Advertising and Perceptual Biases* Andreas Grunewald, University of Bonn** and MatthiasKr¨akel, University of Bonn*** Abstract Consumers are often prone to perceptual biases. These biases (e.g., the opti- mism bias or conservatism) can make advertising less effective, which dampens firms’ incentives to invest in advertising. If advertising is persuasive and there- fore socially wasteful, biases that make advertising less effective are, at first sight, welfare enhancing. However, biases also change product market compe- tition, which can induce more advertising. We show under which conditions consumers’ perceptual biases lead to a net increase in persuasive advertising and, hence, to a net welfare loss. Our results show that this outcome will be particularly likely if consumers rate competitive goods as rather different prior to the firms’ advertising decisions. Keywords: combative advertising; persuasive advertising; consumer biases; price competi- tion. JEL classification: D03; D11; L1; M37. * Financial support by the DFG, grant SFB/TR 15, is gratefully acknowledged. ** Andreas Grunewald, Department of Economics, University of Bonn, Ade- nauerallee 24–42, 53113 Bonn, Germany, phone: +49-228-739217, fax: +49-228- 739210, e-mail: [email protected]. *** Matthias Kr¨akel, Department of Economics, University of Bonn, Adenauer- allee 24–42, 53113 Bonn, Germany, phone: +49-228-739211, fax: +49-228-739210, e-mail: [email protected]. 1
31
Embed
C:/Users/Grunewald/Dropbox/Working Papers/Advertisment and ... and... · Andreas Grunewald, University of Bonn** and Matthias Kr¨akel, University of Bonn*** Abstract Consumers are
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Advertising and Perceptual Biases*
Andreas Grunewald, University of Bonn**and
Matthias Krakel, University of Bonn***
Abstract
Consumers are often prone to perceptual biases. These biases (e.g., the opti-mism bias or conservatism) can make advertising less effective, which dampensfirms’ incentives to invest in advertising. If advertising is persuasive and there-fore socially wasteful, biases that make advertising less effective are, at firstsight, welfare enhancing. However, biases also change product market compe-tition, which can induce more advertising. We show under which conditionsconsumers’ perceptual biases lead to a net increase in persuasive advertisingand, hence, to a net welfare loss. Our results show that this outcome will beparticularly likely if consumers rate competitive goods as rather different priorto the firms’ advertising decisions.
Q18720131127.3See for example http://time.com/103175/heathrow-airport-samsung-galaxy/4Moreover, conservatism has been extensively discussed as possible explanation for finan-
cial anomalies, see, for example, Barberis et al. (1998), Hirshleifer (2001), Brav and Heaton
(2002), Barberis and Thaler (2003), Brandt et al. (2004).5For example, consumer conservatism can induce overly loyal behavior of consumers as
documented by Prince (2011) and Hortacsu et al. (2015). In consequence, consumer choice
may hardly be affected by advertising.
2
firms’ application of persuasive advertising and analyze the resulting conse-
quences for overall welfare. We focus on persuasive advertising for two rea-
sons.6 First, as persuasive advertising aims at manipulating consumer pref-
erences by influencing their quality perceptions, it is most apparent that the
belief formation process is crucial for its impact. Second, persuasive advertis-
ing is the most problematic form of advertising as it is pure social waste.7
In our model, we consider two firms each supplying one good. The firms
decide on advertising at the first stage and compete in prices at the second
stage. The building blocks we use to model the two stages are well known
in the literature. To address possible manipulation of consumer preferences
at the first stage, we use the Bayesian learning framework with normally dis-
tributed beliefs, which is applied in economics (e.g., Prendergast and Stole
1996, Meyer and Vickers 1997, Holmstrom 1999, Hoffler and Sliwka 2003) as
well as marketing (e.g., Erdem and Keane 1996, Mehta et al. 2003, Janaki-
raman et al. 2009, Goettler and Clay 2011). In this setting, advertising can
influence consumer expectations about the goods’ uncertain perceived quali-
ties. The more a firm invests in advertising, the higher will be the consumers’
posterior perceived quality of the firm’s good. For modeling price competition
at the second stage, we build on the seminal paper by Shaked and Sutton
(1982).
The major novelty in our setting is that we allow for consumers to violate
Bayes’ rule when processing information. Our modelling of perceptual biases
builds on the observation by Koszegi (2014, p. 1085) that decision makers
are typically prone to two general mistakes – ”systematically incorrect priors,
and mistakes in updating beliefs based on information.” We simultaneously
incorporate both kinds of mistakes in our setting. Hence, consumers might be
too optimistic or too pessimistic when assessing a good’s consumption quality
prior to the updating process. At the same time, they might put too much or
too little weight on updating when new information about a good’s consump-
tion quality arrives. We allow for all possible combinations of these perceptual
6Advertising can influence consumer behavior in three different ways (Bagwell 2007,
1708–1724; Belleflamme and Peitz 2010, 135–139). First, it can provide consumers with
useful information about the existence, the price, and specific characteristics of a good.
Second, advertising may complement a good by enhancing its degree of popularity and,
thereby, the consumers’ utility from using or consuming it. Third, advertising may be
neither informative nor complementary but purely persuasive.7See, among others, Braithwaite (1928), Von der Fehr and Stevik (1998), Bloch and
Manceau (1999), and Buehler and Halbheer (2012).
3
biases. However, to avoid redundancies, we will focus on the most prominent
cases (see, for example, Offerman and Sonnemans 1998) – the existence of too
optimistic priors (i.e., consumer optimism) and too little updating (i.e., con-
sumer conservatism). All the remaining biases can be analyzed analogously.
At first sight, one would expect that firms reduce persuasive advertising
if it is less effective due to perceptual biases. For example, if conservatism is
widespread among consumers, i.e., they partially ignore new information and
stick to their prior beliefs about a good’s consumption quality, the influence of
advertising may be severely limited. Firms that anticipate such biases should
choose less persuasive advertising compared to a situation without percep-
tual biases. However, we show that the biases also influence product market
competition, which can result in more advertising.
To understand the effects of perceptual biases on product market compe-
tition consider the price competition at stage two of the model. The firm that
supplies the good with the higher perceived quality – the posterior market
leader – has a clear competitive advantage. Contrary to the rival firm, it can
charge a higher price without losing too many consumers because the higher
perceived quality of its good leads to greater consumer loyalty. We show that
the posterior market leader earns strictly larger profits than the rival firm, but
the profits of both firms are increasing in the spread between the perceived
qualities of the two goods. At the first stage, a firm can choose advertising
to increase the perceived quality of its good. As both firms’ profits increase
in the perceived quality spread, each firm faces the following trade-off. On
the one hand, advertising will be beneficial if the firm is the posterior market
leader when entering price competition. On the other hand, advertising will
be detrimental ex post if the firm is not the posterior market leader.
These two counteracting effects explain how perceptual biases influence
firms’ marketing strategies at stage one. First suppose consumers have sys-
tematically incorrect priors. In this case, consumer optimism will increase a
firm’s advertising if the perceived quality of its good initially exceeds that of
the rival firm and if the beneficial starting point of the former one is pro-
tected by consumer conservatism. In that situation, the firm with the higher
mean quality at stage one – the prior market leader – will also be the pos-
terior market leader with high probability. This firm has strong incentives
to further increase the perceived quality spread and, hence, expected profits
by advertising. On the other hand, the rival firm is discouraged and chooses
low advertising. Overall welfare will decrease with consumer optimism if the
4
extra persuasive advertising of the prior market leader exceeds the decrease in
advertising of the discouraged rival firm.
Second, consider consumers whose information processing is biased by con-
servatism. Two effects on optimal advertising have to be distinguished. On
the one hand, conservatism hampers Bayesian updating and, thus, makes ad-
vertising less effective (attenuation effect). This effect decreases advertising
incentives. On the other hand, conservatism also influences price competition
at stage two (competition effect): The larger conservatism the more likely the
initially perceived quality ranking of the two goods will be preserved at stage
two. Such preservation of an asymmetric starting point encourages advertising
behavior of the prior market leader and discourages the rival firm, because both
maximize their profits by generating a large perceived quality spread. Alto-
gether, the two effects of conservatism make the rival firm decrease advertising,
whereas advertising of the prior market leader will be increased (decreased)
if the competition effect dominates (is dominated by) the attenuation effect.
Welfare losses due to persuasive advertising are boosted by consumer conser-
vatism, if the competition effect is sufficiently strong so that the additional
advertising of the prior market leader exceeds the reduction in advertising of
the rival firm. This outcome is particularly likely if consumers rate the goods
as rather different prior to the firms’ advertising decision.
Due to its wasteful characteristic there has been an ongoing discussion if
persuasive advertising should be regulated and how to implement regulation.
For example Peltzman (1973) investigates the U.S. drug market and the role
of consumer protection legislation. His starting point is the observation that
pharmaceutical firms often invest in advertising to exaggerate the effectiveness
of new drugs.8 A central topic is the question whether pharmaceutical firms
should be forced by law to prove the efficacy of new drugs. Our findings add
to this discussion as they show under which circumstances perceptual biases
lead to a strong need for regulatory policy and suggest how regulation could
be successful in enhancing welfare. Recall that consumer conservatism will
lead to a net increase in persuasive advertising in situations with a clear mar-
ket leader. Thus, policy measures counteracting a high degree of perceived
heterogeneity of goods can be useful to prevent the negative effects of con-
sumer conservatism. In particular, the prohibition of persuasive comparative
advertising may help to avoid that consumers perceive similar goods as overly
8Comanor (1986) reports that advertising expenditures in the pharmaceutical industry
considerably exceed R&D expenditures.
5
different, which would end up in excessive advertising by the perceived market
leader. Moreover, investment in consumer protection that lowers uncertainty
about the products’ consumption qualities (e.g., mandatory identification of
product characteristics) may make Bayesian updating less important for con-
sumers. Hence, it could also contribute to mitigate detrimental welfare effects
of consumer optimism and consumer conservatism.
Our paper is organized as follows. The next section summarizes the liter-
ature related to our paper. The model is introduced in Section 3. In Sections
4 and 5, we analyze the impact of consumer biases on firms’ advertising deci-
sions in monopoly and duopoly, respectively. Section 6 focuses on the welfare
implications of consumer biases. Section 7 concludes.
2 Related Literature
Our paper is related to two strands in the literature – the part of the behavioral-
economics literature that considers consumer biases, and the economic litera-
ture on advertising. The first strand of literature considers firms’ behavior in
response to consumers with non-standard preferences or cognitive biases (for
an overview, see Koszegi, 2014). In particular, we build upon the work that
investigates selling strategies in competitive markets populated by cognitively
limited consumers. Gabaix and Laibson (2006) and Dahremoller (2013) ex-
plore firms’ decisions if some consumers are not able to observe information
about a product add-on. They argue that, as a consequence of this consumer
myopia, hidden product details are shrouded by firms even in highly competi-
tive markets. Bordalo et al. (2013) go one step further and endogenize which
attribute of the good receives attention by consumers.
More closely related to our analysis are papers considering marketing poli-
cies. Li et al. (2014) investigate the incentives of a monopolist to disclose
potentially adverse product effects if some consumers are unaware of their
existence. They argue that firms will only disclose information through ad-
vertising if the share of unaware consumers is small. Otherwise, mandatory
disclosure policies may enhance welfare. In a similar spirit, Eliaz and Spiegler
(2011) assume that firms can engage in marketing to influence the set of al-
ternatives that a consumer perceives as relevant. Their paper fundamentally
differs from ours in the way how marketing influences consumers. While in
their setting marketing is used to directly manipulate the consideration sets
of consumers, we analyze firms’ persuasive advertising choices that enter the
6
decision process more indirectly through a manipulation of beliefs. The latter
approach is also taken by Shapiro (2006) who builds on the idea that ad-
vertising influences the likelihood of consumers remembering a positive past
experience with the advertised good. Considering only monopolists, the pa-
per argues that advertising expenditures are highest for intermediate product
qualities. Complementary, we find for persuasive advertising that firms with
perceived quality advantages tend to advertise most.
The second strand of literature deals with advertising in economics. Bag-
well (2007) offers a comprehensive overview of the relevant literature. Our
paper is most closely related to the economic literature on combative and per-
suasive advertising, as we assume that total demand by consumers is exoge-
nously given and advertising does not provide additional utility for consumers.
Hence, if a firm attracts additional consumers by investing in persuasive ad-
vertising, rival firms will typically be harmed by losing these consumers.
According to Leiter (1950), the problem of combative and persuasive adver-
tising was first addressed by Marshall (1919), who emphasizes the social waste
of redistributing consumers from rival firms. In his empirical study, Lambin
(1976) uses data on advertising, sales, quality, and prices from 107 individual
brands to analyze how an increase in advertising expenditures by one firm in-
fluences own sales as well as the sales of rival firms. His results support the
notion that advertising is combative. Further evidence comes from Metwally
(1975, 1976) and Kelton and Kelton (1982). Similarly, Cubbin and Domberger
(1988) and Thomas (1999) show that firms respond to entry by rival firms with
increased combative advertising. More recently, Chen et al. (2009) theoret-
ically and experimentally investigated combative advertising. Theory shows
that while in some situations combative advertising leads to a more intense
price competition, there are other situations in which combative advertising
dampens price competition. The experimental results strongly support the
theoretical findings. Chioveanu (2008) argues that persuasive advertising may
not only influence the intensity of price competition but also lead to asym-
metric advertising and price choices by firms. This provides an alternative
explanation for price dispersion phenomena. Contrary to ours, none of these
papers offer an analysis of the impact of consumer biases on combative adver-
tising.
7
3 The Model
We consider a two-stage market game in which two risk-neutral and profit-
maximizing firms decide on advertising at stage one and, thereafter, compete
in prices at stage two. Each firm i (i = A,B) offers a good i whose consump-
tion quality is uncertain in advance, because it depends on the specific match
of individual consumer preferences and the characteristics of the good. For
example, suppose that each firm produces an innovative consumption good
(e.g., a gadget). In this case, some individuals may enjoy consumption of the
good, whereas others may be less enthusiastic after consumption. Concerning
the Samsung example from the introduction, on the one hand the consump-
tion quality crucially depends on the technical features of the new smartphone
whose usefulness is uncertain for a consumer. On the other hand, consumption
quality also depends on the specific preferences of an individual consumer (e.g.,
whether an individual extensively uses all new features). This is uncertain for
the firms. Altogether, neither consumers nor firms know the consumption
quality of a good ex ante with certainty.
Our analysis of the price competition at stage two is based on the model of
Shaked and Sutton (1982).9 Production costs are normalized to zero and the
mass of risk-neutral consumers is assumed to be one. Consumer types θ are
uniformly distributed over [¯θ, θ] with
¯θ ≥ 0, θ ≥ 2
¯θ and θ =
¯θ+1 such that the
density is 1. A consumer of type θ receives utility θ · qi − pi from purchasing
one unit of good i (i = A,B) at price pi. The variable qi denotes the ex
ante uncertain consumption quality of good i and is normally distributed with
positive mean qi0 and variance σ2i0, i.e., qi ∼ N (qi0, σ
2i0). Consumer type θ
indicates the importance a consumer attaches to the consumption of the given
good. While consumers with high levels of θ enjoy the consumption of a good
with positive qi more intensely than their counterparts with lower θ, they will
also suffer more intensely if qi is negative.
At stage one, firms choose the intensity of advertising. Subsequently, con-
sumers and firms observe signals sA and sB about the consumption quality
of each good in the market. The value of each signal is determined by the
underlying true consumption quality of the good and the previously chosen
9The following model set-up with vertical differentiation is frequently used in the in-
dustrial organization literature. It can also be found in the marketing literature. See, for
example, Mehta et al. (2003), who use the same consumer utility function with normally
distributed product qualities to discuss consumer search under price uncertainty.
8
advertising by the firm:
si = qi + ai + εi (i = A,B). (1)
Here, εi ∼ N (0, σ2ε) denotes exogenous noise and ai ≥ 0 firm i’s advertis-
ing choice. Advertising ai leads to costs c (ai) with c (0) = c′ (0) = 0 and
c′ (ai) , c′′ (ai) > 0 for ai > 0. Moreover, we assume that c′′ is bounded from
below and above with c′′ ∈ [¯c, c]. To avoid technical problems, we assume
that ai has a finite upper bound. All random variables are assumed to be
statistically independent.
After having observed si, consumers update their prior beliefs about the
distribution of the consumption qualities, qi ∼ N (qi0, σ2i0), and form a pos-
terior distribution. From DeGroot (1970) we know that Bayesian updating
conditional on si would lead to a posterior distribution qi ∼ N (qi1, σ2i1) with
qi1 = qi0 +σ2i0 (si − ai − qi0)
σ2ε + σ2
i0
and σ2i1 =
σ2i0σ
2ε
σ2ε + σ2
i0
,
where ai denotes consumers’ belief about firm i’s advertising decision. Hence,
consumers’ posterior mean of good i’s consumption quality will be larger than
the prior mean, if and only if the realized quality signal exceeds its expected
value.
However, we intend to study the consequences of perceptual biases on the
advertising choices of firms. In general, consumers may be prone to two dif-
ferent kinds of mistakes (see also Koszegi 2014, p. 1085). First, empirical
evidence shows that consumers have systematically incorrect priors. In partic-
ular, they are overly optimistic about future life events (Weinstein 1980, 1982,
1989). Second, consumers make systematic mistakes when updating their pri-
ors according to new information. For example, they may firmly stick to their
priors and partially ignore new information (Philips and Edwards 1966). We
simultaneously incorporate both kinds of mistakes into our setup in a stylized
way.10 For this purpose, we assume that consumers deviate from Bayes’ rule
and form their posterior beliefs according to
qi1 = α · qi0 + β ·σ2i0 (si − ai − α · qi0)
σ2ε + σ2
i0
(2)
with i = A,B. Each kind of bias is described by one parameter (α > 0 and
β > 0, respectively). Here, α indicates consumers’ misperception of the prior.
10Brav and Heaton (2002) and Brandt et al. (2004) model conservatism by overweighting
the prior mean and underweighting Bayesian updating at the same time. However, we follow
the suggestion of Koszegi (2014) and disentangle both effects.
9
If α > 1 for example, consumers will be optimistic about future consumption
utilities. If in contrast α < 1, consumers are pessimistic. We name the first
part of the updating formula (αqi0) the biased prior mean of the consumption
quality of good i.
The parameter β describes the weight consumers attach to new information.
The smaller β, the fewer consumers update their prior information. If β < 1
for example, consumers will tend to stick to their priors and update these more
conservatively than a Bayesian consumer would do. In this case, we speak of
consumer conservatism. If, in contrast, β > 1, consumers will put excessive
weight on new information about consumption quality.
The timing of the game is the following. At the beginning of stage one, the
two firms simultaneously choose their advertising intensities aA and aB, leading
to publicly observable signals sA and sB, respectively. At the end of this stage,
consumers update their beliefs according to (2). At the beginning of stage
two, firms simultaneously choose prices pA and pB. Thereafter, consumers
make their purchasing decisions.
As a solution concept we apply perfect Bayesian Nash-equilibrium, with
the only distinction that consumers form their quality perceptions according
to (2).11 Thus, an equilibrium of the game consists of a strategy profile in-
corporating the strategies of both firms and all consumers and a belief system
such that the following three statements hold. First, both firms play mutually
best responses, anticipating consumer behavior. Second, on the equilibrium
path consumers derive their quality perceptions from firms’ advertising choices
according to (2). Third, consumers’ choices maximize their perceived utility.
The following analysis investigates how perceptual biases – i.e., variations
of α and β – affect equilibrium advertising and overall welfare. We analyze
variations of α and β for their entire domain. However, to ensure readability of
the analysis, our wording is oriented to the prominent case in which consumers
are overly optimistic regarding consumption quality (α > 1) and characterized
11We are interested in how systematic mistakes in information processing changes firms’
advertising decision. Hence, we apply a solution concept that is as closely related as possible
to Nash-equilibrium and at the same time is adequate to incorporate perceptual biases. This
ensures comparability to the case of rational consumers, which is indeed incorporated in our
setting. The approach to use Nash-equilibrium, despite agents having cognitive limitations
or non-standard preferences is also common to the literature (see, for example, Gabaix and
Laibson 2006, Eliaz and Spiegler 2011 and Bordalo et al. 2013). An interesting alterna-
tive to model perceptual biases would be to adopt an equilibrium notion that incorporates
systematically wrong beliefs of the consumers about firms’ advertising intensity.
10
by consumer conservatism (β < 1).12
4 The Monopoly Case
We start our analysis with the benchmark case of a monopoly. Hence, only
one firm, say A, is active whereas the other firm stays outside the market. We
assume that each consumer buys at most one unit of good A. A consumer of
type θ will buy one unit if θqA1 − pA ≥ 0 and has zero utility otherwise.
When deciding on the unit price at stage two, firm A anticipates that the
fraction θ − pA/qA1 of consumers will purchase good A if it chooses price pA.
The firm, thus, maximizes
pA ·
(
θ −pAqA1
)
,
which yields the solution p∗A = θqA1/2 and the optimal profit θ2qA1/4. Since
qA1 positively depends on sA (see (2)) and, hence, on aA, advertising increases
the posterior mean of the uncertain consumption quality, leading to a higher
monopoly price and higher monopoly profits.
At stage one, firm A decides on the level of advertising and solves
maxaA
E
[
θ2qA1
4
]
− c(aA)
=maxaA
θ2
4
(
αqA0 + βσ2A0 (qA0 + aA − aA − αqA0)
σ2ε + σ2
A0
)
− c(aA)
with E denoting the expectation operator with respect to qA and εA. Straight-
forward computations lead to the following result:
Proposition 1 In the monopoly case, firm A’s optimal advertising, a∗A, is
described byθ2
4β
σ2A0
σ2ε + σ2
A0
= c′(a∗A) . (3)
Proposition 1 shows that the stronger consumer conservatism (i.e., the
lower β), the less effective is advertising. As a consequence, a monopolist
decides to advertise less. Since this deterrent effect on advertising choices is
due to an attenuated impact of advertising on consumer decisions, we call it the
attenuation effect of consumer conservatism. In contrast, consumer optimism
does not impact advertising choices in the monopoly case. Although optimism
12See, e.g., the experimental findings by Offerman and Sonnemans (1998).
11
leads to a higher biased prior mean and therefore also a higher posterior mean,
it does not influence the (marginal) impact of advertising.
Furthermore, the proposition shows that optimal advertising decreases in
the magnitude of noise. Intuitively, with a less precise quality signal sA (i.e.,
a low value of 1/σ2ε), it becomes unattractive for the monopolist to invest in
advertising since consumers react little to the signal. Finally, the higher the
prior uncertainty about consumption quality (i.e., the higher σ2A0), the higher
is the optimal level of advertising.13 This effect is also intuitive. If, initially, the
uncertainty about the unknown consumption quality is large, by advertising
the firm can shift a lot of probability mass under the density of qA to the right.
5 The Competition Case
5.1 Equilibrium Behavior
In this section, both firms, A and B, are active and compete for consumers
via advertising at stage one and setting prices at stage two. For simplicity,
we concentrate on the case of combative advertising where total demand is
fixed and advertising only influences the allocation of consumers across firms.
Hence, we assume that each consumer buys exactly one unit of either good
A or B.14 We will see that a key determinant of firms’ advertising choices at
stage one and price choices at stage two is the relative perceived quality of the
goods. At stage one, the relative perceived quality, α(qi0 − qj0), depends on
consumer optimism and the expected true consumption qualities; at stage two,
it is given by qi1− qj1. For an easy understanding of the driving forces, we will
say that firm i is the prior market leader if it provides higher mean quality at
stage one (qi0 > qj0). Depending on advertising choices and resulting quality
signals, consumers update their quality perceptions before they purchase a
good. If a firm achieves a higher perceived quality after the advertising stage
(qi1 > qj1), we will call this firm the posterior market leader.
To determine the equilibria, we use backward induction and start by solving
the price competition game at stage two. Suppose, w.l.o.g., that firm i is the
posterior market leader. The resulting equilibrium prices are derived in the
13The same outcome does not necessarily hold for the competition case. See the note by
Grunewald and Krakel (2015).14In our setting, this corresponds to assuming that the utility of consuming no good is
negative infinity as done for example in Heidhues and Koszegi (2008).
12
appendix.15 They are given by
p∗i =(qi1 − qj1)(θ + 1)
3and p∗j =
(qi1 − qj1)(2− θ)
3, (4)
leading to equilibrium profits
π∗
i = Θ · (qi1 − qj1) and π∗
j =¯Θ · (qi1 − qj1) (5)
with Θ := (θ + 1)2/9 > (2 − θ)2/9 =:¯Θ. Due to its higher perceived quality,
firm i can charge a higher product price than firm j without losing too much
market share. The derivations in the appendix show that prices are strategic
complements, which is a typical finding for price competition. Altogether,
firm i chooses a high price in equilibrium and firm j reacts by choosing a high