ArticlePDF Available

Suppliers Caught in Supermarket Price Wars: Victims or Victors Insights from a Dutch Price War


Abstract and Figures

During retailer-initiated price wars (PWs), hundreds of brands are involved simultaneously, affecting brands’ and retailers’ positioning and ultimately making the performance outcome for individual brands difficult to predict. Likewise, the impact on brand performance after the PW, when prices are restored, is unclear. The authors use a natural experiment approach to track brand sales and shares before, during, and after a long-lasting supermarket PW in the Dutch grocery market. They find that PWs are not truly revenue, sales, or share generators for most brands, unless prices remain reduced permanently by the retailer. Only after the PW, when rivals’ prices are restored and the focal brand’s reduced retail price is maintained, can substantial sales, revenues, and share gains be realized. Moreover, restoring prices without additional price promotion support can severely damage brands’ performance. Overall, national brands can gain share, sales, and revenue, but at the cost of not restoring regular prices, while private labels can benefit even when prices are restored after the PW ends.
Content may be subject to copyright.
During retailer-initiated price wars (PWs), hundreds of brands are involved
simultaneously, affecting brandsand retailerspositioning and ultimately
making the performance outcome for individual brands difcult to predict.
Likewise, the impact on brand performance after the PW, when prices are
restored, is unclear. The authors use a natural-experiment approach to track
brand sales and shares before, during, and after a long-lasting supermarket
PW in the Dutch grocery market. They nd that PWs are not truly revenue,
sales, or share generators for most brands unless prices remain reduced
permanently by the retailer. Only after the PW, when rivalsprices are restored
and the focal brands reduced retail price is maintained, can substantial sales,
revenues, and share gains be realized. Moreover, restoring prices without
additional price promotion support can severely damage brandsperformance.
Overall, national brands can gain share, sales, and revenue, but at the cost of
not restoring regular prices, while private labels can benetevenwhenprices
are restored after the PW ends.
Keywords: retailing, price wars, retailersupplier relationships, consumer
packaged goods
Online Supplement:
Suppliers Caught in Supermarket Price Wars:
Victims or Victors? Insights from a Dutch
Price War
In an effort to drive trafc, grocery retailers often initiate
price wars (PWs) on national brands (NBs) (Heil and
Helsen 2001). In general, supermarket PWs entail sustained
reductions of regular prices1of hundreds of brands across
categories and retailers, affecting several brand competitors
within the same category. For example, in 2010, Wal-Mart
started a soda warby reducing the preferred manufacturer
price from $6 to $5 for a case of Coca-Cola, Pepsi, and other
top brands, forcing grocery competitors to follow suit
(Tuttle 2010). These reductions turned out to be benecial
for the involved brands; demand was so strong that sup-
pliers could barely keep up. In the United Kingdom, Tesco,
Sainsburys, and Asda slashed prices storewide on thou-
sands of items, again involving many brands (Grierson
2011). A Tesco executive noted that manufacturers should
welcome these retailer-initiated cuts because they have the
potential to boost brand volume at those retailers (Hegarty
2011). Still, in Australia, a PW between the two largest
players in the market, Woolworth and Coles, reportedly has
taken its toll on both local and international manufacturers,
including Heinz and Unilever. Heinz even reported a sales
drop of 13% in 2011 and proclaimed Australia to be its
worst market (Maxwell 2011). Because of the mixed
outcomes of PWs, brand manufacturers often feel trapped
*Francesca Sotgiu is Assistant Professor of Marketing, Vrije Universiteit
Amsterdam (e-mail: Katrijn Gielens is Associate Professor of
Marketing and Sarah Graham Kenan Scholar, University of North Carolina at
Chapel Hill (e-mail: The authors gratefully ac-
knowledge Alfred Dijs (AiMark) for providing the data. They also thank
Marnik Dekimpe and Marc Vanhuele for their helpful comments and sugges-
tions. Special thanks are due to Marnik Dekimpe, Maarten Gijsenberg, Jan-
Benedict Steenkamp, and Harald van Heerde for kindly providing advertising
data for the U.K. market. Peter Danaher served as associate editor for this article .
1We use the term sustainedto indicate that the price reductions are not
typical temporary price promotions but are maintained for the entire length of
the PW, if not longer. The distinction between a sustained price reduction
and a temporary price promotion is usually clearly communicated to con-
sumers. For example, in our setting, retailers informed consumers of their
intentions to maintain the prices at the reduced level in the long run,
communicating them as permanentlower prices.
© 2015, American Marketing Association Journal of Marketing Research
ISSN: 0022-2437 (print) Ahead of Print
1547-7193 (electronic) DOI: 10.1509/jmr.13.01801
by retailers engaging in PWs and even try to avoid getting
(further) involved. For example, in the Dutch PW that is the
topic of the current research (see Van Heerde, Gijsbrechts,
and Pauwels 2008), Campina, the market leader in most
dairy categories, increased its wholesale prices in the hope
of staying out of the PW (Van Aalst et al. 2005).
So far, no systematic evidence exists on how supermarket
PWs alter the performance of brands trapped in these wars.
Extant literature on PWs (e.g., Van Heerde, Gijsbrechts, and
Pauwels 2008) has mainly focused on the outcomes for the
initiators of the PW, which, in a grocery setting, means su-
permarket retailers. In this study, we shift the focus to the third
parties involved in PWsthat is, the brands whose prices are
used to ght the supermarketPW. On the one hand, brands can
be beneciaries of PWs because absolute price reductions
stimulate demand at the retailers involved in those wars. If
brands indeed thrive during PWs (as retailers claim), suppliers
may be willing to embrace PWs. On the other hand, these
volume boosts do not always materialize. Because a retailer
reduces prices not for individual brands in a category butrather
sometimes for a full set of direct competitors, relative prices
could remain fairly stable, and absolute price reductions may
not lead to the expected volume sales boost, thus resulting in
lower revenue sales at that retailer. In summary, because the
competitive position of individual brands may not improve, the
outcome of PWs for individual brands2is difcult to predict.
Moreover, during PWs, supermarkets do not directly involve
all brands in a category immediately or simultaneously. Usually,
retailers announce storewide sustained price cuts for a sub-
stantial, though not fully exhaustive, set of brands in different
categories. For example, in the Dutch PW, Albert Heijn, the
market leader and initiator of the PW, did not begin reducing the
price of some category leaders (e.g., Lays chips, Heinz ketchup)
until more than a year after the start of the PW. Thus, even when a
brands absolute price remains untouched, its relative price
positioning within the category and store may deteriorate. In this
article, we refer to a PW price drop (PW_PDrop) or a PW price
constant (PW_PCst) setting, depending on whether the retailer
decreased a brands price and thus used the brand to ght the PW.
Although a brands absolute price level does not change in a
PW_PCst setting, its relative price positioning is still affected.
To correctly assess the impact of PWs, we acknowledge these
differences between PW_PDrop and PW_PCst scenarios.
To complicate matters further, when PWs end, retailers are
tempted to reverse the price cuts they initially announced as
permanentduring the PW. However, even less is known
about brand performance when the PW ends, and thus retailers
need to be cautious. For example, although Wal-Martssoda
war price cuts were benecial initially, retailers realized that
the ongoing cuts could undermine their efforts to restore prices
over time (Tuttle 2010). Thus, can brand prices be reestablished
easily to their pre-PW levels without negatively affecting
performance? Previous research on temporary deal retraction
(e.g., Kahn and Louie 1990; Wathieu, Muthukrishnan, and
Bronnenberg 2004) suggests that when prices are restored,
market shares can fall below their initial levels. To provide a
complete picture of PW consequences, we examine the
implications of brandsinvolvement after the PW, when the
retailer decides to increase brand prices again, a scenario we
label the post-PW price lift setting (post-PW_PLift). We
contrast this with a setting in which the retailer maintains
prices at the PW level after the PW, a setting we refer to as the
post-PW price constant scenario (post-PW_PCst).
In doing so, we also shed more light on which types of brands
stand to lose or win more share, sales, and revenues at retailers
involved in PWs. Previous work (e.g., Gupta and Cooper 1992)
has demonstrated that consumers tend to amplify or discount
(temporary) price changes depending on several brand- and
retailer-related contextual cues. These effects may also occur when
dealing with more sustained price changes, as in a PW setting.
More specically, the brands ownership (NBs vs. private labels
[PLs]) and positioning, its promotion and communication efforts,
and the retailers positioning can all interfere with the impact of
PW involvement. Combined, insights into these moderating effects
may afford deeper understanding of (1) what type of brands may
stand to gain or lose more from involvement in a PW and (2) how
brands can avoid potential downsides of retailer-initiated PWs.
In summary, this article aims to contribute to the existing
literature by (1) investigating the impact of retailer-initiated
PWs on brand performance; (2) distinguishing between
brands whose prices are kept constant, decreased, and
increased; (3) capturing the evolution of brand performance
before, during, and after a PW; and (4) controlling for the
moderating impact of brandsand retailerspositioning and
actions. To illustrate the different PW scenarios, we turn to
the soft drink category at a Dutch retailer during the PW
between 2003 and 2005 (see Figure 1).
At the start of the PW, the retailer immediately reduced
the price of both the leading NB (NB1) and its PL, leaving
the price of the second NB (NB2) untouched. Thus, the PL
and NB1 entered the PW_PDrop setting immediately, while
NB2 remained in a PW_PCst scenario until the retailer
reduced its price as well. During this PW_PCst scenario,
NB2 faced a new, less favorable pricing position than NB1,
and the price gap relative to the PL increased. When the PW
ended, the retailer increased NB1s price and left the price
of NB2 and the PL untouched. This move signaled the end
of the PW in the category, and NB1 entered the post-
PW_PLift setting, while NB2 and the PL entered the
post-PW_PCst scenario. Finally, NB2 and the PL stayed in
this post-PW_PCst stage until the retailer increased their
prices as well. While the prices of the PL and NB2 stayed
below their pre-PW level, NB1s price surpassed this level.
To examine the impact of these PW settings, we use a
natural-experiment approach, in which we track brands before,
during, and after a long-lasting supermarket PW. More spe-
cically, we assess the performance of 162 brands in 25 cat-
egories at 5 national retailers in the Netherlands between 2002
and 2007. In this time span, a major PW started and ended.
PW Literature
Extant literature has dened the PW phenomenon, ex-
amined its antecedents, and focused on the performance
2In this study, we focus mainly on brandsperformance within each re-
tailer, though we acknowledge that brands are also concerned about their
position in the entire market. However, by conducting a simulation at the
national level (in the Total Revenue Effect of PW Scenario Sequences
subsection), we also shed some light on the overall brand demand. In the
remainder of the article, we refer to sales and share at the retailer level, with
the exception of the section on the simulation results. Moreover, we use the
word saleswhen referring to volume sales and revenueswhen referring
to dollar sales throughout the text. Sharealways refers to volume share.
consequences for direct participants. Heil and Helsen
(2001) were among the rst to outline the criteria for on-
going competitive price interactions to qualify as PW be-
havior.3Furthermore, the PW literature provides insights
into three main antecedents of PWs: (1) economic down-
turns (e.g., Green and Porter 1984) and expansions (e.g.,
Rotemberg and Saloner 1986), (2) competitive entry (e.g.,
Elzinga and Mills 1999; Milgrom and Roberts 1982), and
(3) nancial (e.g., Busse 2002) and market share (Grifth
and Rust 1997; Leeang and Wittink 1996) conditions. How-
ever, although the consequences for the initiators of PWs have
been explored, the insights are often inconclusive: Some studies
emphasize potential revenue losses (Brandenburger and
Nalebuff 1996), while others identify conditions when partici-
pants may gain (Busse 2002; Elzinga and Mills 1999).
To the best of our knowledge, the only marketing study,
other than Heil and Helsens (2001), to investigate the con-
sequences of PWs is that by Van Heerde, Gijsbrechts, and
Pauwels (2008). The authors nd that during supermarket
PWs, price awareness and price sensitivity increase, and price
becomes a more important purchase criterion driving consumers
decisions of what and where to buy. They also provide evidence
that as consumer basket sizes increase, retailers may still gain
from a PW, though not every retailer stands to win or lose
equally. The resulting increased price sensitivity mainly benets
traditional price ghters, while it harms more conventional re-
tailers. Indeed, when price ghters participate in PWs, the
boundaries of the options available to consumers along the price
dimension may expand downward. In the eyes of consumers,
these price ghters may become even cheaper than their com-
petitors, and thus the perceived expensiveness of the traditional
supermarkets may increase (see Cunha and Shulman 2011).
Regular supermarkets that try to mimic price ghters with
sustained price cuts may end up as suboptimal options, or de-
coys, further reinforcing the price ghtersdominant position
(Huber, Payne, and Puto 1982). In summary, the PW literature
has (1) mainly focused on the causes leading to a PW, only
partially covering consequences; (2) taken the point of view of
the initiators, which in the case of supermarket PWs are retailers,
not individual brands; and (3) identied moderating effects of
price positioning (i.e., price ghters are the PW winners).
Translating these insights to our third-party brand setting sug-
gests that economy-priced brands could gain more from PWs.
Price Promotion Literature
The vast literature on price promotions has extensively
detailed how temporary price cuts affect brand performance
Figure 1
ular price PL Re
ular price leadin
brand NB1 Re
ular price NB2
Regular Price
2003 February
2003 April
2003 June
2003 August
2003 October
2003 December
2004 January
2004 March
2004 May
2004 July
2004 September
2004 November
2004 December
2005 February
2007 February
2007 March
2007 May
2007 July
2007 September
2007 November
2005 April
2005 June
2005 July
2005 September
2005 November
2006 January
2006 March
2006 April
2006 June
2006 August
2006 October
2006 December
Notes: The gure is a stylized representation of actual data for three leading brands in the soft drinks category at a leading Dutch retailer before, during, and after
a PW that took place from 2003 until the end of 2005. The regular price refers to the nonpromotional price.
3Accordingto Heil and Helsen (2001), a PW occurswhen one or more of the
following criteria are met: (1) the direction of the pricing is downward; (2) the
pricing interaction among the actors occurs at a much faster rate than normal;
(3) the actions and reactions focus mainly on the competitors, rather than on the
customers; (4) the competitive interaction violates industry norms; (5) the
pricing interaction as a whole is undesirable to competitors; (6) the competitors
neither intended nor expected to ignite the PW through their preceding
competitive behavior; and (7) this pricing interplay is unsustainable.
Insights from a Dutch Price War 3
(for an overview, see Neslin and Van Heerde 2009). In this
study, we borrow from this body of work to infer how con-
sumers subjectively encode or interpret sustained price changes
accompanying the different PW scenarios (Gupta and Cooper
1992; Winer 1986). Consumers evaluate and encode price
information, and it is their perceptions of the information, not
the information itself, that affect their behavior (Mazumdar,
Raj, and Sinha 2005). In general, consumersperceptions of
price are dependent not only on the actual price but also on their
adaptation level of what that price should be (Monroe 1973). As
such, consumers can either discount or amplify price reductions
or increases (Gupta and Cooper 1992; Pauwels, Srinivasan, and
Franses 2007). Price reductions or consumer gains may be
discounted because consumersdo not fully regard the new price
as much lower than the benchmark and adjust the gain to more
reasonable levels. By the same token, price increases or con-
sumer losses may also be discounted because consumers try to
rationalize buying products at a higher price. In contrast, both
price reductions and increases can be encoded as more sub-
stantial than what is objectively true (Berkowitz and Walton
1980). The extent to which these discounting or amplication
processes occur depends on the context in which consumers
evaluate price changes (Thaler 1985). Research has identied
the relative price positioning of both brands and retailers as
major factors (e.g., Gupta and Cooper 1992; Mazumdar, Raj,
and Sinha 2005).
Indeed, a substantial body of work has demonstrated that
premium-priced brands stand to gain more from price dis-
counts than economy-priced brands (e.g., Allenby and Rossi
1991; Bronnenberg and Wathieu 1996). In general, temporary
price discounts by premium players (brands and/or retailers)
that do not typically compete on price may be perceived as
more substantial and important, subsequently inuencing
consumerspurchase decisions (Wathieu, Muthukrishnan, and
Bronnenberg 2004). Although previous research has focused
considerably less on the effects of the subsequent price in-
creases following price discounts (for notable exceptions,
see Kahn and Louie 1990; Wathieu, Muthukrishnan, and
Bronnenberg 2004), we can apply a similar reasoning to price
increases, suggesting that price increases by premium players
may also appear more substantial, as the initial price cut
preceding the increase was more salient from the beginning
(Wathieu, Muthukrishnan, and Bronnenberg 2004). Yet
Campbell (1999) posits that brand reputation also comes into
play, such that brands that rely on a higher reputation suffer
less from the negative consequences of price increases. The
question remains, however, of how these effects play out in
contexts in which several price moves occur simultaneously
(as is the case with PWs) rather than sequentially (as is the case
in price promotion settings) (Kahn and Louie 1990). In
summary, the price promotion literature has (1) assessed the
impact of temporary price cuts extensively, but focused less on
the subsequent effects of price increases; (2) suggested that
price cuts are amplied or discounted depending on the po-
sitioning and reputation of the players; and (3) demonstrated
that premium-priced players win when prices are cut.
Translating these insights to our context of sustained price cuts
suggests that premium-priced brands stand to gain the most
during PWs but stand to lose more when PWs end.
Building on both the PW and price promotion literature,
we conclude that the effects of PWs and post-PWs are
conditional on the positioning of the brands included in the
PW. Yet whereas the PW literature has suggested that
economy-priced brands stand to gain most, the price pro-
motion literature has declared premium-priced players the
winners. In this article, we rst reect on the main effects of
the different PW settings and then elaborate on the mod-
erating impact of brand positioning and actions.
Impact of PW Scenarios on Brand Performance of Retailers
Involved in the PW
As explained previously, we distinguish four PW scenarios:
a PW_PDrop, a PW_PCst, a post-PW_PLift, and a post-
PW_PCst setting. In the PW_PDrop and post-PW_PCst
settings, a brands absolute or relative price decreases,
whereas in the PW_PCst and post-PW_PLift scenarios, the
absolute or relative price increases. Basic economic reasoning
predicts that absolute price changes result in demand changes.
This may occur even more so in the PW context, in which the
importance of price is amplied as price awareness and sen-
sitivity increase (Van Heerde, Gijsbrechts, and Pauwels 2008).
Moreover, even when a brands absolute price is not altered, the
brands relative price positioning may change because will-
ingness to pay often depends on theprices of rival products and
brands (Reibstein and Wittink 2005). As such, we propose that
when retailers reduce the regular prices of brands in a
PW_PDrop setting, demand increases. However, demand may
decrease when a brand is part of a PW_PCst setting because the
retailer does not lower the focal brands price while some of the
rival brandsprices are reduced. Although the absolute price
level does not change, the focal brands relative price de-
teriorates, which could lead consumers to switch from the focal
brand. Post-PW_PLift settings entail absolute price increases,
which should decrease demand for the brands involved in the
direct post-PW. Still, in a post-PW_PCst scenario, the regular
price is not (yet) restored, while rivalsprices are increased, and
thus a brands relative price positioning improves. This case can
result in brand switching in favor of the focal brand.
Moderating Role of Relative Brand Positioning
Relative brand price. Because price reductions for more
expensive brands are more salient and bring the focal brand
within reach of more budget-conscious consumers, enabling
them to enjoy benets they otherwise would not (e.g.,
Bronnenberg and Wathieu 1996; Chandon, Wansink, and Laurent
2000), the perceived gains in PW_PDrop and post-PW_PCst
settings may be amplied. Still, when economy-priced brands
are involved in PW_PDrop or post-PW_PCst scenarios, the
range of prices available to consumers may be extended
downward, making other brands appear relatively more ex-
pensive(Cunha and Shulman 2011), which suggests that price
reductions will be discounted more for relatively more ex-
pensive brands. When prices are increased, however, con-
sumers experience potential losses. These losses may also
appear more substantial for premium brands because they were
more salient from the beginning. Moreover, increasing the price
of premium-priced brands may extend the price range upward,
making rivals appear cheaper(Cunha and Shulman 2011).
Therefore, premium-priced brands may stand to lose more
when involved in post-PW_PLift and PW_PCst settings.
Brand ownership. Brand name is another important
contextual cue that affects the impact of PW scenarios
(Gupta and Cooper 1992). Because PLs typically rely less
on a loyal consumer base (Sethuraman and Raju 2012) and
PL prices are more difcult to compare across retailers
(Vanhuele and Dr`eze 2002), consumers may have poorer
price knowledge about PLs. Consequently, PL price moves
are less salient, and the effect of PW_PDrop and post-
PW_PCst scenarios may be noticed less. Still, because PLs
are mostly perceived as price ghters, their downward price
moves may turn all rivals relatively more expensive, thus
implying that PLs stand to gain from PW_PDrop and post-
PW_PCst settings. Price increases, however, may be less
salient and do not extend the range of available price options.
Therefore, PLs may be affected less negatively in post-
PW_PLift and PW_PCst scenarios than NBs.
Moderating Role of Retailer Positioning
The perceived gains and losses attributed to a brands
price changes may vary by store (Berkowitz and Walton
1980), depending on the overall relative price position of
the store (across a full basket of goods). Price reductions in
premium-priced stores may be more salient and believable
because consumersperceived value of price reductions
may be higher when encountered in a store that does not con-
sistently make price claims. Price reductions in discount-oriented
store environments may therefore go less noticed (Gupta
and Cooper 1992). Yet when price ghters engage in PW ac-
tivities, rival regularretailers may appear disproportionally
expensive, and their price cuts may be viewed as unsuccessful
attempts to mimic the price ghters. Thus, engaging in PW_PDrop
and post-PW_PCst settings may be less successful at premium-
priced retailers (Van Heerde, Gijsbrechts, and Pauwels 2008),
reducing the positive effects for brands sold there. In addition,
when premium-priced retailers increase prices, such increases
are further amplied because they extend the range of available
price options, and rivals may seem less expensive. The neg-
ative effects of post-PW_PLift and PW_PCst settings are thus
reinforced at premium-priced retailers.
Moderating Role of Marketing-Mix Actions
Feature and price promotions. These types of promotions
lower consumersexpected price levels (Alba et al. 1999;
Kalwani and Yim 1992). As such, the reduced regular prices in
PW_PDrop and post-PW_PCst settings may be less salient.
Moreover, the effectiveness of sustained low price claims in
PW_PDrop scenarios may be further weakened when products
are offered on promotion because consumers may be waiting
for even betterdeals (Mela, Gupta, and Lehmann 1997). Thus,
the perceived gains following PW_PDrop and post-PW_PCst
scenarios may be discounted when a brand is on promotion. In
contrast, when prices are increased, price promotions and
features feed a(n incorrect) lower perception of the regular
price, thus reducing potential losses linked to post-PW_PLift
and PW_PCst settings.
Brand advertising. In general, (nonprice) brand-oriented
advertising strengthens brand image, causes greater
awareness, differentiates products, and builds brand equity
(Keller 1993). Advertising signals product quality, leading
to an increase in brand equity (Kirmani and Wright 1989).
Because brand-oriented advertising may increase product
differentiation, it can make price elasticities less negative
(Mitra and Lynch 1995). Consequently, the perceived gains
of PW_PDrop and post-PW_PCst settings may be less
substantial, while the potential losses of post-PW_PLift and
PW_PCst settings may be reduced as well.
Natural-Experiment Setting
We investigate the nationwide supermarket PW that took
place in the Netherlands between 2003 and late 2005 (see
also Van Heerde, Gijsbrechts, and Pauwels 2008)4to assess
the impact of PWs on brand performance. On October 20,
2003, Albert Heijn, the market leader, announced price
reductions for more than 1,000 products, which immedi-
ately triggered the PW: the main competing retailers reacted
within the same week, matching or even exceeding Albert
Heijns price reductions. Over time, prices were lowered
in a series of waves, involving different product categories
and brands (NBs and PLs) in every new round. In total, 16
rounds of price reductions were announced in the media.
Most of these rounds were initiated by the market leader
and involved between 250 and 2,000 NBs and PLs each
time, with an average decrease of 11% of the NB price
levels. On average, 8 weeks elapsed between two rounds,
with a minimum of 1 week at the start of the PW and a
maximum of 20 weeks during the last two rounds. The PW
ended on October 31, 2005 (Van Heerde, Gijsbrechts, and
Pauwels 2008), when Albert Heijns competitors did not
follow its last round of price reductions.
We use Dutch GfK Benelux household panel data, which
feature advertising, volume share, and volume and value sales
series for all brands in each retailercategory combination, to
construct weekly regular and promotion retail prices. We ob-
tained weekly brand advertising spending information from
Nielsen. The data cover a span of 309 weeks, from January
2002 to January 2008. Thus, the analysis includes 93 weeks
before the start and 113 weeks after the last round of the PW.
As such, we can examine the market before, during, and after
the PW (Ailawadi, Lehmann, and Neslin 2001). We restrict
the analysis to ve leading national supermarket retailers:
Albert Heijn, C1000, Edah, Plus, and Super de Boer. The
retailers vary in price positioning, service level, and extent of
promotional activities, with Albert Heijn scoring high on both
price and service, Super de Boer and Plus scoring average on
price and assortment, C1000 having lower prices and good
service, and Edah having lower prices and lower service.5We
screened up to ve leading brands within each of 25 cate-
gories.6The 25 categories represent some of the most fre-
quently purchased categories in the Netherlands. However,
the ve brands were not consistently available on the retailer
shelves in all categoryretailer combinations in the observation
window. In total, we track 324 brandcategoryretailer
combinations (corresponding to 162 unique brands); we used
4As Van Heerde, Gijsbrechts, and Pauwels (2008) discuss, all criteria that
dene a PW (Heil and Helsen 2001) were met during the Dutch PW.
5For consumer packaged goods, retailers set prices centrally and advertise
these prices nationwide in the Netherlands, implying that no price-zoning
practices are used. Note also that in May 2006, Laurus, the holding company
behind Edah, announced that it sold Edah to the Sligro Group, after which the
Edah chain was dissolved (Planet Retail 2006). As such, we censor the Edah
observations from May 2006 onward, which leaves us with fewer time
observations in the post-PW settings for Edah.
6The categories we analyzed are beer, biscuits, candy bars, cheese, chips,
chocolatemilk, coffee, coffee milk, cream, detergents,drinkable yogurt, frozen
pizza, soft drinks, ice tea, juice, low-fat spreads, margarine, milk, muesli,
softener, biscuits, laundry detergent, laundry powder, water, and yogurt.
Insights from a Dutch Price War 5
97,036 observations in our estimations (i.e., 79 brands at
Albert Heijn × 309 weeks + 81 brandsat C1000 × 309 weeks +
44 brands at Edah × 239 weeks + 47 brands at Plus ×
309 weeks + 73 brands at Super de Boer × 309 weeks).
Performance variables. We capture brand performance
by volume shares, volume sales, and value sales. The rst
two metrics enable us to test our predictions, and the latter
provides the required input for our discussion on how PW
scenarios affect the brands revenue at the retailer level. We
capture each brands market share in a brandcategory
retailer combination as the ratio of the sales volume of all
stockkeeping units (SKUs) of that brand in a category c at a
specic retailer r to the total category sales volume at that
retailer. The pattern of shares indicates that shares level off
when brands are involved in a PW_PCst setting (1.8%
change relative to pre-PW setting) and gain momentum
when they enter PW_PDrop (share: +5.9%) and post-
PW_PCst (share: +6.4%) settings. Overall, increasing
prices in the post-PW_PLift scenario negatively affects
shares only marginally (1.5%). The effects of the different
PW stages on volume and value sales are less clear-cut.
Although the descriptives indicate the average impact of
PW stages on brand performance, they do not allow us to
separate the shifts related to PW involvement from those
induced by price, promotion, or other uctuations.
Assessing the true impact of the PW beyond these con-
founding effects calls for a formal model estimation.
PW scenarios. We distinguish four PW scenarios:
a PW_PDrop scenario, a PW_PCst scenario, a post-
PW_PLift scenario, and a post-PW_PCst scenario. We
compare these settings with the pre-PW setting, which we
use as a reference scenario. The scenarios are represented
by a set of mutually exclusive step dummy variables that
take the value of 1 when a brand is exposed to a particular
setting and 0 otherwise. Using this classication enables us
to gain explicit insights into potential asymmetries between
scenarios in effects of price increases versus price decreases
and relative versus absolute price changes. To that extent,
our operationalization is in line with prior work that ex-
amines asymmetries in promotion effects that differentiate
between losses and gains (e.g., Bronnenberg and Wathieu
1996; Sethuraman, Srinivasan, and Kim 1999). To
classify a brand over time in the different scenarios, we
track the regular retail price per volume of the brands
leading SKU (for a similar approach, see Bronnenberg,
Mela, and Boulding 2006).7Because the price scenarios
reect retailersdecisions to include brands in the PW to
enhance their overall performance, we do not treat these
PW variables as exogenous in the empirical model. We
instrument the PW dummy variables using a set of variables
that reect the cost of doing business at the brand, category,
and retailer levels. (We provide more details on this ap-
proach in the Methodologysection.) Next, we explain
how we assign brands to the different PW scenarios and
then elaborate on the operationalization of the different
brand and retailer contextual moderator variables. Web
Appendices A and B provide information on the frequency
of the individual PW scenarios and the different sequences
of PW scenarios.
PW scenarios denition. A brand enters a PW_PDrop
scenario when a particular retailer reduces its regular price
permanently in a given category. In our sample, 87 of the
brandcategoryretailer cases were immediately involved
in the PW and never experienced a PW_PCst setting. In
total, 249 cases entered a PW_PDrop setting at some point,
66% of which entered in the rst three months; only ap-
proximately 3% entered during the last wave of price re-
ductions in 2005. For the PW_PCst scenario, the retailer
does not reduce a brands regular price in a given category.
In total, 237 cases experienced a PW_PCst scenario. Brands
that experienced a PW_PCst scenario remained, on aver-
age, for 57 weeks in that scenario and only 29 weeks if they
also entered the PW_PDrop scenario. Overall, 73 cases
stayed in the PW_PCst scenario and thus never experi-
enced a price reduction.
The last round of price reductions occurred on October
31, 2005. From that moment on, prices of brands directly
involved in the PW slowly began to increase. We assign a
brand to the post-PW_PLift scenario as soon as its regular
price is increased. If the price of the brand is not increased,
but the retailer increases a competing brands regular price,
we set the focal brand as involved in the post-PW_PCst
scenario. Prices increased for only eight cases immediately
after October 31, 2005. On average, retailers waited 44
weeks before increasing regular prices.8
Brand positioning. We capture a brands relative posi-
tioning as the ratio of the brands regular price to the
category price across all brands in category c at retailer r at
time t. On average, the relative price positioning tends to
improve during the PW_PDrop scenario (3.4%) but
worsens during the PW_PCst (+4.7%) and post-PW_PLift
(+2.7%) settings. We further capture brand ownership by
means of a dummy variable that equals 1 if the brand is a PL
and 0 otherwise. In total, we have 81 PL and 243 NB cases.
Retailer positioning. We capture a retailers relative price
positioning using the average price at the retailer of a basket
containing all products carried by all ve retailers in the 25
tracked categories, relative to the average across all ve
retailers (for a similar approach, see Ailawadi, Pauwels, and
Steenkamp 2008; Gijsbrechts, Pauwels, and Van Heerde
Marketing-mix actions. To capture the impact of price
promotion activities, we include a price index, measured as
the weekly shelf price of the brands leading SKU divided
by its nonpromotional regular price (e.g., Van Heerde,
7The regular price of an SKU refers to the nonpromotional price within
a given brandcategoryretailer combination. Hereinafter, regular price
refers to the regular price of the leading SKU. To assess the robustness of our
classications, we checked up to ve leading (in terms of volume sales)
SKUs of each brand. This check enabled us to verify and conrm the exact
timing of price increases and reductions.
8Involvement in a PW can vary widely across retailers for one and the same
brand. For example, the retailers used Amstel, a beer brand, in four ways in
our data. At one retailer, Amstel was immediately included in the PW
(PW_PDrop), and its price was not raised again after the PW (post-
PW_PCst); another retailer did not include Amstel immediately in the
PW but focused on its rivals (so Amstel also went through a PW_PCst
setting). At yet another retailer, Amstel was immediately included in the PW,
but the retailer did restore the price after the PW (post-PW_PLift). Finally,
at the two remaining retailers,the brand went through all fourPW scenarios, but
even in those cases, the timing of entering the scenarios is not the same.
Leeang, and Wittink 2004). To capture feature promotions,
we include a dummy that equals 1 in the weeks when the
retailers featured the focal brand in their circulars and 0 oth-
erwise. Following Mela, Gupta, and Lehmann (1997), we use a
brands national television advertising spending to capture
brand-oriented advertising. Because all marketing decisions
are ultimately endogenous, we instrument all price-based,
feature, and advertising variables. We elaborate in detail on
this correction in the Methodologysection.
Control variables. We control for a deterministic trend
and retailer xed effects. In addition, we control for the
price evolution of the brand, captured by the brands regular
retail price in a given week divided by its regular retail price
before the PW, and detail direct cross-retailer effects
(Leeang et al. 2000; Wittink et al. 1998) by including the
focal brands price at competing retailers. To this end, we
use the average shelf price of the brand at competing re-
tailers to correct for competition across retailers at the brand
level. Moreover, we account for differences in the national
availability of a brand by including its weighted distribu-
tion. We capture this weighted distribution as the per-
centage of stores in which a product is sold, weighted by the
importance of the outlets, expressed in category volume
share. The shares used as weights pertain to the average
share in the previous year. Table 1 provides an overview of
the key descriptive for all the variables. The Web Appendix
presents the bivariate correlations between all variables
To test the effects of PW involvement and the moderating
effects of brand and retailer positioning on a brands sales
and market share, we use a multiplicative response model
(Neslin and Van Heerde 2009) and a multiplicative com-
petitive interaction market share attraction model (Cooper
and Nakanishi1988), respectively. Formally, we can ex-
press these as follows:
bc,t =a0
bc,t +eSales
c, and(1)
bc,t =Attrr
bc,t +eShare
lc,t +eShare
where Salesr
bc,t represents the volume sales of brand b in
category c at retailer r at time t and Sharer
bc,t is expressed as the
attraction of the brand, Attrr
bc,t, relative to the summed at-
tractions of the B brands offered in the category, thus en-
suring that market shares sum to unity and are between 0 and
1. We linearize the attraction model by applying the log-
centering transformation (see Cooper and Nakanishi 1988).
We specify a rst-order autoregressive residual structure to
account for autocorrelation in the error terms due to the time-
series nature of the data ður
bc,t =rur
Parsons, and Schultz 2001).9To accommodate the
hierarchical nature of the data, we correct for the grouping of
time observations within brands within categories. As such,
the error terms usharer
bc,t and usalesr
bc,t are normally distributed with
mean 0 and variances2. The random effects eshare
bðcÞand esales
are normally distributed over brands within retailer cate-
gories, with an expected value of 0 and variance equal to t
while eshare
cand esales
care normally distributed over categories,
with an expected value of 0 and a variance equal to t
Both sets of random components allow for a random co-
efcient specication across categories and stores on the
intercept, yielding a multilevel model (Raudenbush and Bryk
The term Xr
bc,k,t (k = 1, ...,K)representsasetofin-
dependent variables that enable us to tie the effects of the four
PW settings and the moderating brand and retailer effects to
brand sales and attraction. For easeof exposition, in Equation 3
we depict both brand attraction and sales as Perfr
bc,t to formally
connect both constructs with the covariates. We also use one
set of coefcients (p) for both relationships, though we esti-
mate the two equations separately and thus permit the esti-
mates of the coefcients to differ.
i,bc,t +
i,bc,t ×PL
p12+ iPWr
i,bc,t ×lnðRetRPr
p16+ iPWr
i,bc,t ×lnðPromor
p20+ iPWr
i,bc,t ×Ft
bc,t +
p24+ iPWr
ADVbc,t+p29 lnðBrRPr
bc +p31lnðRetRPr
bc,tÞ+p32 lnðPromor
bc,t +p34lnðADVbc,t Þ+p35lnðPr
bc,tÞ+p37 lnðDistbc,tÞ
+p38Trend + ebðcÞ+ec+u
bc,t +rur
where PWr
i,bc,t (i = 1, ..., 4) represents the focal constructs in
this analysis and expresses the four PW scenarios in which a
brand can be involved (using the before-PW setting as ref-
erence). The terms BrRPr
bc,t and PLr
bc capture the brands
relative price positioning and ownership, respectively,
whereas RetRPr
bc,t is the relative price positioning of the
retailer. In addition, Promor
bc,t and Ftr
bc,t reect the weekly
price and feature promotions for the focal brand, respectively,
and ADVbc,t captures brand advertising for brand b. The price
evolution since the start of the PW for the regular retail price
for brand b is represented as Pr
bc,t, and the brands price at
competing retailers is captured by CompRetPr
bc,t. We further
control for the weighted distribution for the brand (Dist
and a deterministic trend (Trend).
Retailers do not decide randomly which brands to in-
clude in a PW, nor do they set retail prices arbitrarily.
9To test the need for a rst-order autoregressive specication of the error
term, we regress the residuals of the sales and market share attraction models
on the independent variables used in those models and the lagged residuals.
The signicant p-values (<.01) suggest the presence of autocorrelation
(Franses and Paap 2001, p. 40).
Insights from a Dutch Price War 7
Rather, they select brands and prices with the potential to
maximize category and store performance. Similarly,
manufacturers may anticipate how consumers will react to
their marketing mix and adjust their promotional efforts
(advertising, price, and feature promotions) accordingly.
To account for possible endogeneity in these decisions, we
use an instrumental variables approach for the potentially
endogenous variables in the model.
Selection. Overall, suitable instruments must satisfy three
requirements: (1) they must be uncorrelated with the error
term of the performance model, (2) they must be suf-
ciently correlated with the endogenous elements of the
independent variables, and (3) they must be uncorrelated
with potentially omitted variables in the performance
equation (Wooldridge 2010, p. 112). Therefore, nding
suitable instruments is always a difcult challenge.
In general, prices and advertising will be set by the cost
and markup structure as determined by the nature of the
product, category, and retail environment. Costs are ex-
pected to be correlated with advertising and prices but to be
uncorrelated with demand and unobservable demand
shocks. Although cost drivers may be determinants of price
and advertising, Nevo (2001, p. 546) concludes that these
are rarely observed. To address this issue, we follow Nevos
approach of using marketing variables from a similar but
different market as instrumental variables. The logic is that
shocks in costs that cause exogenous variation in marketing
variables in one market will cause similar exogenous
variation in the focal market. For example, costs of in-
gredients may drive price variation in a market in the same
way that they drive price variation in a different and
noncompeting market. Thus, marketing variables in one
market can be used as instruments for marketing variables
in the other market (Hausman 1997; Nevo 2001; for recent
applications, see Dinner, Van Heerde, and Neslin 2014;
Lamey et al. 2012; Ma et al. 2011; Van Heerde et al. 2013).
For these instruments to be valid, no common demand
shocks may occur across markets, nor can advertising or
promotion activities be coordinated across markets. This is
more likely when a different set of retailers is active in the
different markets.
Giventhatweareworkingwithdatathatarespecied at
the national level (the Netherlands) and that cover a wide
variety of product categories, we select information for
the same set of categories from another national market
(i.e., the U.K. market), to ensure the absence of such
common demand shocks. In the United Kingdom, none of
the retailers involved in the Dutch PW are active, nor do
U.K. retailers operate in the Dutch market. Moreover, we
double-checked the business press to ensure that no PWs
were taking place at the same time in the United Kingdom.
We further validated this empirically by regressing U.K.
prices and advertising on the Dutch PW event, and the
effect was not signicant (p>.10). Overall, the U.K. retail
market is considered similartotheDutchmarketwith
respect to logistics (e.g., infrastructure, population den-
sity), market structure (e.g., dominant trading format, own-
label penetration, concentration), and extent of vertical
integration and supplier dependence (e.g., revenues derived
[%] from wholesalers and manufacturers) (for more detail,
see Fernie and Staines 2001), making it an excellent
candidate from which to derive instruments. Thus, for every
category in our sample, we identied the leading NBs and
PLs at three leading U.K. retailersnamely, Tesco,
Sainsbury, and Asda10and matched them to the corre-
sponding NBs and PLs in our sample. For each brand, we
collected same and competitor prices and advertising11 in
thesameperiod(20022008). In summary, the U.K. prices,
Table 1
Before PW
PW Price
PW Price
Constant Post-PW Price Lift
Post-PW Price
All 324 324 249 237 144 267
NBs 243 243 191 173 116 196
PLs 81 81 58 64 28 71
Market share (%) 21.9b(.19)c22.4 (.16) +5.9%d1.8% 1.5% +6.4%
Sales (volume in thousands) 47.7 (120.8) 42.6 (75.6) +17.6% +.7% +35.6% +50.8%
Revenues (thousands of euros) 5.6 (10.8) 5.1 (6.9) +6.3% .7% +32.3% +41.7%
Brand price positioning 1.00 (.34) 1.00 (.32) 3.4% +4.7% +2.7% .4%
Retailer price positioning .99 (.05) 1.00 (.05) .1% .0% +1.6% +.5%
Price promotion index .97 (.07) .97 (.07) +.5% +.1% +.8% +.5%
Feature (%) 9.11 (.28) 9.14 (.10) 3.4% 6.6% +1.0% 14.9%
Advertising (thousands of euros) 27.56 (87.00) 34.91 (59.89) +20% 7% 24% +23%
Regular price .23 (.20) .25 (.21) 14.4% .2% 4.8% 9.7%
Distribution .75 (.34) .74 (.34) +2% +1% 1% +2%
No. of observations 97,036 29,160 25,634 13,416 8,521 20,305
aCases refer to brandcategoryretailer combinations. In total, we follow the performances of 162 unique brands. Not all 162 brands are available across all
retailers (e.g., PLs). To avoid confusion, we report the total number of brandcategoryretailer combinations.
bMean value.
cStandard deviation.
dAll percentages reported in the table represent the percentage change versus the pre-PW (baseline) setting for the cases included in that scenario.
10These retailers are similar in terms of assortment, prices, service, and
quality to the retailers in our data set.
11The price data were obtained through Aimark and derive from the U.K.
Kantar household panel data. The advertising data are from Nielsen Media
averaged across the three retailers, and advertising series
for each NB and PL and its top four competitors (ten in-
struments in total) enable us to partially capture evolutions
in production and advertising costs.
To further capture the evolution in overall production
costs, we use the following additional instruments from
the Dutch market: quarterly gross domestic product per
capita, monthly fuel (energy) index (this index is based on
the prices of oil, petroleum, natural gas, and coal), and the
overall consumer price index (data available from Statis-
tics Netherlands [
htm]). Furthermore, to capture cost and differentiation
drivers related to systematic differences across categories,
we select additional instruments that reect the competitive
structure within every category not yet captured by the
previous sets of instruments (see Hausman 1997; Luan and
Sudhir 2010; Nevo 2001; Srinivasan, Nijs, and Pauwels
2008). The idea is that the competitive structure drives
prices and advertising (e.g., brands in categories with a
higher level of concentration may have lower markups). As
such, we include sets of variables that, on the one hand,
reect the concentration level, the level of price dispersion,
the extent of SKU proliferation, growth, and purchase
frequency12 and, on the other hand, capture a set of
perceptual variables that reect consumer category in-
volvement, perceived quality, the value difference between
NBs and PLs in the category, and the perceived promotion
intensity in the category.13 These variables are time in-
variant, measured in the U.K. market in 2001, the year
before the Dutch PW erupted. By using U.K.-based data,
we again lter out the potential impact on demand while
capturing the impact on the cost structure.
To capture cost drivers at the retailer level, we include
store service and assortment quality, which reect the re-
tailerslogistic and operational costs.14 Finally, to capture
cost and markup differences due to innate brand and retailer
characteristics, we use xed-effects dummies for each re-
tailer and brand. These dummy variables reect the location
of the brands and retailers in the characteristics space and
can be assumed as exogenous or at least predetermined.
Characteristics of other products are correlated with price
and advertising because the markup and differentiation of
every product depend on the distance from the nearest
neighbor. Because brand characteristics are exogenous,
they are valid instruments (Nevo 2001).
Finally, we use the fourth lag of the instruments. The
sufciently long lag ensured exogeneity, as conrmed by
the Sargan test, which we discuss subsequently (for a
similar practice, see Ataman, Van Heerde, and Mela 2010).
Strength and validity. We formally assessed the validity
for overidentifying restrictions (Wooldridge 2010, pp.
13435). The Sargan test rmly conrmed their validity:
we could not reject the null hypothesis that the residuals and
the instruments are uncorrelated at any of the conventional
signicance levels. Second, we also checked the strength of
the individual instruments and removed instruments for
which the signicance level exceeded .10. The p-values of
our instruments are below .01, suggesting that the variables
are sufciently strong. Moreover, we regressed each en-
dogenous variable rst against the exogenous variables in
the performance model and then added the instruments to
conduct an incremental F-test for the explanatory power of
these independent variables. Overall, the set of remaining
instruments was still sufciently strong, as evidenced by
the R-square and F statistics. Across the four scenarios, we
obtained an average (median) R-square of 45% (49%), and
all incremental F-values exceeded the common threshold
of 10 (on average, the incremental F-values are 1,236,
median = 893; see the Web Appendix). Thus, we conclude
that the instruments are not weak. Given that the in-
struments are valid and not weak, we used the
HausmanWu test to formally probe for endogeneity in
the performance equations by comparing the ordinary least
squares and the independent variable estimates (see
Wooldridge 2010, p. 120). We can reject the null hy-
pothesis that the estimates are equal, conrming that
endogeneity exists. The results for the auxiliary regressions
appear in the Web Appendix.
Following the instrumental variables methods for ran-
dom coefcient models as proposed by Heckman and
Vytlacil (1998) (see also Wooldridge 2010, p. 145), we use
the resulting instruments and all exogenous variables of the
performance model to estimate the four PW dummy var-
iables, the log of the relative price, the log of the basket
price, the log of the relative change in prices, the log of
competing retailersprices, the log of the price promotion
index, the log of advertising, and the feature support using
two-stage least squares with generated instruments.15 The
correct variancecovariance matrix for the second-stage
regression must take into account that the instrumented
regressors were predicted from a previous (rst-stage)
regression (Wooldridge 2010, p. 145). To obtain the ad-
justed standard errors, we compute the residuals from the
second-stage equation using the parameter estimates ob-
tained by substituting the instrumented variables (the
predicted values of the endogenous variables) for their
original values (see Verbeek 2004, p. 145).
We rst discuss the estimation results for the volume
sales and market share attraction model and discuss the
empirical support for our hypotheses. Then, we use the
parameters to simulate revenue gains and losses of PW
involvement at both the chain and national levels. Table 2
depicts the parameter estimates for both the sales and
market share attraction models.
12These variables derive from the U.K. Kantar household panel data.
13The perceptual category variables were made available by Aimark and
were previously used by Steenkamp, Van Heerde, and Geyskens (2010) and
Steenkamp and Geyskens (2014). For more information on the operation-
alization, see Steenkamp, Van Heerde, and Geyskens (2010, pp. 101516).
14Perceptions of store merchandise quality and service come from
members of GfKs national household panel. The data cover an eight-year
period, from January 2001 to December 2008, and include 12 (biannual)
waves of surveys, each conducted during one week (around weeks 16 and 40
every year). The store-level scores represent the average across all panel
members (for a similar approach, see Lourenço and Gijsbrechts 2013).
15We used the log-transformed variables as dependent variables in our
auxiliary regressions, in line with Wooldridges (2010, pp. 26768) warning
against forbidden regressions.
Insights from a Dutch Price War 9
Volume Sales and Share Parameters
Main effects of PW scenarios. Across both models, the main
effects of the different PW scenarios behave as expected. In
the PW_PDrop and post-PW_PCst scenarios, both volume
sales and shares increase signicantly, whereas in the post-
PW_PLift and the PW_PCst settings, they both decrease
signicantly. The effect of a post-PW_PCst on sales and
share is about two times as pronounced as the effect of the
PW_PDrop scenario (sales: p
= .28 vs. p
.11 vs. p
= .25), implying that the true effects of PW
involvement can be reaped only after rival prices increase:
when consumers can contrast the sustained price reductions
to the increased regular prices of rivals, the perceived gains
become substantially more appealing. Moreover, the
negative effect of a post-PW_PLift is clearly more pro-
nounced than the effect of a PW_PCst setting. Finally, the
absolute effect of a post-PW_PLift is by far the largest of all
PW settings (sales: p
=2.26; share: p
=.31), followed
by the impact of the PW_PCst (sales: p
=.74). These
results corroborate the extant literature in that price in-
creases and decreases trigger asymmetric responses, with
price increases being more harmful overall than price de-
creases being benecial. At the same time, the negative
damaging consequences of not being involved in the PW
outweigh, in absolute value, the positive effects gained
when a brand is involved (sales: |p
= .28; share:
= .11).
Brand positioning. During a PW_PDrop, premium brands
gain less than economy brands (sales: p
=1.20, p<.01).
This result is in line with the nding in the PW literature
(Van Heerde, Gijsbrechts, and Pauwels 2008) that PWs
further accentuate the price dimension and benettradi-
tional price ghters not just at the chain level but also at the
brand level. Still, this does notapplytopost-PW_PCst
sustained price decreases than economy brands, as would
be expected from the ndings of the price promotion lit-
erature (sales: p
= 1.39, p<.01; share: p
= .27, p<.01).
So, during a PW_PDrop, price ghters stand to win. In line
with this notion, we nd that economy brands lose more
when stalling PW involvement (sales: p
share: p
= .83, p<.01).However,assoonassomerivals
prices are restored, premium brands gain more in sales and
Table 2
During PW Post-PW
Price Drop Price Constant Price Lift Price Constant
sign Sales Share
sign Sales Share
sign Sales Share
sign Sales Share
PW Scenario p
+ .28*** .11*** p
−−.74*** .15** p
−−2.26*** .31*** p
+ .44** .25***
× Relative brand price p
+/−−1.20### .03 p
1.61*** .83*** p
−−1.29** 1.65*** p
+/1.39### .27###
×PL p
.31### p
+ .08 .91*** p
+ 1.86*** 1.18*** p
+/−−.80## .46###
× Relative retailer price p
+/−−.84 .74### p
−−1.33*** 2.06*** p
8.06*** 2.23*** p
+/−−2.81### .03
× Price promo index p
+ 12.67*** .95*** p
−−19.48*** 4.45*** p
−−29.13*** 28.14*** p
+ 6.12* 2.22***
× Feature promotion p
−−1.05** .69*** p
+ 2.74** 1.35** p
+ 12.58*** 3.07*** p
−−5.01*** .18
× Brand advertising p
.04* .05*** p
+ .01 .21*** p
+.21*** .11*** p
.05 .02
Control Variables
Rel. brand price p
2.78*** 1.20***
PL p
3.71*** 1.01***
Rel. retailer price p
4.79* 1.75
Price promo. index p
3.50* 6.67***
Feature promotions p
1.23** .24
Brand advertising p
.06** .18***
Price evolution p
2.02*** 2.12***
Price brand comp.
.12** .12***
Distribution p
6.97*** .35***
Trend p
.00*** .00***
Intercept .46
AR(1) r.41*** .38***
N 97,036 97,036
.42 .37
2LL 314,062 134,624
*p<.10 (one-sided).
**p<.05 (one-sided).
***p<.01 (one-sided).
#p<.10 (two-sided).
##p<.05 (two-sided).
###p<.01 (two-sided).
Notes: We do not display parameter estimates for the brand and retailer dummy variables owing to space limitations, but they are available on request.
Notes: AR(1) = rst-order autoregressive; 2LL = 2 log-likelihood.
share: the reduced price gap with rivals whose prices were
restored mainly favors premium players. During a post-
PW_PLift and PW_PCst, premium brands are harmed more
than other brands by the unfavorable absolute or relative
price changes (p
=1.29, p<.05; p
=1.65, p<.01). This
result is in line with our expectations, which were based on
the salience argument.
During PW_PDrop settings, PLs seem to gain more than
NBs (sales: p
=.04,p<.05; share: p
= .31, p<.01), again
suggesting that PWs tend to favor all traditional price
ghters. However, in post-PW_PCst settings, when rivals
prices increase while the PLs lower price is maintained,
PLs do not manage to gain as much from a contrast effect
between their own sustained lower price and rivalsin-
creased prices (sales: p
=.80, p<.05; share: p
p<.01). In contrast, PLslower salience and consumer
knowledge favor PLs when confronted with higher absolute
prices because they lose less than NBs in post-PW_PLs
(sales: p
= 1.86, p<.01; share: p
=1.18,p<.01), which
is in line with our expectations. Yet when PLs face relative
price increases in the PW_PCst setting, PLs stand to lose
more (share: p
=1.91). Indeed, their price-ghter image
does not allow the retailer to exclude them from PW
Retailer positioning. As we expected, not being involved
in the PW tends to harm brands more when they are sold at
premium-priced retailers (sales: p
=1.33, p<.01; share:
=2.06, p<.01). At the same time, a brand may benet
more from PW involvement in premium store environments
because of the positive impact on share, while sales volume
is not harmed signicantly (share: p
=.74,p<.01). As we
expected, after the PW, brands gain less from having their
prices kept constant by more premium retailers (sales:
=2.81, p<.01; share: p
=.03, p>.05). Moreover,
although they may still stand to lose share, when premium-
priced retailers do restore brand prices after the PW ends,
brands stand to lose less sales at those premium retailers
(sales: p
=8.06,p<.01). Indeed, consumers may more
easily forgiveprice increases at premium retailers and
justify their price increases (Campbell 1999).
Marketing-mix actions. We expected promotions to re-
duce the perceived gains of sustained price reductions and
the perceived losses of regular price increases. This is
conrmed by the positive, signicant moderating effect of
the price promotion index (sales: p
= 12.67, p<.01; p
6.12, p>.05; shares: p
= .95, p<.01; p
and the negative, signicant moderating effects of the
PW_PCst and post-PW_PLift settings (sales: p
p<.01; p
=29.13, p<.01; shares: p
=4.45, p<.01;
=28.14, p<.01).16
The moderating effects of features conrm that
promotion-related information reduces both the gains of
price reductions and the losses of price decreases. The
interactions with the PW_PDrop and the post-PW_PCst
settings are negative for both sales and shares (sales: p
1.05, p<.05; p
=5.01, p<.01; share: p
=.69, p<.01;
=.18, p>.10), whereas the interactions with the post-
PW_PLift and PW_PCst scenarios are positive (sales: p
2.74, p<.05; p
=12.58,p<.01; share: p
Finally, we expected brand advertising to reduce price
sensitivity. This is indeed the case during the PW for the
brandsshareaswend a negative moderating effect
(share: p
=.05, p<.01) when the price is decreased and a
positive moderating effect when the brand is not involved in
the PW (share: p
= .21, p<.01). Advertising may help
mitigate the negative effects of not being part of a PW while
direct competitors are. This is, however, not the case when
it comes to sales (sales: p
=.04,p>.05; sales: p
p>.05). Moreover, after the PW dies out, advertising tends
to reinforce price sensitivity, as we nd that the moderating
effects of the post-PW_Lift setting are negative and sig-
nicant (sales: p
=.21, p<.01; share: p
=.11, p<
.01). This result may be an indication that brand advertising
stimulates consumers to contrast the advertised brands with
their rivals. A PW_PDrop and post-PW_PCst reinforce the
relative price decrease in favor of the advertised brand,
while advertising helps further differentiate the brand from
its rivals.
Impact of PW Involvement on BrandsRevenue and Gross
Prot at the Retailer Level
Ultimately, brand manufacturers want to assess the ex-
tent to which PW involvement will affect their brands
bottom line. Therefore, we simulate the average weekly
revenue gains and losses of each PW scenario for all the
brands in the data set while distinguishing between dif-
ferent types of brands at both the chain level and the overall
country level. We also assess the cumulative sales gains or
losses in total after passing several sequential PW scenarios
and evaluate how brands fare in total at the national level
after the PW ends. Finally, we offer some insights into the
impact of wholesale price renegotiations on brandsgross
To evaluate the revenue gains and losses of PW in-
volvement while ltering out all other confounding effects,
we reran the model using value sales (expressed in thou-
sands of euros) as our dependent variable. Because value
sales capture the combined effect of both volume sales and
price changes, we can use the parameters from this model to
assess the extent to which PW involvement leaves more or
less money on the table than before the PW erupted.17
Impact of individual PW scenarios. For each brand in our
data set, we calculate, ceteris paribus, the weekly revenues
the brand would have obtained in each scenario. Therefore,
we use the parameters of the revenue model and set all
relevant variables to their means in each respective sce-
nario, while assuming no additional marketing-mix sup-
port. First, we examine the effect at the chain level for both
NBs and PLs and also extrapolate the effects to the country
level18 for NBs involved in the same scenario at different
chains. Moreover, we classify brands into three price tiers
based on the distribution of their average relative price in
16The promotion index is below 1 in the case of a promotion and equals 1
in business-as-usual settings. Therefore, a positive (negative) sign implies
that promotions lead to a more negative (positive) impact of a PW scenario.
17Compared with the volume sales model, the parameters of the revenue
model seem robust and are similar with respect to signicance and sign. The
full set of results can be obtained from the authors on request.
18By denition, PLs are only available in one chain; thus, chain-specic
results reect the national level.
Insights from a Dutch Price War 11
the pre-PW stage. Second, we compute the revenue gains or
losses of the different PW stages as the difference between
expected revenues in the pre-PW setting and the projected
revenues in each PW scenario. Table 3 reports weekly revenue
gains and losses, extrapolated from the panel to the total Dutch
market, based on conversion rates provided by GfK.
As Panel A in Table 3 shows, on average, the increase in
revenue following a PW_PDrop involvement is limited and
even slightly negative (at the chain level: V360; at the
national level: V1,350), implying that the resulting vol-
ume sales lift does not compensate for the price cut. Still, at
budget retailers (i.e., price 10% below market average),
theselossesturntoslightgains(+V580), whereas at pre-
mium retailers (i.e., price 10% above market average), the
same brand incurs a more pronounced loss (V630). More
substantial revenue gains, however, can be obtained during
the post-PW_PCst setting: if the regular price is maintained
at the lower level while rival prices increase, brands can
expect revenue increases relative to the pre-PW baseline of
V5,370 and V38,150 at the chain and national levels, re-
spectively. Although the price remains well below its pre-
PW level, the resulting increases in volume sales offset
those lower prices substantially. In contrast, in settings in
which the brands price increases (relatively and abso-
lutely), substantial losses can be expected. On average, a
brand stands to lose up to V3,570 (nationally: V14,680)
in PW_PCst scenarios and up to V34,370 (nationally:
V132,100) in post-PW_PLift settings. This suggests that
although not being involved in the PW leads to substantial
revenue losses, caution is warranted because after the PW,
brands have a difcult time restoring prices. These losses
are most pronounced at budget retailers (V68,020) and
least pronounced at more premium players (V10,890).
The revenue effects of different brand tiers. When dis-
tinguishing between gains and losses for premium versus
economy brands, we nd that in PW settings involving
price decreases, the effects are mixed (see Table 3, Panel
B). In PW_PDrop settings, economy NBs and PLs gain
more than standard and premium NBs; on average, pre-
mium and standard NBs incur losses in direct PWs (V790
and V570, respectively), whereas the other brands earn
small gains (economy NB: +V350; standard PL: +V800;
economy PL: +V260). Nevertheless, for premium and
standard NBs, the most substantial gains of PWs unfold
after the PW, when rival brandsprices are restored while
focal brandsreduced PW prices are maintained. Indeed,
we nd weekly gains of V16,190 and V8,830 for premium
and standard NBs, respectively, while the gains for
economy NBs and standard PLs are more or less the same as
during the PW (+V270 and +V630, respectively) and
economy PLs incur a small loss of V870.
In PW settings involving relative or absolute price in-
creases, NBs suffer disproportionally more than PLs. In
PW_PCst, premium, standard, and economy NBs stand to
lose V3,180, V6,210, and V3,110, respectively, while
standard and economy PLslosses are limited to V260 and
V140, respectively. Thus, delaying PW involvement can be
costly for NBs and especially the premium-positioned
players. Yet the losses that can be incurred when the PW
ends and prices are restored are also most substantial for
Table 3
A: Chain and National Levels
PW Price Drop PW Price Constant Post-PW Price Lift Post-PW Price Constant
of V
of V
of V
of V
At Chain Level .36 8.2% 3.57 37.4% 34.37 88.0% +5.37 +45.1%
Premium chain .63 17% 5.80 45.4% 10.89 76.0% +1.1 +38%
Budget chain +.58 +3.2% 2.35 27.2% 68.02 95.4% +11.7 +60%
Total gain/loss across all
supermarkets (NB only)
1.35 9% 14.68 44.5% 132.1 90% +38.14 +42%
B: Impact of Brand Positioning
PW Price Drop PW Price Constant Post-PW Price Lift Post-PW Price Constant
of V
of V
of V
of V
Premium .79 8.5% 3.18 31.5% 37.42 95.7% +16.19 +48.3%
Standard .57 9.1% 6.21 43.2% 47.81 90.3% +8.83 +48.9%
Economy +.35 +9.4% 3.11 54.1% 12.62 67.0% +.27 +7.1%
Standard +.80 +.25% .26 28.5% +.24 +18.2% +.63 +25.0%
Economy +.26 +19% .14 15.2% +.05 +12.2% .87 50%
Notes: All gures are expressed in thousands of euros and are extrapolated from the panel to the total Dutch market on the basis of conversion rates provided by
GfK. The table presents the weekly gains or losses per PW setting using the results from a model with value sales (expressed in K euro) as the dependent variable.
The table reads as follows: A brand that enters the PW_PDrop tends to lose on average V.36K (V360) per retail chain (V.63K [V630] at premium chains, but
gain +V.58K [V580] at economy chains). Summing all revenue gains and losses across all supermarkets, a brand that enters the PW_PDrop tends to suffer 9%
revenue losses. For the purpose of this exercise, we use abstraction from all potentially correcting marketing efforts, such as advertising and price promotions.
Thus, the numbers may appear larger than observed in real life.
NBs (premium NB: V37,420; standard NB: V47,810;
economy NB: V12,620). Overall, PLs fare better and can
even gain (standard PL: +V240; economy PL: +V50).
Total revenue effect of PW scenario sequences. In addition,
we assess the cumulative revenue gains or losses a brand
stands to incur in total, after passing sequential PW settings,
relative to a business-as-usual setting that assumes that the
PW had not erupted. To do so, we multiply the weekly
revenues obtained during a particular PW scenario by the
number of weeks the brands remain in that setting (see the
Web Appendix). Moreover, we distinguish between sce-
narios that do not involve additional marketing-mix
spending and scenarios that allow for price and feature
price promotions after the PW when the brandspriceis
restored, because these levers are particularly helpful to
brands in these settings. For the former, we assume that a
15% price discount or feature is used every six weeks
during the direct post-PW setting on the basis of the average
interpromotion interval in our data set. Table 4 depicts the
total revenue shifts for the most frequent PW sequences
(see also the Web Appendix).
Every sequence of scenarios that involves a post-
PW_PLift phase will result in the lowest level of reve-
nues and even substantial losses (Sequences AD). The
losses are most substantial when a PW involvement is
delayed and the brand is the rstinthecategorytorestore
the price level (Sequence A: V5.47 million). When the
brand is immediately involved, the loss drops considerably
(Sequence B: V2.556 million). If the retailer stalls re-
storing the brand price after the PW ends, the average loss
drops further, and more so if the brand was not immediately
included in the PW (Sequence C: V1.711 million; Se-
quence D: V158,000). In contrast, if the brandspriceis
not increased when the PW ends, considerable revenue
gains can still be obtained, rising up to V1.185 million
(Sequence E) and V1.308million(SequenceF),withand
without incurring a PW_PCst setting, respectively. In
summary, signicant revenue gains can be achieved from
PWs, but they come at the cost of maintaining the price at
the reduced PW level.
Table 4 also reveals that using either increased price
promotion or feature intensity, on average, does not always
sufce to offset the negative effects of a post-PW_PLift
involvement. Only when involvement in PWs is stalled can
additional price promotions turn PW involvement into a
gain, and more so when the price increase after the PW is
also delayed (Sequence C: +V1.542 million; Sequence D: +
V3.104 million).
Total effect of retailer PWs on brands. We also assess the
extent to which retailer rivalry damages or benets brands
overall at the national level, regardless of differences in PW
involvement across retailers. Therefore, we evaluate the
total gains or losses a brand incurs relative to the start of the
PW, calculated as explained previously, across all retail
chains in which the brand is available.19 Across the board,
50% and 51% of all NBs and PLs included in this study
reported a gain, after we ltered out the potential gains of
additional marketing support. Taking this analysis one step
further, we explore what types of brands were able to
weather the downside of retailer-initiated PWs. To do so,
we ran two additional regressions, one for NBs and one for
PLs, in which we regressed the estimated total losses or
gains without additional marketing-mix support to a set of
covariates all measured before the start of the PW and
captured the brands (1) relative advertising spending, (2)
feature intensity, and (3) price promotion intensity and the
categorys (4) concentration and (5) SKU proliferation.20
For the NBs, we found that gains were, on average, higher if
NBs had invested more in advertising before the start of the
PW and less in feature communications and promotions.
So, whereas advertising does not help mitigate the negative
effects of price increases after the PW ends (while feature
communication and price promotions do), investing in
advertising before a PW helps shield brands from revenue
erosions from PWs. In contrast, PLs were better off when
they invested less in advertising before the PW. So, ad-
vertising may weaken the price-ghter image of PLs, which
hurts them during PWs. Still, PLs perform better during
PWs if their promotion intensity was lower before the PWs:
Table 4
Sequence Gains/Losses (in Thousands of V)
Sequence D
PW Price
PW Price
Price Lift
Post-PW Price
Constant No Support
Price Promo in
Post-PW Price Lift
Feature Promo
in Post-PW Price Lift
AUUU 5.470 1,657 2,169
BUU 2,556 876 1,315
CUUU U 1,711 +1,542 651
DUUU158 +3,104 +262
EUU U +1,185
No PW participation: +68
Notes: All gures are expressed in thousands of euros and are extrapolated from the panel to the total Dutch market on the basis of conversion rates provided by
GfK. The table presents the total gains or losses for the most frequent PW sequences that brands experienced in our setting, using the results from a model with
value sales as the dependent variable and the average number of weeks brands stay in each setting displayed in the Web Appendix. The table reads as follows: In
sequence A, a brand that enters the PW_PDrop and the post-PW_PLift setting tends to lose, on average, V5,470K (V5.47 million). These losses are reduced to
V1,657 (V1.657 million) or V2,169 (V2.169 million) if the brand uses 15% discounts or is featured every six weeks during the post-PW_PLift scenario.
19For the PLs, the national and chain level collapse because PLs are, by
denition, only available at one chain.
20Both regressions were signicant, with R-squares of .19 and .43 for NBs
and PLs, respectively. The results are available on request.
Insights from a Dutch Price War 13
price promotions may make potential consumer gains of
permanent price cuts less salient. We further highlight our
ndings by considering the 2 × 2 matrix presented in
Figure 2, in which we distinguish between brands that rate
low and those that rate high on advertising and promotion
intensity before the PW, on the basis of an approximate
median split. For each cell, we provide the median, the
25th, and the 75th percentile revenue loss or gain, to il-
lustrate the revenue values for the best- and worst-case
scenarios for both NBs and PLs.
Margin and wholesale price implications. Because of the
price cuts during PWs, retailers may want to renegotiate
wholesale prices21 to maintain their gross protlevelsand
put pressure on manufacturers to reduce these wholesale
prices. Retailers often argue that manufacturers stand to
benet from PWs because price reductions stimulate
market performance, such that even when manufacturers
partially carry the burden of the price reduction, their net
result is positive. Prots are, however, extremely sensitive
even to the slightest declines in average prices because any
decrease goes straight to the bottom line (Garda and Marn
1993). Therefore, we raise the question of how manufacturer
gross prots will be affected by renegotiated wholesale prices.
Although we do not have full margin and cost information for
all brands and retailers, we have gross margin information for
all brands at one retailer (i.e., the difference between the retail
price and the buy-in price). We dene gross prot as the gross
margin multiplied by the number of units sold. This enables us
to assess the effects of wholesale price reductions on brands
gross prots in case retailers aim to shift the full burden of the
PW retail price reductions to the brand manufacturer. More
specically, we calculate the impact on brandsgross prots
if the retailer renegotiated the wholesale price in such a way
that the retailer maintains (1) the pre-PW margins with the
new reduced regular retail price or (2) the same gross prot
level as before the PW.
If the retailer wants to maintain the same gross margin as
before the PW, the wholesale price will need to decrease, on
average, by 13%. In approximately 72% of all cases, the
resulting wholesale price shift will lead to an average gross
prot loss of 27%. If a brand can still gain, the gross prots
increase by 39% on average. If the retailer renegotiates the
wholesale price to maintain the same pre-PW gross protlevel,
wholesale prices decrease by 24%. In this case, brandsgross
prots decrease by 32% in approximately 70% of all cases.
However, in the remaining 30% of cases, brands still gain,
and their gross prots increase by 55%.
Brand manufacturers are often at the mercy of retailers when
their brands become involved in supermarket-initiated PWs. As
such, they feel pressured to decrease the buy-in price during the
PW and not to pass on price increases after the PW is over
(Aitamer and Dubreil 2013), even though the implications of
these actions on brand performance are unknown. Moreover,
suppliers must cope with retailersefforts to promote their PLs
to maintain their margins. Such intense price competition
distorts consumersprice perceptions, potentially jeopardizing
premium brands in favor of economy brands and PLs. Using a
natural experiment of a Dutch PW, we investigate what happens
to brand performance (1) during a PW, both when the brand is
not involved in the PW (while the direct competitors are) and
when the brand is included directly in the PW and regular prices
are reduced, and (2) after a PW, both when the brandspriceis
not increased while some rival brandsprices are restored and
when the brands regular brand is restored to a higher level. We
allowed brand and retailer positioning and marketing-mix ac-
tions to moderate the effects of these PW scenarios. Combining
the effects of different PW scenarios and sequences, we pinpoint
when brands are more likely to be harmed by supermarket-
initiated PWs in the following subsections.
PWs are not truly revenue, sales, or share generators. In
contrast with retailershopes of increased volume sales, the
resulting volume sales and revenue sales at the individual
brand level are relatively modest in most cases, suggesting that
volume sales lifts hardly compensate for the price drops. The
true gains of PW involvement can be reaped only if the retailer
does not restore the brands price in the aftermath of the PW,
Figure 2
Price Promotion
High Low
Advertising Intensity
P25: –3,922,782
Median: 102,038
P75: 2,416,120
P25: –1,047,943
Median: 449,893
P75: 1,089,220
P25: –1,132,231
Median: –6,128
P75: 217,676
P25: –215,905
Median: 1,385
P75: 1,629,884
Price Promotion
High Low
Advertising Intensity
P25: –16,428
Median: – 332
P75: 71,943
P25: –3,957
Median: 2
P75: 321,535
P25: –23,003
Median: –70
P75 6,132
P25: –330
Median: 2,369
P75: 21 , 0 51
A: NBs: National Total Revenue Gains/Losses
B: PLs: National Average Revenue Gains/Losses
21Typically, wholesale prices are negotiated onceat most, twice
annually. Conversely, retailer prices uctuate all the time, as do volume sales.
Thus, both a brands gross margin and prot (which is determined by the
gross margin and the volume sales) also uctuate all the time. Our purpose is
to determine how retailers can renegotiate wholesale prices to maintain
relatively steady gross prot levels when facing strongly uctuating retail
prices and volume sales.
which comes at the cost and risk of permanently devaluating
brand value and equity in the long run.
No reduction in price during PWs by retailers also hurts
brands. Even when revenues are considered, the best-case
scenario for most brands is when retailers enter brands in PWs
immediately by reducing the retail prices. The share, sales, and
revenue losses brands stand to incur by not being involved in
PWs even outnumber the gains they can reap during the PW.
This recommendation for (third-party) brands contradicts the
usual guideline given to potential direct PW participants
(i.e., the retailers in grocery PWs), which is to stay out of PWs
as long as possible (Rao, Bergen, and Davis 2000). Brands,
unwillingly caught in PWs, may benet frombeing involved in
PWs as quickly as possible.
The true gains or losses of PW involvement are revealed
only after the PW ends. Only when rivalsprices are restored,
while the focal brands reduced retail price is maintained,
can substantial sales and share gains be realized. Still, this
comes at the cost of maintaining a lower retail price and
thus risking brand equity losses in the long run. However, if
retail prices are increased after the PW, losses are sub-
stantial (in volume, value sales, or shares).
Staying out of the PW altogether is not necessarily
destructive. When brand prices are not reduced at any point
in time during the PW while rivalsprices are reduced,
brands may not be harmed and may even record a small
revenue gain. Indeed, potential gains can be reaped in the
aftermath of the PW when rival prices are increased and the
nonparticipatingbrands become the relatively less ex-
pensive options. Dissuading retailers from using brands
through buy-in price manipulation or other negotiation
tactics is a strong negotiation stance that may pay off (albeit
at the risk of souring the relationship with the retailer).
Moreover, if the retailer still decides to engage the brand in
its PW tactics, not only do these hypothetical stay-out-of-the-
PWgains vanish, but the deteriorated relationship may also
harm the brand in the long run. Thus, negotiations to stall PW
involvement may be a high-risk game.
Not all brands are equal: PWs jeopardize NBs but boost
PLsperformance. Premium (national) brands gain less in
volume sales and share and even risk incurring revenue
losses when retailers actively engage them in their PW
battles. In contrast, PLs still win with respect to revenue.
This may indicate that retailers can strategically use PLs
and NBs in PWs to further improve the strategic position of
their own store brands relative to NBs. In addition, given
that PLs lose more when their price is not reduced in PWs,
retailers may have more than one reason to actively engage
and prioritize their PL portfolio in PWs. In the aftermath of
the PW, NBs seem at rst to be the clear winners with
respect to shares and volume and value sales when the
reduced prices are maintained by the retailer. Still, this may
turn out to be a Pyrrhic victory in the long run; although
immediate revenues rise, ultimately the NBsequity and
leverage in relation to PLs is reduced, especially because
the price gap between both (one of the most important
drivers of sustained NBspower over PLs) will likely
shrink. Because PLs can withstand the negative effects of a
price increase in the PW aftermath better than NBs, they can
afford to restore their prices, which is not the case for NBs.
Indeed, all NBs can expect to incur severe losses when
increasing prices again at the end of the PW if they do not
take corrective marketing-mix steps. This nding may
leverage. For most other brands, any sequence that involves
price increases as the PW ends is bound to result in severe
revenue losses. Compensating with higher unit prices may
be the only way to reduce revenue losses without extra
investments in price and feature promotions (which ulti-
mately also affect brand equity). Overall, NB manufac-
turers should try to negotiate their way out of PWs if they
know how not to jeopardize the relationship with the re-
tailer completely.
Not all retailers are equal: higher performance at premium
chains. During PWs, brands stand to lose more in both share
and sales when premium retailers do not include them in
their PW activities. At the same time, brands may gain more
in share in these more premium shopping environments
when their price is reduced in the PW. In addition, after the
PW, restoring prices in premium store environments is
substantially less harmful to brand sales. This suggests that
when brand manufacturers want to negotiate their position
with retailers during PWs, they should pick their ghts
because they may fare better at premium retailers.
The role of the marketing mix tends to change before and
after PWs. Advertising does not help mitigate the negative
effects of price increases in the aftermath of PWs, whereas
feature communication and price promotions do. Again, this
may be a short-term x because increased price promotion
tools only feed consumersprice discount cravings, which may
ultimately erode brand value in the long run. Still, advertising
support before a PW creates a buffer that helps prevent revenue
erosion resulting from PWs. In contrast, PLs weather the
downsides of PWs better when they have invested less in
advertising in non-PW times, because advertising may
weaken the PLsprice-ghter image. At the same time, like
NBs, PLs suffer less from PWs if their promotion intensity
was lower before the PWs.
Margin renegotiations during PWs will likely be difcult.
Retailers hoping to engage brands in nancing the price
reduction may expect solid opposition from manufacturers.
During PWs, wholesale price reductions are not feasible in
the vast majority of cases, unless the brandsgrossprotis
reduced as well. In contrast, if the brands price is reduced
permanently in the PW aftermath, margin renegotiation
options become more realistic because revenue gains are
more substantial, creating more leeway on both the man-
ufacturer and the retailer sides. Still, because most re-
nancing is often required during the nancially strenuous
PW stages, interests will likely not be aligned, and nego-
tiations will be cutthroat.
Several issues for further research remain. In this article,
we mainly consider the effects on market share and only
briey touch on prot implications. Still, margin re-
negotiations and, thus, prot shifts are crucial elements in
retailermanufacturer relationships during PWs. Although
we provide insights into the potential outcomes for two
brands in one category, this exercise is limited in scope and
does not project the full cost spectrum. This is, however, not
uncommon in the literature, as full retailer and supplier margin
data are rarely available. Although we have limited infor-
mation on wholesale prices, with a full set of (time-varying)
Insights from a Dutch Price War 15
margin and cost data, a more complete prot picture could
be drawn. Further research should examine these prot
implications. Moreover, we gleaned some rst insights into
PWstotal brand demand by extrapolating the brandsre-
tailer sales to the national level. However, to gain deeper
insights into total brand demand, an alternative approach
could be to model total sales as the dependent variable and
all retailersindividual actions as independent variables.
Furthermore, research should explore the generalizability of
our results for other products, industries, and geographic
markets. In this article, we observe the performance of the
consumer packaged goods categories. Although we can
argue that PWs are fought mainly with brands that offer the
potential to affect store trafc, different effects could arise
for minor brands belonging to more niche-like or ll-in
types of categories. In addition, this research is set in the
Netherlands, a market in which PLs are well accepted and
not necessarily perceived as inferior options. Extending this
research to countries and markets in which PL penetration is
less high would verify whether PLs are as likely to thrive in
PW settings.
Ailawadi, Kusum L., Donald R. Lehmann, and Scott A. Neslin
(2001), Market Response to a Major Policy Change in the
Marketing Mix: Learning from Procter & Gambles Value Pricing
Strategy,Journal of Marketing, 65 (January), 4461.
———, Koen Pauwels, and Jan-Benedict E.M. Steenkamp (2008),
Private-Label Use and Store Loyalty,Journal of Marketing,
72 (November), 1930.
Aitamer, Gildas and Magali Dubreil (2013), Casino Lowers Prices,
Adding Fuel to the French Price War Fire,Planet Retail Daily
News, (February 14), (accessed October 2, 2015), [available at].
Alba, Joseph W., Carl Mela, Terence Shimp, and Joel E. Urbany
(1999), The Effect of Discount Frequency and Depth on Consumer
Price Judgments,Journal of Consumer Research,26(2),99114.
Allenby, Greg M. and Peter E. Rossi (1991), Quality Perceptions
and Asymmetric Switching between Brands,Marketing Science,
10 (3), 185204.
Ataman, M. Berk, Harald van Heerde, and Carl F. Mela (2010), The
Long-Term Effects of Marketing Strategy on Brand Sales,
Journal of Marketing Research, 47 (October), 86682.
Berkowitz, Eric N. and John R. Walton (1980), Contextual In-
uences on Consumer Price Responses: An Experimental
Analysis,Journal of Marketing Research, 17 (August), 34950.
Brandenburger, Adam M. and Barry J. Nalebuff (1996), Coopeti-
tion. New York: Doubleday.
Bronnenberg, Bart J., Carl F. Mela, and William F. Boulding (2006),
The Periodicity of Pricing,Journal of Marketing Research,
43 (August), 47793.
——— and Luc Wathieu (1996), Asymmetric Promotion Effects
and Brand Positioning,Marketing Science, 15 (4), 37994.
Busse, Meghan (2002), Firm Financial Condition and Airline Price
Wars,RAND Journal of Economics, 33 (2), 298318.
Campbell, Margaret C. (1999), Perceptions of Price Unfairness:
Antecedents and Consequences,Journal of Marketing Research,
36 (May), 18799.
Chandon, Pierre, Brian Wansink, and Gilles Laurent (2000), A
Benet Congruency Framework of Sales Promotion Effective-
ness,Journal of Marketing, 64 (October), 6581.
Cooper, Lee G. and Masao Nakanishi (1988), Market-Share
Analysis. Boston: Kluwer Academic Publishers.
Cunha, Marcus and Jeffrey D. Shulman (2011), Assimilation and
Contrast in Price Evaluations,Journal of Consumer Research,
37 (February), 82235.
Dinner, Isaac M., Harald van Heerde, and Scott Neslin (2014),
Driving Online and Ofine Sales: The Cross-Channel Effects of
Traditional, Online Display, and Paid Search Advertising,
Journal of Marketing Research, 51 (5), 52745.
Elzinga, Kenneth G. and David E. Mills (1999), Price Wars
Triggered by Entry,International Journal of Industrial Orga-
nization, 17 (2), 17998.
Fernie, John and Harry Staines (2001), Towards an Understanding
of European Grocery Supply Chains,Journal of Retailing and
Consumer Services, 8 (1), 2936.
Franses, Philip Hans and Richard Paap (2001), Quantitative Models in
Marketing Research. Cambridge, UK: Cambridge University Press.
Garda, Robert A. and Michael V. Marn (1993), Price Wars,
McKinsey Quarterly, 3 (August), 87100.
Gijsbrechts, Els, Koen Pauwels, and Harald J. van Heerde
(2014), Conict Journalism: Oil on the Fire? How Media
Coverage of a Price War Impacts Retailers, Consumers, and
Investors,working paper, Department of Marketing, Tilburg
Green, Edward J. and Robert H. Porter (1984), Noncooperative
Collusion Under Imperfect Price Information,Econometrica,
52 (1), 87100.
Grierson, Jamie (2011), Tesco Set to Trigger New Price War,The
Independent, (September 22), (accessed October 2, 2015), [available
Grifth, David E. and Roland T. Rust (1997), The Price of
Competitiveness in Competitive Pricing,Journal of the Acad-
emy of Marketing Science, 25 (2), 10916.
Gupta, Sunil and Lee G. Cooper (1992), The Discounting of
Discounts and Promotion Thresholds,Journal of Consumer
Research, 19 (December), 40111.
Hanssens, Dominique M., Leonard J. Parsons, and Randall L. Schultz
(2001), Market Response Models: Econometric and Time Series
Analysis, 2nd ed. Norwell, MA: Kluwer Academic Publishers.
Hausman, Jerry A. (1997), Valuation of New Goods Under Perfect
and Imperfect Competition,in The Economics of New Goods,
T.F. Bresnahan and R.J. Gordon, eds. Chicago: University of
Chicago Press, 20937.
Heckman, James J. and Edward J. Vytlacil (1998), Instrumental
Variables Methods for the Correlated Random Coefcient Model:
Estimating the Average Rate of Return to Schooling When the
Return Is Correlated with Schooling,Journal of Human Re-
sources, 33 (4), 9741002.
Hegarty, Ronan (2011), Waitrose Demands a 5% Cut from Sup-
pliers,The Grocer, (October 10), [available at http://www.
Heil, Oliver P. and Kristiaan Helsen (2001), Toward an Un-
derstanding of Price Wars: Their Nature and How They Erupt,
International Journal of Research in Marketing, 18 (1), 8398.
Huber, Joel, John Payne, and Christopher Puto (1982), Adding
Asymmetrically Dominated Alternatives: Violations of Regu-
larity and the Similarity Hypothesis,Journal of Consumer
Research, 9 (1), 9098.
Kahn, Barbara E. and Therese Louie (1990), The Effects of Re-
traction of Price Promotions on Brand Choice Behavior for
Variety-Seeking and Last-Purchase Loyal Customers,Journal of
Marketing Research, 27 (August), 27989.
Kalwani, Manohar U. and Chi Kin Yim (1992), Consumer Price
and Promotion Expectations: An Experimental Study,Journal of
Marketing Research, 29 (February), 90100.
Keller, Kevin Lane (1993), Conceptualizing, Measuring, and
Managing Customer-Based Brand Equity,Journal of Marketing,
57 (January), 122.
Kirmani, Amna and Peter Wright (1989), Money Talks: Perceived
Advertising Expense and Expected Product Quality,Journal of
Consumer Research, 16 (3), 34453.
Lamey, Lien, Barbara Deleersnyder, Jan-Benedict E.M. Steenkamp,
and Marnik G. Dekimpe (2012), The Effect of Business-Cycle
Fluctuations on Private-Label Share: What Has Marketing Conduct
GottoDowithIt?Journal of Marketing, 76 (January), 119.
Leeang, Peter S.H. and Dick R. Wittink (1996), Competitive
Reaction Versus Consumer Response: Do Managers Overreact?
International Journal of Research in Marketing, 13 (2), 10319.
———, Michel Wedel, and Philippe A. Naert (2000), Building
Models for Marketing Decisions. Boston: Kluwer Academic
Lourenço, Carlos and Els Gijsbrechts (2013), The Impact of Na-
tional Brand Introductions on Hard-Discounter Image and Share-
of-Wallet,International Journal of Research in Marketing,
30 (4), 36882.
Luan, Jackie and K. Sudhir (2010), Forecasting Marketing-Mix
Responsiveness for New Products,Journal of Marketing Re-
search, 47 (June), 44457.
Ma, Yu, Kusum L. Ailawadi, Dinesh K. Gauri, and Dhruv Grewal
(2011), An Empirical Investigation of the Impact of Gasoline
Prices on Grocery Shopping Behavior,Journal of Marketing,
75 (March), 1835.
Maxwell, Miranda (2011), Analysis: Food Multinationals Meet
Their Match in Australia,Reuters, (December 14), (accessed
October 2, 2015), [available at
Mazumdar, Tridib, S.P. Raj, and Indrajit Sinha (2005), Reference
Price Research: Review and Propositions,Journal of Marketing,
69 (October), 84102.
Mela, Carl F., Sunil Gupta, and Donald R. Lehmann (1997), The
Long-Term Impact of Promotion and Advertising on Consumer
Brand Choice,Journal of Marketing Research, 34 (May), 24861.
Milgrom, Paul and John Roberts (1982), Predation, Reputation,
and Entry Deterrence,Journal of Economic Theory, 27 (2),
Mitra, Anusree and John G. Lynch Jr. (1995), Toward a Recon-
ciliation of Market Power and Information Theories of Adver-
tising Effects on Price Elasticity,Journal of Consumer Research,
21 (March), 64459.
Monroe, Kent B. (1973), BuyersSubjective Perceptions of
Price,Journal of Marketing Research,10(February),
Neslin, Scott A. and Harald J. van Heerde (2009), Promotion
Dynamics,Foundations and Trends in Marketing, 3 (4),
Nevo, Aviv (2001), Measuring Market Power in the Ready-To-Eat
Cereal Industry,Econometrica, 69 (March), 30742.
Pauwels, Koen H., Shuba Srinivasan, and Philip Hans Franses
(2007), When Do Price Thresholds Matter in Retail Categories?
Marketing Science, 26 (1), 83100.
Planet Retail (2006), Laurus sells Edah stores to Sligro/Sperwer
Consortium,(May 29), (accessed October 13, 2015), [available
Rao, Akshay R., Mark E. Bergen, and Scott Davis (2000), How
to Fight a Price War,Harvard Business Review,78(2),
Raudenbush, Stephen W. and Anthony S. Bryk (2002), Hierarchical
Linear Models: Applications and Data Analysis Methods.
Newbury Park, CA: Sage Publications.
Reibstein, David J. and Dick R. Wittink (2005), Competitive
Responsiveness,Marketing Science, 24 (1), 811.
Rotemberg, Julio J. and Garth Saloner (1986), A Supergame-
Theoretic Model of Price Wars During Booms,American
Economic Review, 76 (3), 390407.
Sethuraman, Raj and Jagmohan Raju, (2012), Private Label
Strategies: Myths and Realities,Handbook of Marketing
Strategy in V. Shankar and G.S. Carpenter, eds. Glasgow, UK:
Edward Elgar Publishing, 31835.
———, V. Srinivasan, and Doyle Kim (1999), Asymmetric and
Neighborhood Cross-Price Effects: Some Empirical General-
izations,Marketing Science, 18 (1), 2341.
Srinivasan, Shuba, Vincent Nijs, and Koen H. Pauwels (2008),
Demand-Based Pricing Versus Past-Price Dependence: A
Cost-Benet Analysis,JournalofMarketing, 72 (March),
Steenkamp, Jan-Benedict E.M. and Inge Geyskens (2014), Man-
ufacturer and Retailer Strategies to Impact Store Brand Share:
Global Integration, Local Adaptation, and Worldwide Learning,
Marketing Science, 33 (1), 626.
———, Harald J. van Heerde, and Inge Geyskens (2010), What
Makes Consumers Willing to Pay a Price Premium for National
Brands over Private Labels?Journal of Marketing Research,
47 (December), 101124.
Thaler, Richard H. (1985), Mental Accounting and Consumer
Choice,Marketing Science, 4 (3), 199214.
Tuttle, Brad (2010), Soda Wars: $3.99 for a Case of Coke,Time
Magazine, (June 18), (accessed October 2, 2015), [available at
Van Aalst, Marcel, Laurens Sloot, Leo Van der Blom, and
Leo Kivits (2005), Het Grote VoordeelVan ´en
Jaar Prijsoorlog,Erasmus Food Management Instituut
Van Heerde, Harald J., Els Gijsbrechts, and Koen H. Pauwels
(2008), Winners and Losers in a Major Price War,Journal of
Marketing Research, 45 (September), 499518.
———, Maarten Gijsenberg, Marnik G. Dekimpe, and
Jan-Benedict E.M. Steenkamp (2013), Advertising and
Price Effectiveness over the Business Cycle,Journal of
Marketing Research, 50 (April), 17793.
———, Peter S.H. Leeang, and Dick R. Wittink (2004),
Decomposing the Sales Promotion Bump with Store Data,
Marketing Science, 23 (3), 31734.
Vanhuele, Marc and Xavier Dr`eze (2002), Measuring the Price
Knowledge Shoppers Bring to the Store,Journal of Marketing,
66 (October), 7285.
Verbeek, Marno (2004), A Guide to Modern Econometrics. Chi-
chester, UK: John Wiley & Sons.
Wathieu, Luc, A.V. Muthukrishnan, and Bart J. Bronnenberg
(2004), The Asymmetric Effect of Discount Retraction on
Subsequent Choice,Journal of Consumer Research,31(3),
Winer, Russell S. (1986), A Reference Price Model of Demand for
Frequently Purchased Products,Journal of Consumer Research,
13 (2), 25056.
Wittink, Dick R., Michael J. Addona, William J. Hawkes, and John
C. Porter (1988), SCAN*PRO: The Estimation, Validation and
Use of Promotional Effects Based on Scanner Data,working
paper, Cornell University.
Wooldridge, Jeffrey M. (2010), Econometric Analysis of Cross
Section and Panel Data, 2nd ed. Cambridge: Massachusetts
Institute of Technology Press.
Insights from a Dutch Price War 17
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
The growing sales of private labels (PLs) pose significant challenges for national brands (NBs) around the world. A major question is whether consumers continue to be willing to pay a price premium for NBs over PLs. Using consumer survey data from 22,623 respondents from 23 countries in Asia, Europe, and the Americas across, on average, 63 consumer packaged goods categories per country, this article studies how marketing and manufacturing factors affect the price premium a consumer is willing to pay for an NB over a PL. These effects are mediated by consumer perceptions of the quality of NBs in relation to PLs. Although the results do not bode well for NBs in the sense that willingness to pay decreases as PLs mature, the authors offer several managerial recommendations to counter this trend. In countries in which PLs are more mature, the route to success is to go back to manufacturing basics. In PL development countries, there is a stronger role for marketing to enhance the willingness to pay for NBs.
Full-text available
The authors investigate whether, and to what extent, marketing conduct varies over the business cycle and how this contributes to the growing popularity of private labels. To address this issue, they examine a unique data set that combines a broad set of seven marketing-mix instruments with private-label share, using two decades' worth of data for 106 consumer packaged goods categories in the United States. The results show that private-label share behaves countercyclically and that part of the boost in private-label share during contractions is permanent. Retailers' observed practice of supporting their own labels in contraction periods while cutting back in expansion periods helps this cyclical sensitivity even further. In addition, national brands' procyclical behavior in terms of (1) major new product introductions, (2) advertising, and (3) their promotional pressure compared with private labels is associated with more pronounced cyclical fluctuations in private-label share and even with permanent private-label market share gains. Although brand managers cannot be held responsible for the occurrence of economic downswings, they can be held accountable for how much contractions help strengthen their fiercest competitor, the store brands owned by their very customers.
Full-text available
A lthough store brands (SBs) are becoming increasingly important across the world, their success varies dramatically across consumer packaged goods categories and countries. The purpose of this paper is to provide insight into how such differences in SB success originate. Using a unique data set that combines scanner data for a three-to five-year period with consumer survey data (n = 20987) for scores of food, household care, and personal care categories from 23 countries around the world, we identify cross-national regularities as to the role of nine manufacturer and retailer factors in explaining SB market share. For each manufacturer and retailer factor, we determine whether it can be part of a global integration strategy, whether it can be included in a local adaptation strategy, or whether it is a candidate for worldwide learning. Our findings have important implications for national brand manufacturers and retailers. Preyas Desai served as the editor-in-chief and Dominique Hanssens served as associate editor for this article.
Full-text available
The current marketing environment is characterized by a surge in multichannel shopping and increasing choice of advertising channels. This situation requires firms to understand how advertising in one channel (e.g., online) influences sales in another channel (e.g., offline). This article studies the presence, magnitude, and carryover of these cross-channel effects for online advertising (display and search) and traditional media. The analysis considers how these advertising expenditures translate directly into sales, as well as indirectly through intermediate search advertising metrics-namely, impressions and clickthrough rate. For a high-end clothing and apparel retailer, the authors find that cross effects exist and are important and that cross-effect elasticities are almost as high as own-effect elasticities. Online display and, in particular, search advertising is more effective than traditional advertising. This result is primarily due to strong cross effects on the offline channel. Return-on-investment calculations suggest that by ignoring these cross effects, firms substantially miscalculate the effectiveness of online advertising. Notably, the authors find that traditional advertising decreases paid search click-through rates, thus reducing the net cross effect of traditional advertising.
Presentation of price is a common form of retail advertising. The authors examine the influences of semantic, comparison, and store name cues on consumer price perceptions. The research, unlike earlier student-based pricing studies, involved 562 consumers from a major metropolitan area who were each exposed to one of 16 treatment conditions. Results indicate comparison cues produce positive results, whereas the effect of semantic cues depends on the particular stimulus. Findings underscore the necessity for broader price research beyond traditional quality implications and wider representation of product types.