You are on page 1of 16

International Journal of

International Journal of Industrial Organization Organization
ELSEVIER 14 (1996) 227-242

Coupons and oligopolistic price discrimination

H e l m u t Bester a'*, E m m a n u e l Petrakis b't
"CentER, Tilburg University, P.O. Box 90153, 5000 LE Tilburg, Netherlands
bUniversidad Carlos Ili de Madrid, Departamento de Economia, Calle Madrid 126, 28903
Getafe, Spain

Received 21 January 1994; accepted 8 November 1994


This paper studies sales promotion through coupons in a duopolistic market.

Sending out coupons allows the sellers to separate market segments with different
degrees of consumer brand loyalty. This kind of price discrimination is profitable for
the individual seller when the cost of couponing is sufficiently low. In equilibrium,
however, couponing increases competition and reduces profits. A n increase in the
cost of couponing decreases consumer surplus while the impact on profits and social
surplus is ambiguous.

Keywords: Advertising; Coupons; Oligopoly; Price discrimination

J E L classification: D43; D83

1. Introduction

This paper studies sales promotion through coupons and rebates in an

oligopolistic industry. Sales promotion is a complementary, and in some
cases even more important marketing strategy than media advertising for a
firm to increase its market share. Coupons and rebates count for the bulk of
b i l l i o n s o f d o l l a r s s p e n t e a c h y e a r o n t h e s e p r o m o t i o n a l activities. In 1989,
* Corresponding author at: Department of Economics, Free University Berlin, Boltzmannstr.
20, D-14195 Berlin, Germany.
t We wish to thank two referees for their comments.

0167-7187/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved

S S D I 0167-7187(94)00469-2
228 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

nearly $100 billion worth of coupons (potential savings) were sent out to
consumers in the United States through Sunday newspaper inserts (see
Koselka, 1990). Manufacturers in the United States distributed 310 billion
coupons in 1992, representing a 6% increase over the year before (see
H u m e , 1993). Rebates are virtually equivalent to coupons except that they
impose some additional redemption cost on the customer, namely the cost
of mailing and waiting for the cash refund. For our purposes coupons and
rebates play the same role and we will treat them interchangeably in our
We investigate the role of coupons in a market that is segmented due to
location or brand loyalty. By offering a rebate, a manufacturer can attract
some consumers who are otherwise more inclined to buy a competing brand.
In the location interpretation of our model, consumers face a transportation
cost. Supermarkets, department stores, and shopping-malls can use coupon-
ing of specific residential areas to compete more effectively in their rivals'
h o m e markets. In the alternative interpretation, consumers differ in their
degree of brand loyalty. By price discriminating between his own customers
and the clientele of competing manufacturers, the seller can increase his
own market share. Of course, this requires marketers to use tightly focused
coupon distribution programs as they are offered by special marketing
companies. Targeted coupons have been used successfully by Trump's
Castle to lure gamblers from competing casinos, by Stop & Shop stores to
win customers from other grocery stores, and by L'Oreal Visuelle to attract
users of Cover Girl and Revlon makeup (see Zeldis, 1987).
With couponing the manufacturer has at least partial control of the type
of household reached. Thus, he is able to offer a reduced price to a specific
segment of the market. Coupons can serve as a price discrimination device.
Only those consumers who have received a coupon are able to benefit from
the rebate. Since this enables the manufacturer to separate market seg-
ments, coupons are more than just a low price offered to all consumers. It is
worth mentioning that the use of coupons as a discriminating device does
not depend on the market structure. Coupons would also be used in a
m o n o p o l y version of our model. An alternative explanation of coupons has
b e e n proposed by C r e m e r (1984) and Caminal and Matutes (1990). In these
models, coupons create a lock-in effect because the consumer is offered a
rebate on future purchases of the product. The seller can stabilize his
market share by creating an artificial cost of switching suppliers, This is in
direct contrast with our model, where coupons tend to make switching more
attractive. Yet another explanation (Gerstner and Hess, 1991) is that the
seller can motivate retailer participation in the sales promotion by offering
rebates to the consumers.
T h e literature on the price discrimination aspects of coupons is rather
scarce. Narasimhan (1984) studies the consumers' decision to use a coupon.
H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242 229

Only consumers with a sufficiently low opportunity cost of time take

advantage of coupons. Thus, price discrimination can be achieved through
self-selection. Caminal and Matutes (1990) is, to our knowledge, the only
study of oligopolistic behaviour, but in a repeat purchase context. In their
model the consumer receives a rebate only after purchasing a second unit
from the same seller. The role of coupons is to create a switching cost in the
second period. If the firms precommit to a discount, competition in that
period is decreased and the equilibrium profits increase.
In contrast, in our model coupons increase competition between firms.
Each individual seller has an incentive to reduce the brand loyalty of the
other firms' clientele in order to increase his market share. But, offering a
rebate amounts to reducing the consumers' switching cost and so competi-
tion is intensified. In equilibrium, each seller's profits are lower than if
coupons or price discrimination were not allowed. Price discrimination in
combination with oligopolistic competition leads to lower prices; the
consumers as a whole are better off when the sellers compete by using
The finding that firm profits may be lower as a result of price discrimina-
tion is similar to the findings of Thisse and Vives (1988) and Anderson and
L e r u t h (1993). The general idea is that firms compete on more fronts with
discrimination. Levy and Gerlowski (1991) show that "meeting competition
clauses", whereby a seller announces to meet the competitor's price, may
reduce equilibrium profits. We may view such clauses as a special type of
coupon; they allow the seller to discriminate between those consumers who
only receive his own ad and those who receive ads also from other firms.
The paper is organized as follows. Section 2 presents a simplified
example, where we abstract from the costs of issuing coupons and from
differences in the consumers' degree of brand loyalty. Section 3 describes
the general model. We study the equilibrium marketing behaviour of the
firms in Section 4. Section 5 provides some comparative statics and welfare
results. All proofs are relegated to an appendix.

2. An example

T o illustrate the main features of our analysis, we first present a simplified

example of the more general model studied in the following sections. We
study a duopoly market characterized by some segmentation according to
location or brand loyalty. Using the language of the locational application,
the model considers the following situation. There are two sellers, A and B,
located in different neighbourhoods of some geographical market. The
sellers' production costs are normalized to zero. In each locality there is a
unit mass of consumers with a totally inelastic demand for one unit of the
230 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

good for all prices between zero and v > 0. When a consumer purchases the
good from the seller in the distant location, he has to pay a transportation
cost, s < v.
Shilony (1977) studies the equilibrium of this market when the sellers
compete by setting prices in a standard Bertrand fashion. If v <~ 2s, both
sellers will post the price p * = v in equilibrium. Undercutting the com-
petitor's price is not profitable since this would yield a profit of at most
2 ( v - s)~< v. The price-setting game between the duopolists fails to have a
pure strategy equilibrium if v > 2s. Indeed, any combination of prices
(PA, PB) would allow one of the two sellers to gain either by undercutting
the other seller or by increasing his price by some small amount. Shilony
(1977) shows that there is a unique symmetric mixed strategy solution where
each seller gains an expected profit higher than s.
We now introduce oligopolistic price discrimination. Even though the
seller is unable to identify the origin of the customers at his store, he can
separate them through coupons. Each seller is able to offer the good at
different prices in the two regions by mailing coupons to the other region.
Coupons entail a legally binding promise by the seller to offer a rebate upon
presentation. In our example we abstract from mailing costs and assume that
seller i can costlessly send coupons to all consumers in region j. The coupon
entitles the bearer to buy the good from seller i at the price Pi -- ri, while
buyers without a coupon have to pay Pi.
In this simple example, competition between the duopolists, A and B,
results in the following equilibrium outcome:
P -PB * =p* =s, r A* = r * : r* =s.
B (1)
As is usual in this kind of model, the equilibrium is based on the tie-
breaking rule that consumers buy from the closer firm if prices are equal.
This can be justified by taking the limit as heterogeneous consumer tastes
b e c o m e homogeneous (see Anderson and de Palma, 1988). Each consumer
is indifferent between buying the good from seller A or seller B. If he buys
at the neighbouring store, has to pay the price p* = s; otherwise, has to pay
p* - r * = 0 but incurs the transportation cost s. In equilibrium, it must be
the case that each consumer buys at the local store. If not, the local seller
would have an incentive to lower his price slightly below s. Clearly, the
outcome described by (1) constitutes an equilibrium: by charging a price
above s a seller would lose all his customers. Charging a price below s can
never be optimal, since each seller enjoys a local monopoly position for all
prices up to s. Actually, the equilibrium is unique (after elimination of
dominated strategies with Pi - ri < 0)" if some seller i were to charge a price
Pi > s, then the opponent could induce all consumers in region i to switch by
offering pj - rj slightly below Pi - s. Accordingly, a price Pi > s cannot be
part of equilibrium behaviour. The same argument shows that both sellers
H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242 231

must offer the rebate r* = s. Obviously, given p* this is the highest rebate a
seller is willing to offer. If seller i sets r~ < s, then seller j would optimally
charge his local customers the price pj > s, which was already shown to be
inconsistent with equilibrium.
The example demonstrates that price discrimination increases competi-
tion. The consumers have to pay lower prices and the finns' profits are
reduced. Indeed, each seller earns a profit of s, which is lower than the
profit he gets in the absence of discriminatory pricing. In what follows we
will verify this observation in a more general model. In fact, the above
example has some unappealing features because the consumers' purchasing
decisions have to be based on a tie-breaking rule. Even though the
equilibrium requires both sellers to use coupons, this marketing instrument
is actually ineffective since no consumer is induced to switch. If the sellers
had to pay a mailing cost, the above equilibrium would therefore collapse.
The subsequent model will overcome these difficulties by introducing some
consumer heterogeneity.

3. The model

Following Shilony (1977), we study a market where consumers can

purchase costlessly from a neighbourhood store, but incur some visiting cost
if they venture to a more distant store. There are two firms, A and B,
located in different regions that produce a homogeneous good at zero cost.
Each region is inhabited by a unit mass of consumers who have a common
reservation utility v for the good. Each consumer can costlessly visit the
store at his home location, while he has to pay a transportation cost s to go
to the other seller. It is assumed that s is uniformly distributed on [0, s]
across the population of consumers in each region. This assumption differs
from Shilony's (1977) model of mixed pricing in oligopoly, where all
consumers have the same visiting cost. It generates continuous demand
functions, and guarantees the existence of a pure price-setting equilibrium in
the absence of couponing (see Bester, 1992).
The firms offer the good at prices PA and PB, respectively. Without loss of
generality, let 0 ~<Pi ~< v, i = A, B. The seller cannot distinguish buyers from
different regions once they enter the store. Similarly, he is uninformed
about the visiting cost of the individual buyer. This means that he cannot
make his price offer contingent upon such information. But, seller i may
p r o m o t e purchases by offering the consumers at location j ~ i a rebate r~,
with 0 <<-r i <~P i . T h e seller may offer such rebates by mailing coupons to the
other region. Since he cannot discriminate between consumers with different
values of s, offering a single rebate r~ at location j is optimal. Also,
distributing coupons in the local market cannot increase seller i's profit.
232 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

A f t e r all, the purpose of couponing is to attract some of those consumers

who incur switching costs.
W h e n firm i sends coupons to a fraction h i ~ [0, 1] of the consumers at
location j, it has to pay the mailing cost k(hi). Each consumer is equally
likely to receive the rebate. This means, with probability h i that a consumer
in region j has to pay only Pi - ri for the good available at location i. T h e
b u y e r who has not received a coupon is not entitled to a rebate and has to
pay P r We assume that consumers do not interact with each other so that
trading coupons is not possible.
T h e marketing strategy of firm i may be described by x i = (pi, r i, Ai). The
cost function k ( . ) is assumed to satisfy the following restrictions:
k(0)=0, k'(h)>0, k"(h)>0, k'(0)<3/4, k'(1)>5/9. (2)
T h e marginal cost of couponing is increasing in the n u m b e r of households
reached. This assumption is standard in the advertising literature (see
Butters, 1977, and G r o s s m a n and Shapiro, 1984). The underlying idea is
that the manufacturer can distribute coupons via mail and by placing ads in
a set of magazines or newspapers. The probability that a given consumer
receives a coupon through one of these media is independent of receiving a
c o u p o n through the other media. In this case, the cost of making sure that a
fraction A of consumers receives at least one coupon becomes a convex
function of A. The additional assumptions on k ' ( . ) guarantee that each
m a n u f a c t u r e r will choose some advertising intensity 0 < h i < 1. In addition,
we restrict the analysis to the case v >3. This ensures that competition is
sufficiently strong so that setting Pi = v cannot be optimal for seller i.

4. Equilibrium

We assume that each consumer is aware of the availability of the good in

b o t h regions. In addition he knows the price charged by each of the sellers.
In this situation, advertising conveys no information about the existence or
the price of a good. Distributing coupons only serves to price discriminate
b e t w e e n buyers from different regions. By offering a rebate the seller
increases the attractiveness of his product for those consumers who have to
spend the cost s in order to visit his store.
T o c o m p u t e each seller's demand, we have to distinguish four groups of
consumers. In each of the two regions the purchasing decision of the
consumers who have received a coupon differs from those without a coupon.
A consumer at location A who has not received a coupon from firm B
purchases the good at his h o m e location as long as PA ~<PB + S. If, however,
he gets a coupon of value r B, he will buy from seller A only if PA <~PB --
r a + S. W h e n consumer s at location B is not offered a rebate, he will be
H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242 233

attracted by firm A's price offer ifpA + S ~<PB" Otherwise, with a coupon r A,
he will purchase from firm A if PA - rA + s ~<PB.
Given the consumers' demand decisions, firm A's profit, IiA(XA,XB),
depends on both firms' sales strategies and is given by

HA(XA, XB) = PA( 1 -- AB)D(s - PA + PB) + PAABD(~ - - PA + PB - - re)

+PA( 1 -- AA)D(PB - P A ) + (P* - rA)AAD(p. --PA + rA)
- kOA), (3)

D(z)=--z/~, for 0~<z~<~,
D(z)=-O, for z~<0, (4)
D ( z ) = 1, for z ~>~.
By symmetry, firm B's profit equals HB(XA, XB)= HA(XB, XA). To study
equilibrium advertising in this market, we consider the Nash equilibrium in
the sellers' game. Thus, a pair (x*, x~) of marketing strategies constitutes an
equilibrium if IIA(X~, X*) >t IIA(XA, X~) for all XA, and liB(x*, x~) ~ l i s ( x * ,
XB) for all x B.
Our first result characterizes the symmetric pure strategy equilibrium and
provides a sufficient condition for its existence. To state the result, we define
p(h) --= - k " ( h ) / k ' ( A ) . In utility theory, p(- ) is well known as the coefficient
of absolute risk-aversion. Negative values of p(A) indicate the degree of
convexity of k(h) at h.

Proposition 1. Let (XA,XB)

* * be an equilibrium such that X *A = X R
(p*, r*, A*). Then (p*, r*, A*) is given by the solution to
p* =~/(1 + 0.5A*), r* = 0.5p*, k'(A*) = p * / ( 4 + 2A*).
Moreover, a sufficient condition for the existence of a symmetric pure strategy
equilibrium is that p(A) ~< - 2 for all A @ [0, 1].

Coupons increase competition between the sellers and reduce their

profits. Indeed, it is easily established that both sellers would charge the
price/~ =~ if distributing coupons were not possible. By Proposition 1, all
consumers pay a price below ~ in the equilibrium with couponing. The
consumers gain and the sellers lose by the feasibility of couponing.
Fig. 1 illustrates that the equilibrium is indeed unique: The P - P schedule
depicts all p-A combinations such that p =~/(1 + 0.5A); the K - K schedule
describes the function p = k'(A)(4 + 2A). Assumption (2) ensures that the
two functions intersect at some point (p*, A*) within the area (0, ~) x (0, 1).
This intersection determines the equilibrium values p* and A*.
234 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

/ (P*~)
Fig. 1. Equilibrium (p*, h*).

With the help of Fig. 1 we can easily establish some comparative statics
properties of the equilibrium outcome (p*, r*, A*). For instance, an increase
in the marginal cost k ' ( . ) leads to a higher equilibrium price p* and a lower
level of advertising A* because the K - K schedule is shifted upwards. How
does the equilibrium react to a change in the consumers' transportation
cost? Increasing the parameter ~ is equivalent to increasing each consumer's
switching cost by the same factor. In Fig. 1 this leads to an upward shift of
the P - P schedule. Consequently, both p* and A* are increased.
Not all consumers who are couponed will make use of the rebate.
Redeeming the coupon is worthwhile only when s ~< r* = p * - r * . Accord-
ingly, the redemption rate is ( p * - r * ) / ~ = 1 / ( 2 + A * ) . That is, more than
one-half of the coupons are not returned to the manufacturer. Using the
above comparative statics results, it follows that the equilibrium redemption
rate is increasing in the marginal cost of couponing. A cost increase reduces
the n u m b e r of households that are couponed, but the fraction of households
that redeem the coupon is increased. An increase in the consumers'
transportation cost has the opposite effect: more households are couponed,
but a lower fraction return the coupons. Note, however, that the total
n u m b e r of coupons actually redeemed is given by A*/(2 + A*), which is
increasing in 3,*. This number is, therefore, negatively related to the
marginal cost of couponing and positively related to the level of consumer
switching costs.
Of course, the result that the coupon offers a 50% price rebate is specific
to the setting of our model. In particular, the uniform distribution of
switching costs is important in this context. We can show that the sellers
would optimally offer a lower rebate if the distribution of s puts more weight
on low switching costs. The intuition is that attracting high cost consumers
through a rebate is not profitable when these consumers represent only a
small fraction of the total population.
Proposition 1 only provides an implicit characterization of the equilibrium
H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242 235

values p*, r*, and A*. To obtain an explicit solution of the equilibrium, we
have to look at a parametric example of the cost function k(- ). The simplest
example is k(A)= cA 3. This function satisfies restriction (2) if-s/c < 27. The
equilibrium described by Proposition 1 is then given by

p* = 213d + S ( 3 C S ) 1 / 2 ] 1/2 -- 2(3cs) l/z, (5)

A* = [1 + S ( 3 d ) - 1 / 2 ] 1/2 -- 1.

Note that within the family of cost functions k()t)= cA ", Proposition 1
guarantees the existence of a symmetric pure equilibrium as long as a / > 3.
The equilibrium characterization by Proposition 1 is derived from the
first-order conditions that the profit-maximizing marketing strategies xA and
x B necessarily have to satisfy. Unfortunately, the first-order conditions are
not sufficient for profit maximization. By (3), the firms' profit functions are
concave in Pi, ri and Ai, but not jointly concave in (p~, r~, A~). Because of this
non-concavity, we have to be careful that no seller can increase his profit by,
for instance, simultaneously lowering his price and his advertising intensity.
The restriction p(A) ~ - 2 guarantees that such deviations from (p*, r*, A*)
are not profitable. For any given A~, the first-order conditions determine
p*()ti) and r*(Ai) so that seller i's profit can be written as
///(p*(A~), r*(Ai), A~, xj). Since p(A) ~< - 2 ensures that the cost function
k(. ) is sufficiently convex,//~,.(p* (Ai), r* (Ai),)t i, xi) is concave in )ti. Accord-
ingly, the first-order conditions are sufficient for a global optimum if
p(a) <~ - 2.
For specific cost functions, we can prove the existence of a pure strategy
equilibrium even when the condition p ( A ) ~< - 2 is not satisfied for all
A ~ [0, 1]. The following result extends Proposition 1 to the case k(A) = cA ~,
a > 1, by imposing a restriction on ~/c. Also, we consider the cost function
k ( A ) = - c I n ( l - A ) , which Butters (1977) has derived from a simple
underlying advertising technology.

Proposition 2. I f k(A)= cA~, with a > 1, then there is a symmetric pure

strategy equilibrium for s/c sufficiently small. I f k ( A ) = - c l n ( 1 - A), then
there is a symmetric pure strategy equilibrium for s/c sufficiently large.

For the cost functions considered in Proposition 2 the parameter p(A) is a

monotone function of A. This allows us to replace the global condition on
p(. ) in Proposition 1 by a local condition. For k(A)= cA ~ we have p ( A ) =
( 1 - a)/A, which is increasing in A. In this case, the existence result is
obtained because A* becomes small for low values of ~/c. In contrast, for
k(A) = - c l n ( 1 - A) the parameter p(A) equals - 1 / ( 1 - A) and decreases
with A. Accordingly, we have to guarantee that A* is close enough to unity
by assuming that ~/c is large.
236 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

Even though the manufacturer's profit function is not jointly concave in

(Pi, ri, hi), it is continuous. This implies that there is a symmetric mixed
strategy Nash equilibrium when a pure equilibrium fails to exist (see
Dasgupta and Maskin, 1986). In such an equilibrium the sellers choose a
random marketing policy. Since for a given h i, seller i's profit is concave in
(Pi, r~), it follows from the first-order conditions that the optimal Pi is
positively related with h i. Similarly, the optimal r i increases with Pi. In the
mixed strategy equilibrium, therefore, each seller chooses h i stochastically
and he combines higher values of h i with higher prices Pi and higher coupon
values r i.

5. Welfare

In this section we investigate how the market participants' welfare

depends on the advertising cost k ( - ) and the transportation cost parameter
7. As a measure of consumer welfare, we consider aggregate consumer
surplus, C, which depends on (p*, r*, h*) according to
C(p*, r*, A*) = 2(1 - A*)(v - p*)

+ 2A* [ max[v - p * , v - p * + r* - s]f~ds. (6)

The first term represents the utility of the consumers who do not receive a
coupon. The other consumers' purchasing decision depends on the rebate r*
and the switching cost s. Consumer s will make use of the coupon only if
v-p*>~v-p * + r * - s . Finally, we define social welfare as the sum of
producer profits and consumer surplus. We begin by studying the relation
between welfare and s.

Proposition 3. The sellers' equilibrium profit is increasing in -s. Consumer

surplus and social welfare are decreasing in s.

Not surprisingly, producer profits are positively related to the level of

brand loyalty or geographical differentiation. When the sellers are able to
influence the parameter ~ by their choice of product differentiation, they will
seek to maximize 7. This is in line with the familiar principle of 'maximal
differentiation', as described by D ' A s p r e m o n t et al. (1979). An increase in
has a two-fold impact on consumer surplus. First, the consumer has to pay
higher prices. Second, he is more likely to receive a coupon. The above
result demonstrates that the first negative effect outweighs the second
positive effect. Indeed, as was shown before, the coupon redemption rate
decreases with 7. Finally, social welfare is negatively related to 7: the higher
H. Bester, E. Petrakis / Int. J. lnd. Organ. 14 (1996) 227-242 237

is 5, the higher is the coupon distribution intensity, A*, and the number of
coupons actually redeemed, ~ * / ( 2 + A * ) . This means that the resources
spent on couponing and the consumers' aggregate travel costs increase with

Proposition 4. Let k( A ) = cr/(A). Then consumer surplus decreases with c in

the symmetric pure strategy equilibrium. I f ~7(A ) = A ~, then equilibrium profits
and social surplus are increasing in c. I f r/(A) = -In(1 - A), then equilibrium
profits and social surplus are decreasing in c for s/c sufficiently large.

Higher couponing costs increase prices and reduce coupon distribution.

As a result, the consumers become worse off. The impact on equilibrium
profits and social welfare, however, is ambiguous. Of course, seller i gains
by a reduction in c when the marketing strategy xj of his opponent is kept
fixed. Similarly, lower resource costs of couponing increase welfare when
(x A, xB) remains unchanged. In addition to these direct effects, however,
there are indirect effects. Lower advertising costs make competition more
aggressive, which tends to reduce the sellers' prices and their profits. At the
same time, there is a negative impact on welfare as the sellers waste more
resources on couponing and the consumers switch sellers more often. Which
of these effects dominates, depends on the advertising cost function k ( . ) .
As the first part of Proposition 4 demonstrates, an increase in the cost of
couponing may make the sellers better off. A similar observation has been
made in models of informative advertising (see Bester and Petrakis, 1992;
Grossman and Shapiro, 1984; and Peters, 1984).

6. Conclusions

This paper has studied couponing a s a price discrimination device in

oligopolistic competition. By couponing those consumers who have some
preference for a competing brand, a seller can increase his market share.
Coupons may compensate the consumer for a costly movement to another
brand. In contrast to the repeat-purchase explanation, coupons reduce
consumer switching costs in our model. The price discrimination model
predicts that couponing intensifies competition between the sellers, leading
to lower prices and profits.
Our simple model may be extended in several interesting directions. We
assumed that all the consumers have the same valuation for the good. As a
result, total demand was fixed. Introducing some dispersion of consumer
valuations would make aggregate demand elastic. As couponing increases
competition, it would raise aggregate output. Couponing may have a
positive effect on social welfare if the elasticity of demand is sufficiently
238 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

high. Also, the value of the coupon relative to the price of the product
would presumably depend on the distribution of consumer valuations.
Removing the symmetric structure of the model would be another
interesting extension. In our model, the sellers were identical and so they
used the same marketing strategy. This would no longer be the case when
different production technologies are considered. An extension along these
lines could provide insights into the relation between a firm's efficiency and
its marketing policy. Similarly, we could study the role of a firm's size when
the number of consumers differs across market segments. We might expect
that smaller firms have a higher incentive to distribute coupons because they
can gain more by attracting consumers from other market segments.


L e m m a 1. L e t x i be an optimal marketing strategy f o r firm i given that f i r m j

chooses x i. Then the following m u s t hold: "(i)" pj --s <~Pi - ri <Pj; "(ii)"
pj - 5 ~<p~ < P i +5; and "(iii)" p~ >~ min[pj - %, 0.5pj].

Proof. If p i - r~ ~>pj, no consumer from location j will switch to firm i.

Therefore, firm i could increase its profit by not mailing coupons. If
p~ - r~ <p~ -~, all consumers at location j who receive a coupon from firm i
strictly prefer to switch. Therefore, firm i could increase its profits by slightly
decreasing r~. This proves (i). To prove (ii), note that all consumers at j
strictly prefer to switch to firm i if Pi <Pj---S" Therefore, firm i could
increase its profit by slightly increasing Pr If Pi ~>Pj +~, no consumer at
location i buys from firm i. But then firm i could get higher profits by setting
Pi slightly below Pi +~, while keeping pi - r~ constant. Finally, (iii) follows
by differentiating H/ in (3) with respect to p~ and keeping Pi -- r~ constant.
This yields OH~(XA, X B IPi -- re = const.)/Op~ > 0 for pi < min[pj -
r,, 0.5pj]. Q.E.D.

L e m m a 2. Let x i be an optimal marketing strategy f o r firm i given that f i r m j

chooses xj. Then Pi - ri = 0.5pj.

Proof. By the definition of //~(. ), P i -

ri must maximize ( P i - r i ) D [ p j -
( p ~ - r~)]. By Lemma 1, the solution must satisfy the first-order condition
Pi - 2(Pi - ri) = 0. Q.E.D.

L e m m a 3. Let p(A) -- - k " ( ) t ) / k ' ( A ) ~ - 2 f o r all A @ [0, A2]. Then k'(A2) >
(1 + }k2 -- } k l ) 2 k ' ( A 1 ) f o r all 0 ~< h~ < h 2 ~< 1.

Proof. Define G(A~, h2) -= k'(A2) - (1 + A2 - A1)2k'(A~). Then G ( ) t l , A2) = 0

H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242 239

for /~1 = /~2" Therefore, L e m m a 3 holds if 0G(A~, A2)/OA1 < 0 for all 0 ~</~1 <
h2~<l. Differentiation shows that OG(A1,A2)/OAI<O is equivalent to
p(A~) < - 2/(1 + A 2 - A1). Since - 2 < - 2/(1 + h 2 - As), L e m m a 3 holds if
p(A) ~< - 2 for all h ~ [0, h2]. Q . E . D .

Proof of Proposition 1. L e m m a 2 implies r* = 0.5p*. This together with the

first-order condition for the optimal choice of/~i implies k'(A*) = [p*/212/~.
We show that firm A cannot gain by setting PA > P * only if p* ~>~/(1 +
0.5A*). By L e m m a 2, firm A will optimally set PA - rA = 0.5p*. Therefore,
the profit from choosing PA E (p*, p* +~) together with AA = A* equals

//A = [pA(1 - **)(~ - P A + P * ) + PAA*(~ --PA + 0.5p*) + A*(p*/2)2]/~

-- k ( a * ) . (7)

This implies OIIA/OPA= [g -- 2pA + p * ( 1 -- 0.5A*)]/~. If p* <~/(1 + 0.5A*),

then for PA close enough to p* we obtain OIIA/OPA> 0 SO that firm A could
gain by charging a price slightly above p*. As a result, we must have
p* ~>~/(1 + 0.5A*).
Finally, we show that firm A cannot gain by setting PA < P * only if
p* ~<s/(1 +0.5A*). The profit from choosing p A E ( p * - r*, p*) together
with AA = A* and PA -- rA = 0.5p* equals

HA = pA(1 - A*) + [pAA*(~ --PA + 0.5p*) +pA(1 -- A*)(p* -- pA)]/'g

+ A*(p*/2)2/-g - k(A*). (8)

T h e r e f o r e , oHA/OPA= IS -- 2pA + p * ( 1 -- 0.5A*)]/~. If p* >~/(1 + 0.5A*),

then for PA close enough to p* we obtain a//A / OpA< 0 SO that firm A could
gain by charging a price slightly below p*. As a result, we must have
p* ~<s/(1 + 0.5A*).
T o prove the second part of the proposition, let firm B use the strategy
x B = ( p * , r*,A*) such that p*, r*, and A* satisfy the conditions of
Proposition 1. We will show that adopting the same strategy is a best
response for firm A.
First we show that choosing PA > P * is not profitable for firm A. By
L e m m a 2, firm A will optimally set PA - rA = 0.5,o*. Therefore, the profit
from choosing PA ~ (P*, P* +s) together with AA equals

H A = [pA(1 -- A*)(S --PA + P * ) +PAA*(~ --PA + 0.5p*) + AA[p*/2]2]/~

- k(AA). (9)

F r o m the first-order condition it follows that firm A will optimally set

AA = A*. This implies

OlIA/OPA=[(1--A*)(g--2pA +p*)+ A*(S--2pA +O.5p*)]/-g<O, (10)

240 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

where the inequality follows from PA > P * = 3 / ( 1 + 0.5h*). Thus, firm A

cannot gain by adopting some strategy x A ~ (p*, r*, h*) such that PA > P * "
TO complete the argument, we show that choosing p A < p * is not
profitable for firm A. As firm A will optimally set PA -- rA = 0.Sp*, its profit
f r o m choosing PA E (p* - - r * , p*) together with AA equals

H A =pA(1 -- h*) -t- [pAh*(S --PA + 0.5p*) +pA(1 -- hA)(p* -- pA)]/S

+ ha[P*/212/'g -- k(AA). (11)

F o r any PA < P * the optimal choice of h A has to satisfy the first-order

condition PA(P*--PA) = ~ [ k ' ( h * ) - k'(hA)], so that XA < h* by the convexity
of k ( . ). This equation can be rewritten as

pA2 --PAP* +~[k'(h*) - k'(AA) ] = 0 . (12)

N o t e that, by Proposition 1, p , 2 _ 4 ~ [ k ' ( h * ) - k'(hA) ] = 4~k'(hA) and that,

by L e m m a 1, PA >~ 0.5p*. Therefore, the above equation yields the solution

PA = 0.5p* + [-sk'(AA)] 1/2 (13)

Since 2 p A - - p * = Z ~ k ' ( h h ) ] ~/2 a n d - g - O . 5 k , * p * = p * = Z [ g k ' ( ) t * ) ] 1/2, dif-

ferentiation of H A yields

OHA/Opa = [-g -- 0 . 5 a ' p * -- (1 + h* -- ha)(2pa -- p*)]/'S

= [2[gk'(~,*)] 1/2 - 2(1 + a* - AA)[~k'(aA)]~/2]/~. (14)

Since h A < h*, the condition on p ( . ) guarantees alIA/OpA > 0 by L e m m a 3.

Thus, firm A cannot gain by adopting some strategy x A # (p*, r*, h*) such
that PA < P * ' Q.E.D.

Proof of Proposition 2. To prove the existence of a symmetric pure strategy

equilibrium, we have to show that OllA/OPA > 0 in (14). For k ( h ) = c)t ~ we
have O(A)= ( 1 - a ) / h , which is increasing in h. Thus, by L e m m a 3, it
suffices to show that p(h*) ~< - 2, i.e. that h* ~<0.5(a - 1). By Proposition
1, k ' ( h * ) = ach *'~-1 = p * / ( 4 + 2A*) = 0.255/(1 + 0.5}L*)2, SO that s/c =
4 a A * ~ - l ( 1 + 0 . 5 A * ) 2. Therefore, we obtain h * < ~ 0 . 5 ( a - 1 ) for s/c suffi-
ciently small.
For k(h) = - c In(1 - A) we obtain OIIA/OpA > 0 in (14) if H(A A, h*) ------(1 -
A*) - ~ - ( 1 + h * - h a ) z ( 1 - - h a ) - 1 > 0 for all h a < h * . The function H ( . ) is
concave in AA with H(A*, h * ) = 0 and 0H(h*, h*)/OA A < 0 for A* > 0 . 5 .
T h e r e f o r e , H(AA, h*) > 0 for all h A < A* if H ( 0 , h*) > 0 and A* > 0.5. This is
the case for A * > 0 . 6 1 9 . By Proposition 1, k ' ( h * ) = c / ( 1 - h * ) = p * / ( 4 +
2A*) = 0.255/(1 + 0.5A*) 2 so that ~/c = 4(1 + 0.5h*)2/(1 - h*). Therefore, we
obtain A* > 0 . 6 1 9 for ~/c sufficiently large. Q . E . D .
H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242 241

Proof of Proposition 3. Equilibrium profits are given by

=p* - A*p*Z/43 - k(A*) = 4k'(A*)(1 + 0.5A*) - A*k'(A*) - k(A*)
= 4k'(A*) + A*k'(A*) - k(A*). (15)

B y the convexity of k ( . ), //~ is increasing in A*. A s a result, 3 and //~ are

positively related b e c a u s e A* is an increasing function of ~.
C o n s u m e r surplus equals

C = 2v - 2p* + A*p*2/4s = 2v - 8k'(A*)(1 + 0.5A*) + A*k'(A*)

= 2v - 8k'(A*) - 3A*k'(A*). (16)

Since an increase in 3 raises A*, this p r o v e s that c o n s u m e r surplus is

d e c r e a s i n g in 3.
Social welfare equals W = 2//~ + C = 2v - 2k(A*) - A*k'(A*). A n increase
in 3 leads to a higher value of A* so that the social welfare is
reduced. Q.E.D.

Proof of Proposition 4. A n increase in c raises p * and lowers A*. T h e r e f o r e ,

e a c h c o n s u m e r ' s equilibrium utility will decrease with c and so c o n s u m e r
surplus is a decreasing function of c.
F o r k ( h ) = cA ~ we have Ak'(A)/a = k(Z). T h e r e f o r e

//~ = p * - A*p*2/43 - A*k'(A*)/a = p * [ 1 - (1 + a)A*p*/(43a)]. (17)

B u t A ' p * = 2(-s - p * ) . Since an increase in k ' ( A ) raises p * , A ' p * falls w h e n c

is increased. This p r o v e s that H,. and c are positively related for k(A) = cA r.
B y the p r o o f of Proposition 3, social welfare equals

W = 2v - 2A*k'(A*)/~ - A*k'(A*) = 2v - ( a + 2)A*p*2/(4as)

= 2v - (or + 2 ) ~ -p*)p*/(2o~). (18)

Since 2s/3 ~<p* ~<3, W is increasing in p * . T h e r e f o r e , an increase in c will

raise W if k ( A ) = cA ~.
N o w consider k(A) = c0(A) = - c In(1 - A). By the p r o o f of P r o p o s i t i o n 3
we h a v e
OH/Oc = (4 + A*)r/'(A*) -.0(A*) + c(4 + A*)~"(A*)0A*/0c;
OW/Oc = -2.0(A*) - A*n'(A*) - c[3,O'(A*) + A*.0"(A*)]0A*/Oc.
B y P r o p o s i t i o n 1, c . 0 ' ( A * ) = 3 / ( 2 + A*) 2 so that

c a A * / a c = - ( 2 + A*).0'(A*)/[.0"(A*)(2 + A*) + 2.0'(3.*)]. (20)

C o m b i n i n g these equations yields

242 H. Bester, E. Petrakis / Int. J. Ind. Organ. 14 (1996) 227-242

OH (8 + 2A*)'~'(A*) 2 8 + 2A*
0--7 - ff'(a*)(2 + x*) + 2 ~ ' ( a * ) - n ( a * ) - 4 - a* + In(1 - a*);
OW (6 + h*)~'(h*) 2 6 + a*
0~- - f f ' ( a * ) ( 2 + ,X*) + 2 r t ' ( h * ) - 2 n ( a * ) - 4 - A* + 2 In(1 - a * ) .

N o t e that aIl/Oc < 0 and OW/Oc< 0 f o r h* close e n o u g h to unity. By the

p r o o f o f Proposition 2, ~ / c = 4 ( l + 0 . 5 A * ) z / ( 1 - a *) and so a * - - + l as 3/
c - - + ~ . This proves that OPi/Oc<O and OW/be<O for ~/c large
enough. Q.E.D.


Anderson, S.P. and A. de Palma, 1988, Spatial price discrimination with heterogeneous
products, Review of Economic Studies 55, 573-592.
Anderson, S.P. and L. Leruth, 1993, Why firms may prefer not to price discriminate via mixed
bundling, International Journal of Industrial Organization 11, 49-61.
Bester, H., 1992, Bertrand equilibrium in a differentiated oligopoly, International Economic
Review 33, 433-448.
Bester, H. and E. Petrakis, 1992, Price competition and advertising in oligopoly, CentER
Discussion Paper No. 9222, Tilburg University.
Butters, G., 1977, Equilibrium distribution of prices and advertising, Review of Economic
Studies 44, 465-492.
Caminal, R. and C. Matutes, 1990, Endogenous switching costs in a duopoly model,
International Journal of Industrial Organization 8, 353-373.
Cremer, J., 1984, On the economics of repeat buying, Rand Journal of Economics 15,396-403.
Dasgupta, P. and E. Maskin, 1986, The existence of equilibrium in discontinuous economic
games, I: Theory, Review of Economic Studies 53, 1-26.
D'Aspremont, C., J.J. Gabszewicz and J.-F. Thisse, 1979, On Hotelling's stability in
competition, Econometrica 47, 1145-1150.
Gerstner, E. and J. Hess, 1991, A theory of channel price promotions, American Economic
Review 81, 872-886.
Grossman, G. and C. Shapiro, 1984, Informative advertising with differentiated products,
Review of Economic Studies 51, 63-81.
Hume, S., 1993, Coupons set record, but pace slows, Advertising Age 64, 25.
Koselka, R., 1990, How to print money, Forbes 146, 118-120.
Levy, D. and D. Gerlowski, 1991, Competition, advertising and meeting competition clauses,
Economics Letters 37, 217-221.
Narasimhan, C., 1984, A price discrimination theory of coupons, Marketing Science 3,
Peters, M., 1984, Restrictions on price advertising, Journal of Political Economy 92, 472-485.
Shilony, Y., 1977, Mixed pricing in oligopoly, Journal of Economic Theory 14, 373-388.
Thisse, J.-F. and X. Vives, 1988, On the strategic choice of spatial price policy, American
Economic Review 78, 122-137.
Zeidis, N., 1987, Premiums and promotions: Targeted coupons hit non-users, Advertising Age
58, 26.