefficients. We decompose the individual specific coefficients into an observable and

unobservable component as follows:

i* = 1 + Di + i ,

i ~ N (0, Ik + 1),

(2)

10

Here 1 contains the parameters (

) , Di is a vector of demographic variables,

is a

matrix that measures how the tastes for characteristics vary with observable demographics, is
a scaling matrix, and i represents the additional unobserved characteristics not explained by the
observed demographics.
Assuming vector 2 =

( vec ( ) , vec ( ) )

and combining equations (1) and (2), we

have:
uijst = jst ( x j , p jst , j ; 1 ) + ijst ( x j , p jst , i , Di ; 2 ) + ijst

jst = x j p jst + j ,

(3)

ijst = [ p jst , x j ]'( Di + i )

jst represent the mean utility from brand j at store s at time t, that does not vary by individual,
while ijst represents the individual level utility that varies across individuals.
Given this utility specification, a utility maximizing consumer i will purchase one unit of
j if for all k j, if uij > uik. So the probability of an individual i choosing brand j from store s at
time t is given by
Pijst =

exp( jst + ijt )

1 + exp( lst + ilt )

(4)

However we do not observe individual level purchases, but only aggregate store level
shares. Hence for matching to the observed data, we need to integrate out these individual level
probabilities over the population distribution of the observed (demographics) and unobserved
heterogeneity. Given the population distribution functions of Ds and denoted by P*(.) and
assuming independence among these distributions, the market share of j in store s at time t is
given by:
s jst ( x, p.st , .st ; 2 ) =

ijst

dP *( Ds ) dP *( )

(5)

Ds

The demand qjst for each store is obtained by multiplying the market share in (4) by the
total potential market Mst of each store. The demand qjst for product j at time t at the store level is
given by:

q jst = M st s jst ( x, p.st , .st ; 2 )

(6)

Retailer Pricing Equations

11

Dominicks, the chain whose data we analyze, uses zone pricing. Instead of setting prices
for each store separately, they group all their stores into three different pricing zones and select
for each week a unique price for each SKU in all the stores of a given zone. We therefore define
our supply equations at the zone level (denoted by z) rather than the store level.4
A retailer who maximizes category profits for zone z at time t, will maximize the
following objective function.
zt =

( p
j =1

jzt

w jzt ) s jzt M zt

(7)

where j indexes the brands sold, p indicates prices, w indicates wholesale prices, s indicates
shares and M indicates total potential market size in the category.
Suppose this retailer wants to favor or disfavor national brands in order to support the
store brand. We can capture these effects by looking for deviations in prices relative to the
category profit maximizing prices. We operationalize such deviations by augmenting the
category profit maximization objective to have additional weights on the shares of each brand as
follows:
Consider the following as if objective function for the retailer at time t, for zone z:
zt =

[( p
j =1

jzt

w jzt ) s jzt + jzt s jzt ]M zt

(8)

where j indexes the brands sold, p indicates prices, w indicates wholesale prices, s indicates
shares and M indicates total potential market size in the category.
The first part of the objective ( p jzt w jzt ) s jzt M zt is brand profit as before in equation (7).
By summing it over all brands, we capture the category profit. The second term jzt s jzt M zt ,
allows the retailer to aid (or suppress) the share of a given brand by placing a weight on that
brand's share jzt (which needs to be estimated). This interpretation suggests that the retailer
deviates from the single period category profit maximizing price in order to favor or disfavor the
shares of the concerned brands depending on the sign of jzt. This specification is similar to
4

It is inappropriate to estimate the supply model at the store level when the prices are set at the zone level because

it will make it appear that we have far more degrees of freedom than is warranted in the data. This inflates the
significance of the supply side coefficients by a factor of

# Stores
. Given that we have 90 stores and 3 price
# Zones

zones, we will inflate significance by a factor of 5.47.

12

Chintagunta (2002) that incorporates an additional weight on the share of the store brand in the
objective function. Our specification is more flexible in that it allows the retailer to strategically
favor/disfavor all brands.
Rewriting the first order conditions with respect to zone prices, the retailers price
equation for the brands has the following matrix form:
p1zt

p2 zt

. =

pJzt
123

Retail Price

w1zt

w2 zt

wJzt
123

Wholesale Price

s1zt s2 zt . . . . sJzt s1zt

p1zt p1zt

p1zt

s2 zt
s1zt . . . . . . . .

p2 zt
.

.
.
s

s2 zt
s
1zt
. . . . Jzt s
pJzt Jzt
pJzt pJzt
14444442444444
3
Category profit maximizing margin

1zt

2 zt

Jzt
{

(9)

Favor/Disfavor Effect

From equation (9) it is clear that a positive value of jzt implies prices will be set lower
than the category profit maximizing price and thus favors that brand in terms of improving its
market share. A negative jzt would imply a higher price relative to the category profit
maximizing price in order to disfavor the brand.
The Favor/Disfavor Effect

Our primary objective is to understand the factors that affect the favor/disfavor effect

(jzt). Specifically we look at how favor/disfavor effects are different for imitated brands versus
non-imitated brands and how these differences may vary across the different market segments
before and after store brand introductions. To do this, we include appropriate variables in the jzt.
For example, if we are interested in understanding how imitated brands will be priced differently
from non-imitated brands, we introduce an imitation dummy. The imitation dummy is set to 1 for
all national brands that are imitated by store brands. To understand the effect of introduction of
store brand on the prices, we introduce a store brand dummy variable. It takes the value of 0
prior to the introduction of the store brand, and the value of 1 after the introduction of the store
brand. To understand how the prices for an imitated brand differ before and after the store brand
was introduced, we introduce an interaction variable: Imitation X Store brand. To identify these
estimates at the segment level, we estimate the different parameters discussed above for each
market segment. Additionally, we use other control variables in jzt to account for relevant
13

differences before and after store brand introduction One control variable we used was
manufacturer dummies to account for any differences in manufacturer power with respect to the
retailer. We also test for how the retailer responded to promotions by including the promotion
variable. We define zjzt , as a vector of variables related to brand j at pricing zone z, and time t
that explain the favor/disfavor effect.
The effect jzt is parameterized as:

jzt = zjzt + jzt

jzt

N (0, 2 )

(10)

The estimates of allow us to test our hypotheses about the retailers strategic pricing behavior.
Revisiting equation (9), we have data on wholesale prices and retail prices. The category
profit maximizing margin can be computed given the demand elasticities estimated in the
demand equation. Thus the supply side equation enables us to estimate the parameters that
describe the retailers strategic pricing behavior. It is important to note that we are able to
estimate the parameters because we have data on wholesale prices.5

4. Estimation
In contrast to the random coefficients logit models that have been widely estimated using
individual data, our estimation will be using aggregate store level data. To estimate a random
coefficients logit model using aggregate data, we use a Generalized Method of Moments (GMM)
estimation procedure. This procedure was outlined in Berry (1994) and implemented by BLP
(1995) and extended by Nevo (2000).
A key issue in estimating the demand model is the endogeneity problem due to the
correlation between observed prices and the error term jzt in the demand model. This is because
any unobservable (to the researcher) characteristics captured in the demand side error term (jzt)
that attract or detract customers to the brand will be observed by both the retailer and the
consumer This necessitates the use of instrumental variables (IV) for estimation. Since the error
term jzt enters the demand equation non-linearly, we need to transform the equation in such a
way that the error term enters the estimation equation linearly to use IV methods. The

This is in contrast to Sudhir (2001b) and Berto Villas-Boas (2002), where the wholesale prices are inferred. In such

models, we will not be able to separately identify retailer strategic behavior as well as wholesale prices.

14

linearization is performed using the contracting mapping procedure outlined in Berry (1994), by
solving for jzt .
h +1
h
jzt
= jzt
+ ln ( Sjzt ) - ln ( sjzt ( x, p. zt , . zt ; 2) )

(11)

To avoid computing logarithms, we can follow the transformation suggested by Nevo

(2000). wjzt = exp (jzt). In this way, equation 13 can be rewritten as:
whjzt+1 = whjzt Sjt / sjzt ( x, p. zt , . zt; 2)

(12)

We then, iterate this equation until it converges. Then, we compute the errors in the
demand side in the form:

jzt = jzt ( 2) - xjzt 1

(13)

For the supply side, we can compute the margin conditional on the observed share
derivatives (which are a function of demand parameters)

mgjzt ( wjzt , sjzt , st* )

where, st* represents the partial derivatives of shares as a function of prices.
Therefore the supply side error is given by the difference between observed margins and
predicted margins and the deviations:

jzt = Mgjzt - mgjzt ( wjzt , sjzt , st* ) - jzt

(14)

Given these error terms we use Generalized Method of Moments (GMM) to estimate the
model. Using the instruments z, which are assumed to be exogenous, and independent of the
error term; Therefore E(z) = 0 and E(z) = 0 are the moment equations. Let = (, ). and let
be the set of parameters to be estimated. Then, the GMM estimator, given the moment
conditions, is defined as.
min ' z ( zz )-1 z '

(15)

Where is the standard weighting matrix defined by E().

This algorithm is circular because the weighting matrix is a function of the estimated
parameters , and the estimated parameters are a function of . We therefore use an iterative
procedure in which we assume some initial values for , then we compute . With this value of
, we minimize the objective function (equation 15), and repeat the procedure until convergence
of .

15

5. Data
As stated before, we use the cereal category in the Dominicks Finer Foods (DFF)
Database at the University of Chicago for our empirical application. The DFF database is
particularly useful for our research as it has information on the weekly retail margins for the
products. The DFF database consists of data from several stores with different demographic
characteristics that are classified into different price zones that have different retail prices. This
enables us to estimate the demand model with a greater level of richness, because the sensitivity
of prices to the different demographic characteristics can be estimated.
The cereal category from Dominicks is particularly appealing for our purposes for the
following reasons: (1) Store brands are introduced into several segments of this market midway
through the period of our data, thus allowing us to exploit the data for analyzing the impact of
store brand introduction on retailer pricing behavior. Our study covers a period of 52 weeks, and
store brands were introduced during weeks 23-26 of this period. (2) The cereal category consists
of a set of well-defined segments based on previous research (Nevo 2001): (i) Family, (ii) Kids,
(iii) Health and Nutrition and (iv) Taste Enhanced, so it enables us to assess differences in
retailer behavior across segments. By analyzing differences and similarities in the retailers
behavior across segments, we can better understand the motivations underlying the strategic
behavior of the retailer (3) Finally, the cereal category is a very important category to the retailer.
It is the second largest category in terms of dollar sales and therefore is a very likely candidate
for strategic pricing behavior by the retailer.
As we discussed in the section on store brand positioning, Dominicks used an imitation
strategy for introducing the store brand. The introduced store brands typically imitated popular
national brands in their respective segments. The six store brands introduced imitated six of the
top ten brands in terms of chain sales. The imitated brands are: Cheerios (#1), Frosted Flakes
(#2), Rice Krispies (#3), Corn Flakes (#5), Raisin Bran (#6), and Froot Loops (#10). To reduce
the computational burden, we only use the top 40 brands (accounting for 67% of the sales of the
category) for our estimation.
Share of Outside Good

While we observe the sales of all the competing brands in the category, we do not
directly see the number of consumers who do not purchase within the category. Using an

16

estimate of the potential market size, we therefore compute the number of consumers exercising
the no-purchase option. We estimate potential market size as follows: We assumed that each
household member can potentially consume one serving per day on 33% of the days. One
serving is estimated as 30 grams (as defined in the Cheerios box). The potential market (in
servings) is obtained by multiplying the number of customer visits in a given week times the
average number of household members for each store, times the percentage of days the consumer
consumes RTE cereal.6
The share of the outside good is given by 1

Quantity Purchased
.
Market Potential

Instrumental Variables

As discussed earlier, we need to use instrumental variables estimation to estimate the

price coefficient without bias. For instruments, we need to find variables that are correlated with
the price shocks, but are independent of the error term. BLP (1995) consider the average of
product characteristics of competing products as instruments. Sudhir (2001a) uses a similar
average but computes them for each segment. Nevo (2000) uses the average prices of other
regions as instruments for a regions price, since he used data from multiple markets.
Chintagunta (2001) uses the wholesale price. We use the spirit of the instruments used in the
above papers by using (a) the average price in other price zones, (b) the average price of all
competing products in each segment and (c) the wholesale prices.

6. Results
Descriptive Results

We graph the average retail prices, wholesale prices and margins over all brands during
the period of analysis in Figure 1. As can be seen in this graph, the retail prices and retail
margins tend to increase after the store brand introduction. Does this mean that retailers have
gained more power due to the introduction of the store brand? We cannot conclusively answer
this question without a structural model that separates out the demand, cost and strategic effects

We did the analysis with alternative assumptions of 28%, 30%, 40% and 60%. Our results are robust to these

assumptions.

17

as we do in this paper. Nevetheless, we look at more descriptive results to gain additional

insights about the market.
A more detailed segment-wise characterization of the data is provided in Table 1. Here
and in subsequent tables we split the data into three periods: (1) a 22 week pre-store brand
introductory period, (2) a 4 week transition period during which the six store brands were
introduced gradually into different stores of the chain. By this 4 week period, the introduced six
store brands had achieved penetration in 90% of the stores and (3) a 26 week post-store brand
introductory period. We exclude from our estimation the four-week transition period where store
brands are gradually being introduced throughout the chain.
It is evident from Table 1 that there is a sizable increase in retail margins after the store
brands were introduced in all of the segments except the Health/Nutrition segment. However, the
higher retail margin is not just due to an increase in retail prices. In the Family segment, both
retail and wholesale prices fell, but the reduction in wholesale prices was higher than the than the
reduction in retail prices causing retail margins to rise. Thus prima facie, it appears that the
introduction of store brands tend to have given the retailer the ability to increase its retail
margins.
We explore this issue further by separating out the effects by imitated and non-imitated
brands in Table 2. We expected the imitated brands to be treated more unfavorably than the nonimitated brands and therefore retail margins to increase more for the imitated brands. As
expected, the retail margins increased for the imitated national brands in all segments.
Surprisingly, the retail margins also increased for the non-imitated brands except in the
Health/Nutrition segment. But the percentage changes in markup for the retailer is considerably
higher for the imitated brands, indicating these are being treated more unfavorably than the nonimitated brands.
The overall increase in retail prices did not lead to a decline in sales in all of the
segments. In fact, the national brand sales increased (as seen in Table 1) overall indicating that
the cereal category was growing during this period. This growing demand should have also
contributed to the overall increase in retail prices.
From Table 1, we see that there are only insignificant changes in the level of promotion
(feature/display) induced sales in the family, health and taste enhanced segments, but there is a
doubling of promotional activity in the post-store brand introduction period in the kids segment.

18

It is interesting from Table 2 that there is a significant reduction in promotions for imitated
brands in the family and kids segments, but a significant increase in promotions for the nonimitated brands in these segments. Thus it appears there is a big shift in promotions to the nonimitated brands in these segments.
Insert Tables 1 and 2 here.
It is also instructive to see how the effects of store brand introduction affected each of the
manufacturers. In Table 3, we report the average wholesale prices, retail prices, retail margins
and level of promotion-induced sales by manufacturer. Manufacturers differ in their response to
the store brand introduction. Kelloggs aggressively reduced the wholesale price of its brands.
This should perhaps be expected considering that five of the six imitated brands are from
Kelloggs. General Mills, the manufacturer of Cheerios, the other imitated brand, however raised
its wholesale price. All other firms except Quaker also raised its wholesale prices. However
except in the case of Nabisco, the retailer raised the retail prices. This must be especially galling
for Kelloggs since it had reduced wholesale prices in response to the retailers imitation of 5 of
its leading brands.
Further, Kelloggs and General Mills, the firms that were imitated have a smaller
proportion of their sales when promoted but all the other firms that were not imitated find that
the proportion of their sales when promoted increased after store brand introduction. It is not
easy to separate out why this occurred: Did the manufacturers react by reducing promotional
allowances? Or did the retailer not accept national brands promotions as often as it did before
store brand introduction? Nevertheless it is clear that the store brand introduction reduced the
proportion of the brands on promotion for the largest national brand manufacturers: Kelloggs
and General Mills. Curiously, they also are the manufacturers who were most imitated by store
brands.
Insert Table 3 here
Demand Side Estimates

Our demand side estimates shown in Table 4 have considerable face validity. The mean
coefficient of price is negative as expected and the standard deviation of the price coefficient is
significant indicating that there is heterogeneity in the price sensitivity of customers. Income
reduces price sensitivity, though this coefficient is insignificant. Fiber has a mean positive
19

coefficient, indicating that the population on average values the health benefits of a fibrous diet
that cereal is touted to be. However there is heterogeneity in the valuations of fiber in cereal. Not
surprisingly, people with higher incomes value a diet with high fiber, but kids do not value this.
The surprising result is the Education Variable; college educated consumers seem to not value
the fiber attribute. On average, the presence of sugar reduces the valuation for the product. Kids
consumed more of the sugary cereals. High income and college-educated consumers also sought
sugary cereals. Perhaps it is these consumers who buy the more expensive sugary cereals for
their kids rather than add sugar at home to reduce inconvenience. As expected we obtained a
positive coefficient on promotion, indicating that consumers value promotions. In particular,
seniors value them highly. The coefficient for the interaction term between price and promotions
is negative. Its magnitude reflects that price sensitivity increased around 33% in the presence of
promotions, a result that is consistent with other research (Van Heerde, Leeflang and Wittink,
2001, Sudhir 2001b).
Prior research (Hoch and Banerji 1993, Dhar and Hoch 1997, Starzynsky 1993, and Hoch
1996) identified several demographic characteristics of the population to be correlated with the
success of store brands. By incorporating these demographic variables and accounting for their
effects on demand, we can now be more certain that our inferences about the strategic behavior
of the retailer are not contaminated by unobserved demand side factors.
Like Chintagunta et al. (2002), we also find that there were no significant differences in
demand parameters before and after store brand introduction. This is not surprising because the
demand parameters we estimate are individual level characteristics, which should not be affected
by store brand introduction. We would not however expect the parameters to be the same if we
had estimated a reduced form linear or log-linear model. However the elasticities do increase on
average. We show the average self-price elasticities of the brands in Table 5 classified by
segments before and after store brand introduction. The average elasticity after the store brand
has been introduced is greater (based on a paired t-test of the difference in estimated elasticities
for each brand, p<0.001), indicating that the introduction of the store brand increased the
elasticity of the category itself. See the histogram of the brandwise price elasticity differences in
Figure 2 indicating that there is an average increase in elasticity after the store brand
introduction. However, what is also particularly interesting is that the elasticities of all brands do
not increase. In fact, for some brands the elasticities decline. This could be due to the fact that

20

some of the price sensitive customers who purchased national brands now completely switch to
the store brand and the national brands are now left with only the less price sensitive customers.
This highlights the flexible nature of the random coefficients logit model that accounts for
consumer heterogeneity. Even though consumer parameters and therefore the logit model
estimates do not change after the introduction of the store brand, there is a fairly rich change in
the pattern of changes in the elasticities.
Insert Table 5 here
We also check the nature of the estimated elasticities in two other ways to assess the face
validity of our estimates. First in Table 6, we check whether the cross-elasticity between
Dominicks store brands and its corresponding imitated brands are higher than its cross-elasticity
with respect to other brands. In fact, for all of the brands except for Corn Flakes, this pattern
holds. For example, the cross-elasticity of Dominicks Cheerios with respect to General Mills
Cheerios is 1.786, considerably higher than the cross-elasticity with respect to the other brands.
However, the cross-elasticity of General Mill Cheerios with respect to Dominicks Cheerios is
considerably lower indicating a considerable asymmetry in the nature of the elasticity. This
asymmetric elasticity between store brands and national brands is well documented in the
literature (e.g., Blattberg and Wisnewski 1989, Sethuraman 1995) and indeed our random
coefficients logit model is able to capture this asymmetry well. The cross-elasticity between
Kelloggs Corn Flakes and Dominicks Corn Flakes however is not higher than with respect to
other brands. This suggests that Cornflakes is a staple cereal in the consideration set of most
household and therefore it has high cross-elasticity with respect to all other types of cereals, a
finding that is not very surprising with hindsight.
Insert Table 6 here
We also verify whether the general definition of segments that we define have face
validity based on the estimated demand elasticities. Table 7 shows the average cross price
elasticity within and across the segments we define. In general the higher values in the diagonal
provide support for the segment classification used. The only exception is that Health/Nutrition
demand is affected fairly highly by the changes in prices in the family segment. This is primarily
due to the substantial cross-elasticity between Cheerios (which is a relatively healthy cereal and
promoted as a heart-friendly product in the Family segment) and the brands in the

21

Health/Nutrition segment. Removing the impact of Cheerios from the average reduces the
elasticity to .22.
Insert Table 7 here
Estimates of Strategic Behavior

Pricing of Imitated Versus Non-Imitated Brands: No Control for Segment Differences in Behavior
In Table 8, we report our preliminary estimates about the retailers strategic behavior
without controlling for segment-level differences in retailer behavior. We control for
manufacturer effects and promotion effects. The promotion effect indicates that retailer prices
are lower when the brand is featured or displayed. This is consistent with the fact that the price
sensitivity of consumers are higher in the presence of features and displays, thus making it
sensible for the retailer to reduce margins when a brand is being displayed or featured.
Our results are as follows: Prior to the introduction of the store brand, imitated brands are
favored and prices are lower than for non-imitated brands as indicated by the negative coefficient
on imitation. This is not particularly surprising if we recognize that the imitated brands are the
most popular brands in the category and low prices on these brands are consistent with a pull
strategy of the retailer.
After the store brands are introduced, the non-imitated brands are disfavored as indicated
by the positive coefficient on the Store Brand. The Store Brand coefficient is only marginally
significant at the 10% level. However the Store Brand X Imitated interaction variable is
insignificant, indicating that in contrast to our expectation that imitated brands will be disfavored
in order to favor store brands, we find no significant effect.
Our hypothesis that retailers will favor store brands by raising the prices of national
brands (especially the imitated national brand) also does not appear to have support in this data.
Thus our results seem to be inconsistent with the result found in Chintagunta et al. (2001) that
store brands will be favored by the retailer, relative to national brand. In fact, what is particularly
surprising is that there is no significant disfavor effect on the imitated national brands. The
marginally significant disfavor effect we find is on the non-imitated national brands, which are
not in direct competition with the retailers store brands.

22

As we had suggested earlier in developing our expectations about retailer behavior, we

expected that the retailers behavior might be different across different segments based on the
attractiveness of the segment to the retailer. Specifically, we had suggested that behavior in the
more attractive segments of the market might be more consistent with the strategic support of
store brands, because this is where the strategic behavior might be most rewarding to the retailer,
while the effect may not be significant in the less attractive segments. We now proceed to test
this by investigating if and how the strategic behavior of the retailer is different in different
segments of the market.
Insert Table 8 here
Pricing of Imitated and Non-Imitated Brands: Allowing for Segment Differences in Behavior
To help us assess the segment-wise effects, we report some characteristics of the different
segments in Table 9. We find that the Family and Kids segments have the highest share of sales
volume and also the highest concentration. These two segments have 50% more sales and overall
will be more attractive to the retailer. The two segments contribute to 73.7% of the Herfindahl
index of concentration. It is easier to steal share by imitating a popular brand and using an
against positioning strategy towards a true market leader. Thus disfavoring the market leader
to make the store brand attractive can work better. Also the second highest share brand can also
be better favored by the retailer in order to assuage the manufacturers and still get promotional
support. These segments are therefore more attractive to the retailer in terms of the economic
potential for store brands as well as for strategic behavior by the retailer.
Insert Table 9 here
The coefficients for strategic behavior of the retailer based on our segment-wise analysis
are reported in Table 10. Note that we do not report our demand estimates as these are both
statistically not different and numerically very close to those reported in Table 4. Our estimates
of the coefficients of retailer strategic behavior indicate the following.
First as before, we find that imitated brands are priced lower than non-imitated brands
prior to the introduction of the store brands in all segments except the Kids segment. The
exception in the Kids segment is not surprising when we looked back at what was the imitated
brand in the Kids segment. The imitated brand in the Kids segment was Froot Loops, a premium

23

brand relative to the other Kids brands. This is also indicated by the relatively high price of the
imitated brand in Table 2.
Regarding the effect of store brand introduction on the imitated national brands, we find
that the effects are different across segments. In the Family and Kidss segments, which have the
largest share of sales and have the highest concentration, the positive coefficient (which is highly
significant) on the Store Brand X Imitated indicates that these imitated brands are disfavored. In
contrast, the imitated brands are neither favored nor disfavored in the non-attractive segments.
This result is consistent with our hypothesis that strategic behavior is more likely to happen only
in the more attractive segments of the market, because the retailer may not pay much attention to
the less attractive segments. The deviation effects are graphically described in Figure 3.
Insert Table 10 and Figure 3 here
We next look at the effect of store brand introduction on the non-imitated national
brands. While the coefficient is not significant in any of the segments, the sign of the coefficient
differs in a significant way across the different segments. In the Family and Kids segments, that
are more attractive for strategic behavior, the negative coefficient on the non-imitated brand
indicates that these brands are relatively favored. In the other segments with smaller sales and
more evenly distributed sales across the brands in the segments, the general effect tends to more
positive but much less significant. At least directionally, this is consistent with our expectation,
that in order to assuage the losses of national brand manufacturers who may retaliate against the
retailer by cutting advertising and other promotional support, the retailer may favor such imitated
brands. However this result is insignificant.
We suspected that the insignificance of some results might be due to the large number of
parameters that we estimate in the segment-level estimation. Given, that we could not increase
the number of observations, we needed to reduce the number of parameters estimated in order to
improve the power of our tests. We therefore decided to estimate a model where we group
segments based on the relative attractiveness to the retailer. Based on our measures of
concentration and market size, we group the Family and Kids segments as Attractive segments
and the Health and Nutrition and Taste Enhanced segments as Non-attractive segments based on
a median split of the segments. These estimates are reported in Table 11.
Our results indicate the following: The pooling of the segments for the purposes of
inferring strategic behavior helped in improving the significance of the coefficients. It is clear
24

now that imitated brands are treated more favorably prior to the introduction of the store brands
in the attractive segments, but less favorably so after the introduction of store brands. In contrast,
the non-imitated brands are treated more favorably after the introduction of store brands in the
attractive segments. This is a really surprising result, because general intuition would have
suggested that the entry of store brands would cause harm to national brands. What we find is
that some national brands will be more supported by the retailer after the introduction of the store
brands. In contrast, there were no significant differences due to store brand introduction on either
the imitated or non-imitated brands in the non-attractive segments. These results are well
summarized in the effects we see in Figure 4.
Insert Table 11 here
The Impact of Promotions on Retailer Pricing Behavior
In Table 2, we see that after the introduction of the store brand there is a significant
reduction in promotions for imitated brands in the family and kids segments, but a significant
increase in promotions for the non-imitated brands in these segments. This by itself could imply
that the retailer is helping the introduction of the store brand by shifting promotions efforts
toward the non-imitated brand. Given that we see such systematic differences in patterns of
promotions after the store brand introduction, we wanted to analyze how the retailers pricing
behavior changes when accompanied by promotions. We did this by allowing for the effect of
promotions to vary before and after store brand introduction
We report these results in Table 12. The results indicate that a retailer treats brands that
are promoted more favorably but only after the introduction of the store brand (the negative
coefficient on Promoted X Store Brand). Imitated brands continue to be disfavored in the
attractive segments, indicating that these national brands are being disfavored not only through a
lower frequency of promotions but also through increased prices. However we notice that the
favoring effect for non-imitated brands in the attractive segment (store brand coefficient)
becomes insignificant though it still has the right sign. This is because since imitated brands are
less promoted and non-imitated brands are more promoted, the favoring effect on non-imitated
brands is now captured by the Promotion X Store Brand coefficient. Thus the non-imitated
brands are being favored by means of more frequent promotions accompanied with deeper
discounts. Overall it appears that the retailers provide extra support for promotions after
introducing store brands.
25

The positive and significant coefficient for non-imitated brands in the non-attractive
segments indicates that controlling for promotions these brands are disfavored. However, since
these non-imitated brands are promoted more, and promoted brands are more favored, this
disfavoring effect is neutralized by the benefits from more frequent promotions and overall there
is no significant differences between imitated and non-imitated brands in the non-attractive
segments.
Insert Table 12 here
Robustness Checks

We perform robustness checks to see if the results discussed so far are sensitive to
changes in assumptions and specifications of the model. We tested for an alternative definition of
market potential. Rather than the definition we used in the results discussed, where we assumed
that consumers eat cereals 33% of the time, we tested other percentages such as 28%, 30%, 40%
and 60%. Our results are robust to these alternative assumptions that defined the potential market
size.
Since Kelloggs had five leading brands imitated, we tested for a change in the impact of
the Kelloggs effect after store brand introduction. We did not find any significant effect for this
change variable. Thus any change in pricing behavior by the retailer is explained by the store
brand introduction and not due to fundamental changes in the interactions between Kelloggs and
the retailer.
Chintagunta (2001) suggests the use of store traffic as a proxy for retail competition. If
indeed store traffic proxied for retail competition, then this would imply that on the supply side
retail prices would increase when store traffic increased. Instead our results indicated that retail
prices decline with increase in store traffic. Hence we do not believe that retail competition can
be proxied by store traffic. It is perhaps possible that store traffic is a measure of aggregate
demand and recent literature (e.g. Chevalier et al, 2000) has suggested that increases in aggregate
demand will lead to declines in retail prices. Nevertheless our substantive results about strategic
retailer behavior continue to be the same.

26

7. Conclusion
Summary

In this paper we investigated the strategic behavior of the retailer in enabling the success
of the store brand. Specifically, we investigated how a retailer changes its pricing behavior after
it introduced store brands. We appropriately control for changes in demand and cost conditions
due to the store brand introduction and still identify strategic pricing by the retailer to favor store
brands. Specifically, we adapt the methods used in Berry, Levinsohn and Pakes (1995) and
Nevo (2001) but relax the assumptions of Bertrand behavior used in these papers in order to
investigate the deviations from the category profit maximizing price.
To summarize our results, we found that a nave aggregate approach that does not
account for segment level differences when inferring the strategic behavior of the retailer can
lead to counter intuitive as well as misleading interpretation of retailer behavior. By carefully
accounting for the effects of how a segments attractiveness will impact retailer behavior, we are
able to offer a richer description of how the introduction of a store brand can affect national
brands. Our key results are as follows:
Before introduction of the store brands, the retailer favors the popular brands that are
imitated later by acknowledging the pull power of these leading brands in most segments (the
kids segment was an exception, since the imitated brand was a premium brand). But after
introduction of the store brand, the retailer disfavors these imitated national brands (relative to
pre-store brand introduction period) by charging higher margins in order to support the store
brands. However the retailer favors non-imitated brands compared to before store brand
introduction. This result is surprising in that we might have expected all national brands to suffer
due to store brand introduction. However this differential behavior occurs in only the attractive
segments of the market where such strategic behavior has a greater impact on the profitability of
store brands. In the non-attractive segments, there is no statistically significant differential
treatment of either the imitated or non-imitated national brands before and after the store brand
introduction. We also find that the retailer promotes the imitated national brand less often after
introducing a store brand and further reduces margins when a brand is promoted. Thus the
retailer also disfavors imitated national brands by reducing the extent of promotions (features

27

and displays). The differential is further exacerbated by the fact that promoted brands (which
now tend to be the non-imitated brands) are more favored than non-promoted brands.
Managerial Implications

Our results should be of considerable interest to both researchers and practitioners as

well. Researchers interested in retailing in general and store brands in particular should find
several findings in this paper novel and insightful.
Our analysis indicates that retailers deviate from short-run profit maximizing behavior
both before and after the store brands are introduced. But the deviations are systematically
different before and after the store brand introductions. These results indicate that empirical
research on retailer pricing behavior needs to account for such deviations rather than make the
assumption of single period category profit maximization. Further, researchers studying
passthrough should recognize that extent of passthrough vary systematically for the imitated
brands versus the non-imitated brands. The gains in pass through for non-imitated national
brands in segments where store brands have been introduced are greater than those for the
imitated brands.
These findings have clear implications for practitioners. In the absence of a store brand, a
manufacturer with the most popular brand will have great clout due to the pull power it exerts
to get retailers into the store. However when a competing store brand is introduced to imitate
these popular brands, these brands are disfavored. and the non-imitated brands which were
previously not favored are favored. This implies that after a store brand is introduced, the
popular brand (which is imitated) manufacturer may have lower clout and may need to reduce
prices more in order to maintain market share. This is quite the case of Kelloggs, which reduced
its wholesale prices very substantively to maintain share. However, non-imitated brands may be
able to improve their share without even reducing prices. Our results that non-imitated brands get
lower prices when they are promoted further implies that manufacturers should probably shift
their attention to brands that are not in direct competition to the store brand when promoting
brands.
However while this recommendation is for the short run, the results suggest being more
creative in the long run in terms of its product positioning. Our results indicate that retailers can
use the threat of store brands very successfully against manufacturers when there are strong

28

national brands with large market share from which share can be stolen. With such a threat, it
might be a useful strategy to create a number of variants around the core brand, so that the
retailer would have more difficulty imitating any particular brand and stealing share from such
brands. It would however not be as cost effective for the retailer in terms of shelf space or
manufacturing costs to attack a number of national brand variants. Thus our results provide one
justification for national brand manufacturers to create branded variants.
Coughlan and Choi (2001) have developed a theoretical model of competition between
national brands and private labels. Given the asymmetry in cross-elasticity between national
brands and private labels, they show that national brands will find it useful to invest in brand
asymmetries so as to reduce the comparability of national brands and private labels. Our
empirical results also lead us to recommend that national brands should invest in brand
asymmetry with private labels on the grounds that the retailer strategically disfavors brands that
are comparable to its private label brands.
Limitations and Future Research

We have limited our study to price as the strategic variable for the retailer. The retailer
has control over not only the retail price, but also other marketing mix variables such as shelf
space and position, features, displays and promotions. It could be expected that the retailer may
change its strategy with respect to these variables also so as to facilitate the long-term
penetration of the store brand. Future research needs to address issues related to other strategic
variables. For example, it would be interesting to study how the shelf space is relocated from the
national to the store brands. For example, Hoch, 1996 suggests that because 90 percent of people
are right-handed, the retailer invariably places the store brand to the immediate right of the
leading national brand it is imitating. It would be also important to study how national brands are
compensated by the loss of the shelf space that is taken by the store brands, so that retailers dont
lose advertising and promotional support for the category
Another limitation of our study is that our research focused on one category. We chose
the RTE cereal category for our study because the presence of several segments allowed us to
test for heterogeneous strategic retailer behavior across multiple product segments. It would be
of interest to see if our hypotheses continue hold in other categories where multiple store brands
are introduced in different sub-segments.

29

Future research also needs to consider what other dimensions other than the imitated-non
imitated and attractive-non attractive dimensions are appropriate in explaining the retailers
strategic behavior in setting prices. For example, there are cases where the store brand does not
imitate a national brand. For example Presidents choice from Loblaws in Canada, World
Classics from Topco, and Sams choice from Wal-Mart are not imitations of national brands,
but they are introduced as a high quality differentiated product with respect to national brands. In
other cases (as mentioned by Hoch 1996) the strategy for the low-quality tier is to offer either a
white-label generic or a second store brand (e.g., A&Ps Savings Plus line). Understanding
how the retailers strategic pricing behavior accompanies the introduction of these brands could
be of interest for both researchers and practitioners.
Summarizing, our paper takes an important step in studying the strategic role of the
retailer in enabling the success of store brands. We tested for conditions under which national
brands will be disfavored or favored after a store brand is introduced. We find broad support
for our hypotheses, indicating that the long term success of store brands and long term need for
promotional support drive the strategic behavior of the retailer.

References
Ailawadi, Kusum L, and Bari Harlam (2002), The Effect of Store Brand Share on Retail Margins: An
Empirical Analysis Working Paper.
Berry, Steven T., (1994), Estimating Discrete-Choice Models of Product Differentiation Rand Journal of
Economics, 25(2), 242-262.
Berry, Steven T., James Levinsohn and Ariel Pakes (1995), Automobile Prices in Market Equilibrium,
Econometrica, 63(4), 841-890.
Blattberg, Robert C. and Kenneth J. Wisniewski (1989), Price-Induced Patterns of Competition,
Marketing Science, 8 (Fall), 291-309.
Choi, S. Chan and Anne T. Coughlan, (2001), National Brand/Private Label Competition with Good
Deal Effects: Incentives for Investment in Brand Asymmetry, Working Paper.
Chevalier, Judith A., Anil Kashyap, and Peter Rossi (2000), Why Don't Prices Rise During Periods of
Peak Demand: Evidence from Scanner Data, Working Paper, University of Chicago.
Chintagunta, Pradeep K., (2002), Investigating Category Pricing Behavior at a Retail Chain, Journal of
Marketing Research, 39, 2, 141-154
30

Chintagunta, Pradeep K., Andre Bonfrer and Inseong Song, (2002), Investigating the Effects of Store
Brand Introduction on Retailer Demand and Pricing Behavior, Management Science, 48, 10,
2002.
Corstjens, Marcel and Rajiv Lal, (2000), Building Store Loyalty Through Store Brands, Journal of
Marketing Research, 37, (August), 281-291.
Cotterill, Ronald W., William P. Putsis, and Ravi Dhar (2000), Assessing the Competitive Interaction
between Store brands and National Brands, Journal of Business, Vol 73, No.1, 109-137.
Dhar, Sanjay K. and Stephen J. Hoch (1997), Why Store Brand Penetration Varies by Retailer,
Marketing Science, 16(3), 208-227.
Dube Jean-Pierre H., Pradeep Chintagunta, Bart Bronnenburg, Ron Goettler, Amil Petrin, P.B.
Seetheraman, K. Sudhir, Raphael Thomadsen and Ying Zhao Structural Applications of the
Discrete Choice Model, Marketing Letters, 13:3, 207-220, 2002.
Giblen, G. M. (1993). Summit Conference Defines Future. Grocery Marketing, 53 (November), 32-37.
Hoch, Stephen J. (1996), How Should National Brands Think about Store Brands? Sloan Management
Review, 37, 2, 89-102.
Hoch, Stephen J. and Shumeet Banerji (1993), When Do Store Brands Succeed? Sloan Management
Review, 34, 4, 57-67.
Hoch, Stephen J., Alan L. Montgomery, and Young-Hoon Park, (2001), Long-Term Growth Trends in
Private Label Market Shares, Working Paper.
Kadiyali, Vrinda, N. Vilcassim, and Pradeep K. Chintagunta (2000), Power in Manufacturer-Retailer
Interactions: An Empirical Investigation of Pricing in a Local Market, Marketing Science, 19, 2,
127-148.
Mathews, Ryan (1996), The New Age of Retailer Controlled Brands, Progressive Grocer, special
report (November) pp. A2-A19
McMaster, D.,(1987), Own Brands and the Cookware Market, European J. Marketing, 1,21, 83-94.
Narasimhan, Chakravarthi and Ronald T. Wilcox (1998) Store Brands and Channel Relationship: A
Cross-Category Analysis, Journal of Business, Vol 71, No 4, 573-600
Nevo, Aviv (2000), A Practitioners Guide to Estimation of Random Coefficients Logit Models of
Demand, Journal of Economics & Management Strategy, 9, N 4, 513-548.
Nevo, Aviv (2001), Measuring Market Power in the Ready to Eat Cereal Industry Econometrica, 69, N
2, 307-342.
Pauwels Koen and Shuba Srinivasan, (2002), Who Benefits from Store Brand Entry? Working Paper,
Dartmouth College.

31

Raju, Jagmohan, Raj Sethuraman, and Sanjay K. Dhar, (1995), The Introduction and Performance of
Store Brands, Management Science, 41, 6, 957-978.
Sayman, Serdar, Stephen J. Hoch, and Jagmohan Raju (2001), Positioning of Store Brands, Working
Paper.
Schmalensee, Richard, (1978), Entry Deterrence in the Ready-to-Eat Breakfast Cereal Industry Bell
Journal of Economics, 9, 305-327.
Scott-Morton, Fiona, and Zettelmeyer, Florian, (2001).The Strategic Position of Store Brands in Retailer
Manufacturer Negotiations Working Paper, Yale University.
Sethuraman, Raj (1995), A Meta-Analysis of National Brand and Store Brand Cross-Promotional Price
Elasticities Marketing Letters 6:4, 275-286.
Sethuraman, Raj (2000), What Makes Consumers Pay more for National Brands than for Store Brands
Image or Quality? Working Paper, Southern Methodist University.
Starzynski, G (1993), The Private Label Consumer: Is there One? Northbrook, IL: AC Nielsen.
Sudhir, K (2001a), Competitive Pricing Behavior in the US Auto Market: A Structural Analysis
Marketing Science, winter, 2001, pp. 42-60.
Sudhir, K (2001b), Structural Analysis of Competitive Pricing in the Presence of a Strategic Retailer,
Marketing Science, Summer, 2001, pp. 244-264.
Van Heerde, Harald J., Peter S. H. Leeflang and Dick R. Wittink (2001), Semiparametric Analysis to
Estimate the Deal Curve, Journal of Marketing Research, Vol 38, 2, 197-215.
Villas-Boas J. Miguel, and Ying Zhao, (2001) The Ketchup Marketplace: Retailer, Manufacturers, and
Individual Consumers," Working Paper, University of California, Berkeley.
Villas-Boas , Sofia B., (2000), Vertical Relationships between Manufacturers and Retailers: An Empirical
Analysis, Working Paper, University of California, Berkeley.

32

Table 1
Descriptive Statistics of the Data by Segments

Segments

Period

Wholesale
Price ($)

Margin ($)

Price ($)

Proportion
of Brands on
Promotion

Weekly
Sales ($)

Weekly
Sales (Oz)

Family

Before SB

0.1393

0.0219

0.1612

12.9%

182,447

1,131,950

SB Intro

0.1405

0.0257

0.1661

12.5%

165,048

993,405

After SB

0.1295

0.0265

0.1560

12.1%

188,036

1,205,341

Before SB

0.1687

0.0272

0.1959

5.7%

89,794

458,266

Kids

SB Intro

0.1661

0.0257

0.1918

21.1%

106,156

553,377

After SB

0.1689

0.0284

0.1973

11.8%

117,176

593,920

Health

Before SB

0.1455

0.0242

0.1697

3.6%

62,155

366,362

& Nutrition

SB Intro

0.1544

0.0216

0.1760

7.9%

61,538

349,649

After SB

0.1530

0.0237

0.1767

3.4%

79,132

447,832

Taste

Before SB

0.1429

0.0207

0.1636

9.1%

43,215

264,135

Enhanced

SB Intro

0.1413

0.0261

0.1674

31.2%

54,863

327,736

After SB

0.1392

0.0251

0.1643

10.8%

56,839

346,044

33

Table 2
Descriptive Statistics of the Data by Segments and Imitated vs Non-imitated

Segment
Family

Period

Wholesale
Margin ($)
Price ($)

Price ($)

Proportion of
Brands on
Promotion

Weekly
Sales ($)

Weekly
Sales (Oz)

Before SB

0.1490

0.0264

0.1754

7.48%

65,670

374,416

SB Intro

0.1499

0.0312

0.1811

21.66%

84,317

465,632

After SB

0.1534

0.0284

0.1818

11.46%

83,154

457,370

Before SB

0.1678

0.0271

0.1949

1.51%

76,326

391,716

SB Intro

0.1649

0.0257

0.1906

22.76%

98,190

515,111

After SB

0.1681

0.0280

0.1961

11.85%

106,505

543,139

Health &

Before SB

0.1438

0.0242

0.1680

3.85%

55,732

331,701

Nutrition

SB Intro

0.1533

0.0212

0.1745

8.74%

55,299

316,838

(Non Imitated)

After SB

0.1515

0.0230

0.1745

3.87%

69,655

399,073

Taste

Before SB

0.1538

0.0234

0.1773

11.60%

27,751

156,552

Enhanced

SB Intro

0.1467

0.0295

0.1762

42.30%

40,466

229,720

(Non Imitated)

After SB

0.1510

0.0284

0.1794

8.16%

38,166

212,763

Family

Before SB

0.1345

0.0197

0.1542

16.34%

116,776

757,534

SB Intro

0.1322

0.0208

0.1530

5.10%

80,731

527,774

After SB

0.1148

0.0254

0.1402

8.70%

104,882

747,971

Before SB

0.1740

0.0283

0.2024

29.75%

13,468

66,551

SB Intro

0.1820

0.0261

0.2082

0.00%

7,965

38,266

After SB

0.1784

0.0317

0.2101

11.66%

10,671

50,781

Health &

Before SB

0.1613

0.0239

0.1853

1.06%

6,422

34,661

Nutrition

SB Intro

0.1649

0.0253

0.1902

0.00%

6,240

32,812

(Imitated)

After SB

0.1653

0.0291

0.1944

0.00%

9,477

48,759

Taste

Before SB

0.1270

0.0167

0.1437

4.70%

15,463

107,583

Enhanced

SB Intro

0.1286

0.0183

0.1469

0.00%

14,397

98,016

(Imitated)

After SB

0.1203

0.0198

0.1401

16.31%

18,673

133,281

(Non Imitated)
Kids
(Non Imitated)

(Imitated)
Kids
(Imitated)

34

Table 3
Descriptive Statistics of the Data by Manufacturers

Price ($)

Proportion
of Brands on
Promotion

Weekly
Sales ($)

Weekly
Sales (Oz)

0.0191

0.1571

9.3%

166,218

1,057,937

0.1403

0.0188

0.1592

6.6%

133,706

840,026

After SB

0.1274

0.0228

0.1502

7.9%

166,574

1,108,757

Before SB

0.1650

0.0281

0.1931

10.4%

158,386

820,254

SB Intro

0.1688

0.0282

0.1970

21.5%

177,432

900,767

After SB

0.1670

0.0304

0.1974

7.4%

191,937

972,452

Before SB

0.1322

0.0209

0.1531

7.3%

17,543

114,561

SB Intro

0.1381

0.0298

0.1679

0.0%

22,711

135,290

After SB

0.1446

0.0265

0.1711

17.6%

33,733

197,141

Before SB

0.1250

0.0234

0.1484

5.5%

25,258

170,192

SB Intro

0.1197

0.0309

0.1507

43.1%

45,738

303,569

After SB

0.1213

0.0253

0.1466

15.2%

35,093

239,388

Before SB

0.1616

0.0422

0.2038

13.0%

4,414

21,661

SB Intro

0.1825

0.0340

0.2165

0.0%

2,880

13,302

After SB

0.1742

0.0296

0.2038

34.4%

6,812

33,427

Before SB

0.1322

0.0282

0.1604

2.5%

5,791

36,109

SB Intro

0.1405

0.0242

0.1646

0.0%

5,139

31,213

After SB

0.1435

0.0241

0.1676

7.2%

7,034

41,972

Wholesale
Margin ($)
Price ($)

Manufacturer

Period

Kellog's

Before SB

0.1380

SB Intro
General Mills

Post

Quaker

Ralston

Nabisco

35

Table 4
Demand Parameters

Mean

Sigma

Income

Senior

Education
(College)

Children
(<9)

Constant

0.7332
(0.0329)

-0.4129
(0.6054)

-8.2529
(0.1539)

5.8923
(0.4713)

5.0584
(0.2548)

Price

-63.6998
(0. 9561)

-1.6965
(1.1296)

1.8722
(1.1446)

Fiber

0.1327
(0.0020)

0.1062
(0.0393)

0.2765
(0.0058)

-2.8359
(0.5678)

-4.7172
(0.0505)

0.2462
(0.0577)

Sugar

-0.1739
(0.0032)

0.6429
(0.0130)

2.0085
(0.0325)

0.0850
(0.0385)

Promotion

2.3012
(0.9608)

1.2578
(0.6906)

7.6352
(0.6080)

Price x
Promotion

-23.7340
(5.9455)

36

Table 5
Self-Price Elasticities

Segments
Family

Kids

Health & Nutrition

Taste enhanced

Imitated /
Before SB
non imitated

After SB

Average

Non Imitated

-10.62

-11.34

-11.01

Imitated

-8.80

-8.95

-8.88

Non Imitated

-11.85

-12.14

-12.01

Imitated

-11.70

-12.33

-12.04

Non Imitated

-10.85

-11.45

-11.18

Imitated

-9.14

-10.08

-9.65

Non Imitated

-9.93

-10.26

-10.11

Imitated

-7.09

-6.84

-6.95

37

Table 6
Self & Cross-Price Elasticities
Store Brands and Imitated National Brands

Dominicks
Tasteeo's

Dominicks Corn
Flks

Dominicks
Crispy Rice

Dominicks Frost
Flk

Dominicks Fruit
Rings

Dominicks Rais
Bran

Cheerios (GM)

Corn Flakes
(Kellogg's)

Rice Krispies
(Kellogg's)

Frosted Flakes
(Kellogg's)

Froot Loops
(Kellogg's)

Raisin Bran
(Kell, Pst, GM)

Table shows % changes in market share of brands in columns due to % changes in prices of brands in rows.

Dominicks Tasteeo's

-8.172

0.061

0.059

0.043

0.000

0.000

0.118

0.069

0.072

0.063

0.000

0.000

Dominicks Corn Flakes

0.053

-4.730

0.119

0.176

0.000

0.000

0.046

0.091

0.086

0.072

0.000

0.000

Dominicks Crispy Rice

0.053

0.114

-7.538

0.118

0.000

0.000

0.049

0.095

0.091

0.079

0.000

0.000

Dominicks Frosted Flakes

0.050

0.295

0.245

-7.089

0.000

0.000

0.042

0.104

0.103

0.132

0.000

0.000

Dominicks Fruit Rings

0.000

0.000

0.000

0.000

-9.104

0.000

0.000

0.000

0.000

0.000

0.338

0.000

Dominicks Raisin Bran

0.000

0.000

0.000

0.000

0.000

-5.534

0.000

0.000

0.000

0.000

0.000

0.592

Cheerios (GM)

1.786

0.923

0.908

0.676

0.000

0.001

-10.90

0.948

1.024

0.885

0.000

0.000

Corn Flakes (Kellogg's)

0.530

0.918

0.833

0.612

0.000

0.000

0.498

-6.248

0.891

0.805

0.000

0.000

Rice Crispies (Kellogg's)

0.617

0.962

0.890

0.684

0.000

0.000

0.590

0.930

-10.38

0.808

0.000

0.000

Frosted Flakes (Kellogg's)

0.700

1.016

0.964

0.822

0.000

0.000

0.671

1.137

1.035

-9.325

0.000

0.000

Froot Loops (Kellogg's)

0.000

0.000

0.000

0.000

0.961

0.000

0.000

0.000

0.000

0.000

-12.17

0.000

Raisin Bran (Kell, Post, GM)

0.000

0.000

0.000

0.000

0.000

1.347

0.000

0.000

0.000

0.000

0.000

-7.850

38

Table 7
Average Cross-Price Elasticities by Segment

Family

Kids

Health &
Nutrition

Taste enhanced

Family

0.4624

0.1174

0.3198

0.0405

Kids

0.0771

0.3646

0.0170

0.0031

Health &
Nutrition

0.1401

0.0228

0.2904

0.0973

Taste enhanced

0.0110

0.0029

0.1308

0.6071

Table 8
Strategic Pricing Behavior with respect to imitated/non-imitated brands
Main Effect + Control
Variable

Coef

SD

T-stat

Kellogg's

-0.032

0.0013

-24.8288

GM

-0.0264

0.0011

-23.3684

Post

-0.0459

0.0024

-19.2774

Quaker

-0.0172

0.0017

-10.1119

Ralston

-0.0284

0.0044

-6.4203

Nabisco

-0.0313

0.0015

-20.8525

Promotion

-0.0377

0.0089

-4.2269

Imitated National Brand

-0.0201

0.0024

-8.2497

Store Brand Introduced

0.002

0.0012

1.6775

S B Introduced x Imitated

0.0027

0.0033

0.8086

GMM objective = 0.001082

39

Table 9
Segment Attractiveness

Segments

Family
Kids
Health/Nutrition
Taste Enhanced
Attractive Segments
Non-attractive
Segments

Share of Margin (US $)

Segment in
Category
35.4%
0.0246
25.1%
0.0277
19.7%
0.0237
19.9%
0.0233
0.0258
60.42%
0.0235
39.58%

Number of
Brands

17
20
17
25
37
42

Contribution to
Herfindahl
Index
54.6%
19.1%
15.3%
10.9%
73.7%
26.3%

40

Table 10
Strategic Pricing by Segments
By Segments + Control

Control

Family

Kids

Health

Taste enhanced

Coef

SD

T-stat

Kellogg's

0.004

0.0017

2.29

GM

0.0166

0.0018

9.13

Post

0.0032

0.003

1.09

Quaker

0.027

0.0026

10.41

Ralston

0.0089

0.0052

1.72

Promotion

-0.0462

0.0106

-4.37

Intercept

-0.0322

0.0024

-13.19

Imitated National Brand

-0.0008

0.0021

-0.38

Store Brand Introduced

-0.0014

0.0025

-0.57

S B Introduced x Imitated

0.0053

0.003

1.75

Intercept

-0.0429

0.0019

-22.96

Imitated National Brand

0.0118

0.0039

3.01

Store Brand Introduced

-0.0037

0.0024

-1.53

S B Introduced x Imitated

0.0162

0.0052

3.14

Intercept

-0.041

0.0026

-15.98

Imitated National Brand

-0.0493

0.0039

-12.78

Store Brand Introduced

0.0023

0.0027

0.86

S B Introduced x Imitated

0.0012

0.0055

0.22

Intercept

-0.0322

0.0016

-20.04

Imitated National Brand

-0.0519

0.0057

-9.13

Store Brand Introduced

0.0008

0.0016

0.48

S B Introduced x Imitated

0.0016

0.0077

0.20

41

Table 11
Strategic Pricing by Segment Attractiveness
Segments Attractivness + Control
Coef

SD

T-stat

Kellogg's

-0.0009

0.0016

-0.57

GM

0.0137

0.0016

8.63

Post

-0.0012

0.0027

-0.42

Quaker

0.0221

0.0026

8.61

Ralston

0.0124

0.0059

2.08

Promotion

-0.0467

0.0117

-3.99

Intercept

-0.0301

0.0017

-17.70

Attractive

Imitated National Brand

-0.0004

0.0018

-0.20

Segments

Store Brand Introduced

-0.0055

0.0017

-3.20

S B Introduced x Imitated

0.0113

0.0025

4.52

Intercept

-0.0322

0.0015

-22.14

Non-Attractive

Imitated National Brand

-0.0529

0.0032

-16.48

Segments

Store Brand Introduced

0.0015

0.0014

1.06

S B Introduced x Imitated

0.0026

0.0045

0.58

Control

42

Table 12
Strategic Pricing by Segment Attractiveness

(Controlling for Promotional Effectiveness after Store Brand Introduction)

Segments Attractiveness + Control
Coef

SD

T-stat

Kellogg's

-0.0023

0.0018

-1.25

GM

0.0121

0.0017

7.07

Post

-0.002

0.0029

-0.71

Quaker

0.022

0.0027

8.10

Ralston

0.0108

0.0061

1.77

Promotion

0.0046

0.003

1.49

Promot x SB

-0.0809

0.0182

-4.45

Intercept

-0.0318

0.0018

-17.31

Attractive

Imitated National Brand

0.001

0.0017

0.58

Segments

Store Brand Introduced

-0.0015

0.0012

-1.19

S B Introduced x Imitated

0.0114

0.0025

4.50

Intercept

-0.0329

0.0015

-21.79

Non-Attractive

Imitated National Brand

-0.0548

0.0032

-17.07

Segments

Store Brand Introduced

0.0041

0.0015

2.64

S B Introduced x Imitated

0.0033

0.0046

0.71

Control

43

Figure 1
Average Retail Prices, Wholesale Prices and Margins during Period of Analysis

0.21
0.2

Retail Price

0.19
0.18
0.17

Wholesale Price

0.16
0.15

10

20

30

40

50

60

0.03
0.029
0.028

Margin

0.027
0.026
0.025
0.024
0.023

10

20

30

Store Brand Introduction

40

50

60

Weeks

44

Figure 2

Histogram of Differences in Before and After Elasticities

(with Ho and 95% t-confidence interval for the mean)

Frequency

Ho
-1.0

-0.5

0.0

_
X

0.5

1.0

1.5

2.0

2.5

Differences

45

Figure 3
Pricing behavior before and after the Store Brand Introduction by Segments (cents/Oz)
Attractive segments

Family

Kids

-2.5

-1
-2

-3

-3
-4
-5

-3.5
Before SB
imitated

Before SB

After SB

imitated

non-imitated

After SB
non-imitated

Non-attractive segments

Taste enhanced

Health and Nutrition

-2
-4
-4
-6

-6

-8

-8
-10

-10
Before SB
imitated

After SB
non-imitated

Before SB
imitated

After SB
non-imitated

46

Figure 4
Pricing behavior before and after the Store Brand Introduction by Segment Attractiveness
(cents/Oz)

Attractive Segment

Non-Attractive Segment

-2

-2

-2.5

-4

-3

-6

-3.5

-8

-4

-10

Before SB
imitated

After SB
non-imitated

Before SB
imitated

After SB
non-imitated

47