You are on page 1of 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/24103041

Competitive Promotional Strategies

Article  in  The Journal of Business · February 1988


DOI: 10.1086/296442 · Source: RePEc

CITATIONS READS

395 7,731

1 author:

Chakravarthi Narasimhan
Washington University in St. Louis
73 PUBLICATIONS   3,907 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Retargeting, competitive impliations of fairness, MSRP and maximum selling retail price View project

Private labels View project

All content following this page was uploaded by Chakravarthi Narasimhan on 24 March 2014.

The user has requested enhancement of the downloaded file.


Chakravarthi Narasimhan
John M. Olin School of Business, Washington University

Competitive Promotional
Strategies*

Introduction This article analyzes


the equilibrium-pricing
Marketersuse a variety of methods to promote strategiesof brandsen-
their products. Advertising, trade promotions, gaged in a pricing
and consumer promotions are three primary game. Each brandhas
tools that firms use in formulatingtheir market- a monopoly market
(loyal consumers)and
ing strategieswhile competingfor a share of the competes with others
consumer's dollar.' A wide variety of price pro- in a common market
motions are used to provide a lower price to the called brandswitchers.
final consumers. These take the form of price In the case of a duop-
deals, cents-off labels, coupons (in newspapers, oly, it is shown that
the behaviorof the
magazines, mail, in/on pack), and a variety of brandswitchers char-
rebates. Attemptshave been made in the market- acterizes the equilib-
ing and economics literatureto explain the exis- riumbehaviorof the
tence of price promotions.In general,these mod- duopolists. This article
focuses on two key
comparativestatics,
. * I am especially grateful to Robert Blattberg, Charles namely, the depth of
Kahn,andRobertVishnyfor severalvaluablesuggestionsfor
improvingthis article.JohnQuelch,RamRao, JenniferRein- discounts away from a
ganum,V. Srinivasan,and TommyTan providedmanycom- "regular"high price
mentson earlierversionsof this article.The reviewersof this and the probabilityof
journalofferedmany suggestions,and I thankthem for that. giving a deal or the fre-
Responsibilityfor any remainingerrorsis solely mine. This quency of deals. How
articlewas formerlycirculatedunderthe title "Price Compe- these propertiesvary
tition with SwitchingConsumers." with the size of the
1. Trade promotionsare price discounts and other mer- loyal marketsand the
chandisingsupportgiven to the retailers. Consumerpromo-
tions are discounts and premiumsgiven directlyto consum- behaviorof the switch-
ers. According to latest figures available, manufacturers ing populationis dis-
spent 50 billion dollars in 1984 to promote their products, cussed.
spending36%on trade,38%in advertising,andthe remainder
on consumers. Over the last several years, promotionalex-
penditurehas been growingat an annualrate of 12%,while
advertisinghas been growingat a rateof 9%(Advertising Age
[August 15, 1985]).

(Journal of Business, 1988,vol. 61, no. 4)


? 1988by The Universityof Chicago.All rightsreserved.
0021-9398/88/6104-0001$01
.50
427

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
428 Journal of Business

els are set in a monopoly frameworkand use some form of price dis-
crimination.2The informationeconomics literatureexplainsprice vari-
ation in competitive marketsas the equilibriumoutcome of interaction
between oligopolistic firms and consumers who must search for low
prices.3
In this article I propose another explanation for the variation in
prices in competitive markets that is based on the consumer's brand
loyalty. Marketingresearchers(see Blattbergand Sen 1974, 1976)have
developed segmentationschemes that enable us to classify households
into segments using purchasebehavioralvariablessuch as loyalty to a
brand, price sensitivity, and response to promotions.4The purpose of
this article is to derive the equilibrium-pricing
strategiesof rivals com-
peting in a marketcharacterizedby segmentsthat differin theirloyalty
to a product.
The central hypothesis of this article is that firms fluctuate their
prices to induce brand switchers to buy their products while at the
same time minimizingthe loss of profitsfrom their loyal consumers. I
consider a market made up of consumers who are different in their
purchasing behavior. I hypothesize the existence of consumer seg-
ments that are extremelyloyal to a specific brandand a segmentthat is
willing to switch aroundto find a bargain.I show that the behaviorof
these switchers characterizes the equilibriumstrategies. Further, I
show that the equilibriumstrategies are mixed strategies with firms
choosing prices according to a distributionfunction defined over a
rangeof prices. I interpretdeviationof the price away froma maximum
price as a "promotion."While in cases where both firmstend to charge
the high price with positive probabilitythis poses no problem,this may
be problematicwhen one firmdoes not charge any price with positive
probability.However, several earlierstudies (see Varian1980;Gal-Or
1982;and Sobel 1984)have interpretedsuch equilibriumstrategiesas
occurrences of "sales," and I use the same interpretationin the spirit
of these studies. The two key comparativestatics of interest are the
average depth of discounts and the frequency of promotions. How
these will change dependingon the sizes of the loyal segmentsand the
behaviorof the switchingpopulationis the centralfocus of this article.
In the next section, I develop a simplemodel of a duopolyand derive
the pricing strategies. I then extend the basic model by changingsev-
eral assumptionsto illustratethe impactof the switcherson the equilib-
2. Blattberget al. (1981), Spatt (1981), Narasimhan(1984, 1988), and Jeulandand
Narasimhan(1985)are examplesof modelsthatexplainpricevariationacrossconsumers
or over time.
3. There is a vast amountof literaturein this area. They all rely on some form of
heterogeneityacross consumers(preferences,searchcosts) to explainthe pricevariation
across sellers. Examplesare Butters(1977),Salopand Stiglitz(1977),andVarian(1980).
4. Readersmightwonderwhy loyalty and price sensitivityare both mentionedsince
they are two sides of the same coin. The distinctionbetween these is madeto separate
brandchoice from the quantitydemanded.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
PromotionalStrategies 429

rium strategies. I outline managerialand empirical implications and


then conclude the article with directionsfor furtherresearch.

The Model
In this section I describe the basic assumptionsunderlyingthe model.
Then I introduceseveral plausibleassumptionsconcerningthe behav-
ior of the switching segment. Following this, I derive the equilibrium
strategies of the rivals under each one of these scenarios. Finally, I
consider several extensions and outline the equilibriumstrategies.
Assumptions
1. Thereare two firmsmarketingone brandedproducteach. The firms
face identical technologies, and the technology is characterizedby a
constant marginalcost (average cost is constant or declining every-
where) assumed to be zero withoutany loss of generality.The firmsdo
not incur any other cost for marketingtheir product. Firms are risk
neutraland act as expected profitmaximizersand behave noncoopera-
tively, taking the rivals' strategy into account.
2. All consumersbuy only one unit of a brandas long as the brand's
price is less than or equal to r. This r is the same for both brands.Later
on we will consider a case when the reservationprices are different.
Consumersbuy from only one firm in a given period.
3. The consumerscan be groupedinto the followingthree segments.
A segment of size a.1(O< a( < 1) is loyal to firm 1 and buys only from
that firm.A segmentof size (2(0 < Ot2 < a1) buys only from firm2. The
remainingconsumers IP(O< = 1 - - (X2)switch amongbrands
buyingaccordingto rules to be specifiedbelow.5The marketsize-that
is, the total number of consumers in the market-is a constant and
normalizedto one.
4. There are no intertemporaleffects either in terms of quantity
boughtor in terms of changingloyalties due to price changes. Further,
price changes can be made costlessly, and consumersbecome awareof
the changes immediately.
The above assumptionscapture the essence of a marketcharacter-
ized by a variety model (a'la Hotelling [1929]) in which consumers
preferan ideal variety and, for productsaway from this most preferred
bundle, their willingness to pay decreases. The loyal consumers are
characterizedby an extreme dissatisfactionfor any bundleaway from
theirpreferredbundle. If w(Q) is the willingness-to-payfunctionwhere
Q is an attributeon which the products are differentiated,then the (xi
segment is characterized by wi(Qi) = r, and wi(Qj) < MC for ij = 1, 2,
5. Blattbergand Sen (1974)estimate that loyal and switchingbehaviorsdescribethe
purchasingstrategiesof 69.6%,71.9%,64.6%,66.4%,and 78%of consumersin alumi-
num foil, wax paper, detergents,facial tissues, and headacheremedies categories,re-
spectively.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
430 Journal of Business

i # j, where MC is the marginalcost of the product.The 3 segment, in


contrast, switches between the brands, implying that their brand
choice decision is dependenton prices. There are several ways we can
model the behavior of the IPsegment:
1. A randomlychosen memberof the IPsegmentbuys brand1 if PI ?
P2 - d and buys brand2 if PI > P2 - d, where d is a consumerspecific
parameter,and Pi is the price of brandi. Note that if d is positive, the
consumer will buy brand2 for equal prices and, conversely, will buy
brand 1 when prices are equal if d is negative. The parameterd is the
monetaryvalue of inducementthat brand1 has to provideto overcome
the resistance of a buyer to switch to brand 1. If we assume that d is
distributedover the populationaccordingto a smooth continuousden-
sity function defined over an interval (- b, a), then, as shown by De
Palmaet al. (1985),there is a pure strategyequilibriumwith each brand
selecting a price that is the best response to its rival's strategy.
2. The switchers are not heterogeneous in their preferencefor the
brandsas in the above case. Instead we can assume that all switchers
are identical in their switching behavior and all possess the same d
value. Thus, at equal prices all the switchers prefer brand 2 if d is
positive or brand 1 when d is negative.
3. We can assume a variantof the above by assumingthat d equals
zero for all. This implies that the switchers are extremelyprice sensi-
tive and buy whichever brandhas the lowest price. That is, even for a
small price,difference, the entire i segment will switch brands.
When the switchers' behavior can be approximatedby either as-
sumption2 or assumption3 above, it turnsout that there is no equilib-
rium in pure strategies, but a mixed strategy equilibriumexists. I will
interpretthe realizationof prices from such a scheme as a promotional
mechanism used by the brands in competing for the switchers. For
pedagogical reasons I first consider the switchers' behavior as de-
scribed by assumption3 above and develop the equilibriumstrategies.
(Thisbasic modelhas some similaritiesto Narasimhanand Rao [1985].)
I will then extend the model to cover assumption2 above and simply
present the results. I will also discuss other extensions.
Switchers Are Completely Indifferent between the Brands
If the ith firm(i = 1, 2) were to pick a price Pi, and the jth firm(j # i)
were to pick Pj, then we can write down the profitfunctionfor the ith
firmas
i- 1, 2, i $j,
1 if Pi < Pi )
J1i(Pi,D\) = Oti + jpp
QD f
Il1(Pi, PJ)
= oi~i + =, aii 1/2 if Pi = PF,
0 if Pi > Pj.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
PromotionalStrategies 431

A firmpicks its price takingthe other firm'sprice as given. Ourfocus


is to characterizethe equilibrium-pricingstrategies of the rivals, and
the solution concept is the familiarBertrand-Nashequilibrium.
To illustratethe natureof the solution, assume ot, = 0.5, Ox2 = 0.2,
= 0.3, r = $5, and marginal cost is zero. Firm 1 can guarantee itself a
r) $2.5, and firm 2 a profit of at least $1.0 regardless of what
profit of (aOL
the other firm does. However, due to the nature of the competitive
forces, the lure of makinggreaterprofitsmotivatesa firmto price lower
so as to get the P segmentand increaseits profits.This does not pay off
either since the rival is also thinking the same way. For example,
suppose the first firmthinks of pricingat $3 to get the 3 segment, then
firm 2 can immediatelycounter with a strategy of undercutting.What
quicklyemergesis that no singleprice paircan be an equilibriumpairin
the sense of Nash. It turns out that in the stylized frameworkdevel-
oped here, a mixed-strategyequilibriumexists. A firmsimplyhas to be
unpredictableto have a strategy that is sustainableagainst its oppo-
nent. I develop a numberof propositionsto characterizethe solution.
PROPOSITION 1. There is no Nash equilibriumin pure strategies.
Proof. See Appendix A.
The intuitionbehind the above propositionis the following. If a firm
fixes a price, then its rival can always undercutthe price and capture
the entire 3 segment. Neither the monopoly solution (r, r) nor the
competitive outcome (0, 0) is an equilibriumpair since in the former
case the deviant firm can undercutits rival and, in the latter case, the
deviant firm can raise its price to r and make strictly positive profits.
Having shown that there is no equilibriumin pure strategies, I pro-
ceed to show the existence of an equilibriumin mixed strategies.Recall
that a mixed strategyequilibriumis definedas the pair ([4* SV)for i =
1, 2, where ,i is a probabilitymeasuredefinedover the set SVofstrate-
gies such that, given (4; Sr), the best response (maximizing expected
profit)for firmi is to use ([4 SV).The objective now is to characterize
([4, SV) and explore their properties. In the context of the pricing
problemwe have, SV= {PilPi is a price that the ith firmwould like to
chargewith positive density in equilibrium}.The profitsfor the ith firm
when it charges a price Pi and when its competitorcharges a price Pj
are
Lli(Pi) = %iPi + prob[Pj > PJ] Pi + prob[Pj = P] F
Pi, (2)

where Pie S, i &j, and i,j = 1, 2.


I can now formally state the two-player game as
max E(111i)= f11i(Pi) dF-F(Pi),
Fi

Pi E S*,

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
432 Journal of Business

such that
I :- otir,

fdFi(Pi) = 1,

Pi E S

where Fj(P1) is the cumulative distribution function for firm i, and E(7Il)
is the expected profits.
Returning to the numerical example discussed earlier, notice that it
does not pay for firm 1 to price below $3.12 even if it is assured of
getting the entire P segment since the profits it would get (0.8 x 3.12)
are less than what it can obtain from its protected market. Given that
firm 1 would never want to price below $3.12, firm 2 may never want to
be below $3.12 either since, by raising its price and coming arbitrarily
close to $3.12, it can retain the P segment and increase its profits by
charging a higher price. Thus, both firms would want to price in the
$3.12-$5.00 price range, and the equilibrium solution is to develop a
strategy of choosing prices in this range. Notice further tht the small
share brand can guarantee a profit of $1.56 by pricing slightly below
$3.12 since, in equilibrium, the higher loyal share firm will never price
below $3.12. It is interesting to note that this amount ($1.56) is strictly
greater than the $1.00 that the second firm obtains from its own loyal
market. Thus, price competition over the switchers who are indifferent
between the two brands seems to help the smaller-share brand.
The following propositions, which are proved in Appendix A, further
characterize the nature of the solution. I will assume without loss of
generality that a > 2
PROPOSITION 2. The strategy sets SVand SV are convex. That is,
there are no holes.
PROPOSITION 3. Neither firm can have a mass point in the interior
or at the lower boundary (P) of the other's support, nor can either
firm have a mass point at the upper boundary (P) of the other's sup-
port if that boundary is a mass point for the other firm.
PROPOSITION 4. The strategy sets SV and SV are identical when
neither firm has a mass point. If firm j has a mass point at Pj, then
firm i will charge Pj with zero density in equilibrium. That is, firm i will
randomize in the half open interval [P, Pj) when firmj has a mass point
at Pj. A
Remark. This proposition also establishes that P = sup(S*) =
sup(S*j) and P = inf(S*') = inf(SP). The next proposition charac-
terizes P.
PROPOSITION5. sup(S*) = sup(S) = sup(S) = P = r.
From propositions 1-5, I can state the following. The equilibrium
strategies are mixed strategies with each firm selecting a distribution

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
PromotionalStrategies 433

defined over a convex set (a range of prices). Except possibly at r,


there cannot be mass points anywhere else. The probabilitythat both
the firmswill have mass point at r is zero. The strategysets of the two
fikmswill be identicalif there are no mass points and will not includer
for one firmif the other firmis to have a mass point at r. Thus, if one
firmwere to have a mass point at r, its strategyset is the closed interval
[P, r], and the strategy set for its rival is the half open interval [P, r).
Further,the firmwith the smallerloyal sharecan obtainprofitsthat are
greaterthanthe profitsit would obtainfrom its own loyal segment.The
existence of a mixed strategy equilibriumfor this game can be estab-
lished by applyingtheorem 5 of Dasguptaand Maskin(1986). I show
the existence of an equilibriumby construction.If Fi(P) is the equilib-
riumdistributionfunction for firm i, then

max E(H i) = {r {iP + [1 - Fj(P)]PP}dFi(P).


Fi(P)

In order to derive the distributionfunctions Fi's, we must resolve


the issue of mass points and specify what P is. Assume that both
distributionshave positive density at r in equilibrium(i.e., there are no
mass points). This implies that firm i has equilibriumprofits equal to
otir, and therefore Fi should have support over the interval [Pi, r].
However, since PI (= ot1r/(Ito+ Is))> P2 (= ot2r/(ot2 + Is)),the strat-
egy sets are not identical, contradictingproposition4. Therefore, in
equilibriumone firmmust have a mass point at r, and the other should
have zero density there. Suppose that firm2 has a mass point at r. By
proposition3, the profits to firm 2 when it charges r equals ot2r since,
with probabilityone, firm 1 will be priced lower than firm2. However,
firm2 can do strictlybetter than this since it can price with probability
one at PI and obtainthe entire IPsegmentfor a total profitof (t2 + W)P1
> oc2r.This implies that in equilibriumfirm2 cannot have a mass point
at r. Firm 1 therefore has a mass point at r, and, further,P = PI =
+ Is).The profitthat firm1 will obtainin equilibriumtherefore
(otLr)/(otL
equals ot1r since I have just argued that firm 2 will be below r with
probabilityone. The equilibriumprofit that firm 2 obtains must equal
its best alternative,namely, (t2 + 13)PI.
I can write the equilibriumconditionsfor P E S* except at r. These
are
o1P + [1 - F2(P)]PP = otr, otr/(ot, + @)? P < r, (3)
and
t2P + [1 - F1(P)]P = (t2 + A citr
iI( + (4)
otir1(ot1+ ~)?P < r.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
434 Journal of Business

From (3),
F (P)
2 I -o1 l(r- P) (5)
~~~~PP
From (5), we see that F2(P) = 0, F2(r) = 1. Further, note that F2(P)
approachesone as P approachesr from below, suggestingthat F2 has
no mass point and has a continuousdensity function. Given this obser-
vation, and since the probabilityof ties occur on a measureof zero, I
can state that (4) must hold at r as well. Solving for F1(P) from (4), we
get

F1(P) = I + '2 _ r(+13) PcP r. (6)

From (5) and (6) we see that F1 (P) = 12(P) = 0, F2(r) = 1, and F1 (r) =
1 - (clI - U-2)/(Ot + IP).This implies that F1(P) will have a mass point
at r equal to (a1 - U2)/(Ul + I3)
To summarize,when the switchingpopulationis indifferentbetween
buying the two brands, the firms adopt mixed strategiesgiven by the
following distributionfunctions:
O. P < otir/(oti + 13),
F1 = 1 O2/P - otlr(ot2 + P)/IP(Otl + Isr), oLr/(otL + I3)< P < r,

and
0. P < oLIr/(otL+ 13),
F2 = 1- oL(r - P)IPP, otLr/(otL+ I) < P < r,
1, p:r.
Note furtherthat the equilibriumis unique. Thatis, there is no other
pair of strategies-that is, a probabilitymeasure-strategyset combina-
tion that constitutesan equilibrium.This is verifiedby the construction
of the mixed-strategyequilibriumin (5) and (6). To show that (5) and (6)
do indeed characterizea Nash equilibrium,defineF and F2 to be two
arbitrarydistributionfunctionsfor firms 1 and 2, respectively, with the
provisionthat F' and F2 have positive supporteverywherein the inter-
val (PA,r) (by proposition 2). Define the strategies representedby (5)
and (6) as F* and F*, respectively. Then

E(H(FF2)) 2 = {{otP + [1 - F*(P)]PP}dF'(P)

= Clr JdF (P)

aE(nd(F*, F*)),
and

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
PromotionalStrategies 435

E(J2(F2, F*)) = { {oL2P+ [1 - F(P)]13P}dF2(P)

- (t2 + P3)otlr [dFr(P)


(ai + IP)
cE(112(F*, F*2)).
The last inequalityfollows since F2 cannot have a mass point at r by
proposition 3. This proves that F* and F* are best responses to one
anotherin the sense of Nash.
The equilibriumdistributionfunctions are characterizedby (5) and
(6), and from these we can state the following:
i) F1(P) < F2(P) for P E S*;
ii) E(P1) = f[I - A> f[1 - F2(P)]dP - P = E(P2)
- F1(P)]dP
(from [i]);
iii) E(P11P1 < r) = E(P21P2 < r) = (otjr/I) ln[(ao + P)/otl];
iv) the mass point w = (L1 - 02)/(Oli + O) is increasing in (oi - t2);
and
v) E(HI1) = otir and E(H2) = (t2 + 13)otr/(oIc + 1) > U2r.
The above propertieshave the following qualitativeimplications:(i)
states thatP1 is first-degreestochastic dominantover P2, (ii) states that
on the averagethe brandwith the largerloyal franchise(i.e., brand1)is
pricedhigher, (iii) states that the expected value of the price given that
a brand is priced lower than r is the same for both brands, and (iv)
states that the probabilitythat the firmwith the largerloyal franchise
will charge r is higher, the greater the difference is in the size of the
loyal segments. Note that in terms of marketingstrategiesthis implies
that the brand with a larger loyal share should change its price less
often and offer the same expected discount.
The intuition behind (ii) and (iii) is the following. The opportunity
cost of charging a price P < r is higher for firm 1 since it loses a, (r - P)
from its loyal consumers, which is greater than a2(r - P) that the
second firmloses when it chargesa lower price P. Since the incremen-
tal sale is the same (the P segment), firm 2 is simply more aggressive
since it has less to lose. This explains (ii). Given the symmetryin their
pullingpower of the switcher segment, both firmsoffer the same aver-
age discount accordingto (iii).
Another feature of this equilibriumis that the firm with the larger
loyal franchiseearns the same profitin equilibriumthat it would earn if
it were to set Pi = r with probabilityone, ignoringthe other firm.6By
contrast,the firmwith the smallerloyal franchiseearnsOLir < (Oa2+ 0)/
(a0 + P) > at2r,where at2r is the profitit would makefromits own loyal

6. Since firmsare assumed to be risk neutral,a certaintyequivalentis not preferred


over a risky prospect.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
436 Journalof Business

segment. It is instructive to compare this profit with the profit that


would result if the firms were to collude and set Pi = Pj = r. In that
case, firm 2 will earn (a2 + f2)r. When will the profits obtained under
this equilibrium be larger than the collusive profits? This can be charac-
terized by the following inequality
Ot(a( r(Ot 12) +
2 + r as 2- (t2 /2)

This last inequality can be rewritten to state the condition when the
profits by playing this game will be larger than the collusive profits for
firm two. This yields ,B < (a, - 2a2). Since f3 is positive this also
implies that a2 < al/2. That is, firm 2's loyal segment is much smaller
than firm l's loyal segment. Firm 2 is willing to play the Nash game
since it obtains more than 50% of the switchers and stands to gain more
than by colluding when it obtains only 50% of the switchers. Raj (1985)
offers evidence that a brand with a smaller loyal segment tends to
obtain a larger share of the brand switchers than the brand with a larger
share of the loyal consumers. To summarize, if a, is "large enough,"
firm 2 will simply not agree to any collusive behavior since it stands to
gain if it plays the Nash game. This of course will hold only when side
payments are ruled out. The above result is consistent with the obser-
vation in the marketing literature that smaller share brands tend to
benefit more from promotions than larger share brands (see Guadagni
and Little 1983).
Another interesting feature of the equilibrium is that, as (a1, - a2)
0, the mass point disappears and the equilibrium is symmetric. Further,
as P-> 0, one gets the monopoly price as the pure equilibrium strategy,
and, as a,, a2 _>0, we get the familiar Bertrand-Nash outcome.

Switchers Are Not Indifferent


In this section I extend the basic model to explore the sensitivity of the
implications to changes in modeling the switching behavior. More
specifically, I will assume that, at equal prices, the switchers prefer one
of the two brands. Given sufficient inducement in the form of a price
cut, all will switch to the less preferred brand. I will consider two cases
here. First, when switchers prefer brand 1 (firm with a higher loyal
segment) and, second, when they prefer brand 2.
Case 1. In this case switchers prefer brand 1 at equal prices. I
describe the switchers' buying behavior as follows (d > 0):
buy brand 1 if P1 < P2 + d and PI r;
buy brand 2 if PI P2 + d and P2 r;
buy neither if P1 > r and P2 > r.
First note that (r, r) is a pure strategy equilibrium if (a2 + f3)(r - d)
< a2r. If the profits that will accrue to firm 2 when it lowers its price
to (r - d) and obtains all the switchers is less than what it will obtain

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
PromotionalStrategies 437

fromits own loyal segment, then firm2 has no incentiveto cut its price,
and consequentlyit serves its own loyal market.The above inequality
emphasizes that, if IPis too small or if d is too large, then there is no
incentivefor eitherfirmto deviate from r, and consequentlythereis no
need to offer promotions.I will thereforeassume that (cr2 + )(r - d)
> ct2r. I will also assume that d > PI - P2, where Pi = our/(oti + I).
These two conditionscan be combinedto restrictd as jr(ao - t2)/(tl
+ )(a2 + >) < d< jrI(a2 + ). Under this condition, it can be shown
that a mixed-strategyequilibriumexists. In equilibrium,first note that
firm2 will not have any supportover the interval(r - d, r). By pricing
in this interval,the firmdoes not obtainany switcherseven whenbrand
1 is pricedat r. Firm 1 has no incentive to-pricebelow (P2 + d) (this is
greaterthan P1) since firm2 will not price below P2. Thus, in equilib-
rium, firm 1 randomizes over the range (P2 r)A? d, and firm 2 ran-
-
domizes over the range (P2, r d, and r). In AppendixB I sketch the
equilibriumand provide the distributionfunctions. Comparedto the
results from the basic model, there are many differences. First, ob-
serve that there is a range(r - d, r) over which one of the firms(firm2)
has no support.Second, both firmshave mass points at r. Thatis, both
firms will charge r with positive probabilityimplyingthat, unlike the
earliercase, the probabilitythat the firmstie each other at r is nonzero.
This occurs in this model since, by slightlycuttingits price to below r,
firm2 does not obtain any incrementalsales when firm 1 is priced at r.
It is possible for us now to talk about a "regular"price, that is, r, and a
"promotional"price away from r since with positive probabilitywe
should find both firms priced at r. If we assume that a time series of
prices are realizations from the repeated play of such a game, it 'is
possible to talk about the frequencyand averagediscountof promotion
prices. In table 1, I illustratethe results and implicationsof this equilib-
rium. I make the comparisonbetween the relative magnitudesof the
mass points and the expected discount based on simulationexercises
using parametersover a wide range. While the relative magnitudes
of the mass points at r seemed to be reversed for small values of P
(< 20%),the relativemagnitudesof the discounts seemed to be robust.
Based on these results, I can say that, in equilibrium,the firmwith the
larger loyal share promotes more often and offers a lower average
discountthan the smallerbrand(see Quelch [1987]for some arguments
in support of this). In equilibrium,firm 1 earns higher profits than it
would obtainfrom its own loyal consumers,whereasfirm2 obtainsthe
same profits (a2r).
Case 2. In this case, switchers prefer brand 2 at equal prices. I
describe their behavior as follows:
buy brand 1 if P1 c P2 - d and P1 c r;
buy brand 2 if P1 > P2 - d and P2 c r;
buy neither brand if PI > r - d and P2 > r.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
438 Journal of Business

TABLE 1 Properties of the Equilibria When the Switchers Are Not Indifferent

Characteristic Switchers Prefer Brand 1 Switchers Prefer Brand 2*

Price range:
Brand I (P2 + d, r) (1,r-d, and r)
Brand 2 ("2, r - d, and r) (P1 + d, r)
Mass points:
Brand 1 a2dI[3(r - d)] ((X2 + OP1 + d)lpr + Qt2/P
Brand 2 (oLi + )(P2 + d)lpr - otj/I cqdl[P(r - d)]
Probability of
lower price prob [P1 < r] > prob [P2 < rnt prob [P1 < r] < prob [P2 < r]
firm with large share brand firm with large share brand
promotes more often promotes less often
Average discounts E[P1jIP < r] > E[P2 IP2 < r] E[P1 P' < r] < E[P2lP2 < r]
firm with large share brand firm with large share brand
offers smaller average dis- offers greater average dis-
count count
Maximum dis-
count firm 2 offers the largest dis- firm 1 offers the largest dis-
count count

NOTE.-Variables are identified in text.


* Subscripts 1 and 2 refer to large share (a,) and small share (a2) brands, respectively.
t This prediction was reversed for small values of P(< 20%) or for very small values of d within the
allowable range.

Similar to the discussions above, if (a., + 13)(r- d) < a1r, then (r, r)
is an equilibriumpair. Neither firmhas an incentive to cut its price. All
the switchers buy brand2, and brand 1 obtains only its loyal consum-
ers. I will therefore assume that (a1 + ,B)(r- d) > a, r, which reduces
to d < fPr/(a, + P). Further, it is obvious that P1 + d> P2, and
thereforefirm2 has no incentive to price below P1 + d since firm 1 will
not price below PI. Moreover,firm 1 will not have any supportover the
range(r - d, r) since by chargingany price over this rangeit does not
obtain any incrementalsales to the switcher segment. Therefore, in
equilibrium,firm 1 randomizesover the range(Pc,(r - d), and r), and
firm2 randomizesover the range(P1 + d, r). In AppendixB I providea
sketch of the equilibriumand derive the distributionfunctions. As in
case 1, once againwe findthat both firmshave mass points at the upper
end-namely, r. In table 1 the propertiesof the equilibriumare charac-
terized. Now we find that firm 1 has more mass at r than firm 2.
Further, the expected discount for firm 1 is higher than for firm 2.
These results were found to hold over a wide range of values for the
cx's, IP,and r. The implicationthen is that, when the switchers prefer
the lower loyal share brand at equal prices, the larger brand must
promote less often but offer larger average discount than the smaller
brand. In equilibrium,brand 2 obtains a larger profit than it would
obtainfrom its own loyal customers, as firm 1 obtains the same profits
(air).
To summarize,we see that differentpromotionalpatterns are pre-

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
Promotional Strategies 439

dicted for the rivals depending on the behavior of the switching seg-
ments. When switchers are not indifferentbetween the brands, we
obtaindifferentimplicationsthan when they are completelyindifferent
between the brands. Further, depending on the preferences of the
switchers, the promotional strategies are different. This once again
underscoresthe need for managersto evaluate the switchingbehavior
of brand switchers in designingthe promotionalstrategy.
Selective Price Cuts to Switchers
Assume that only a fraction Oiof the loyal segment buys at the lower
"sale" price, with the remainderpayingthe higherprice. This could be
due to deliberate actions taken by the firm to exclude giving promo-
tions to loyal buyers, such as issuing coupons (Narasimhan 1984).
Continueto assume that the entire f3segmentcan avail itself of a lower
price. The following outcomes are easy to verify.
i) If 01 = 02 = 0, firms compete over the ,Bsegment. This corre-
sponds to the classical Bertrandmodel, and we get the familiarresult
that there is a pure-strategy couponing equilibriumwith each firm
choosing coupon value of r (note that I have assumed a constant mar-
ginal cost and set i equal to zero). If there is a cost of $k per redeemed
coupon, then the value of coupon is (r - k). If there is a fixed cost for
couponing, then we get a mixed-strategyequilibrium.
ii) If 01and 02 are differentfromzero, and 0lt1 > 02t2, we get the same
results as in the basic model. By defininga4 = Oao and ot2 = O2t2,
we see that the results developed above continue to hold with cXire-
placed by at. That is, there is a mixed-strategycouponingequilibrium
with the firm with the larger loyal franchise giving coupons less fre-
quently and with a lower average discount than the smallerbrand.
iii) If 0ilt < 02c-2, again we obtain the same result as in the basic
model, but the roles of the two brandsare reversed. It is the brandwith
the largerloyal share that is able to protect a greaterproportionof its
segment from getting the price cut and thus acts as the aggressive
promoter.As far as coupon users are concerned, it is the largerbrand
that assumes the role of a brand with a smaller loyal market in the
above development. This implies that it is the firm with the smaller
loyal franchise that discounts less often.
Competition in a Differentiated Market
In this section, I discuss the equilibriumstrategieswhen the reserva-
tion prices of consumersfor the two brandsare different.Specifically,
assume that all consumersare willingto pay rl for the firstbrandand r2
(< rl) for the second brand. I thus assume that the two brands are
differentiatedwith the firstbrandcommandinga price premiumof 8 =
(rj - r2). Further,the restrictionthat al is largerthan Ot2is removed.
Thus, we can think of brand 1 as a premiumpricedhigh or a low share

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
440 Journal of Business

TABLE 2 Comparison of the Two Equilibria in a Differentiated Market

High Share PremiumBrand Low SharePremiumBrand


Characteristic (1 - a)ri < (1 - (X2)r2 (1 - ol)rl > (1 -2)r2
Price range:
Brand 1 (rbnPi) (r1,P2 + 8)
Brand 2 (r2, P -8) (r2, P2)
Mass points:
Brand I - 2)r2 -(1- )rl none
(Ot1 + f3r2
Brand2 none a2)r2
(a)rl
Probabilityof
lower price prob[Pi < r1] < prob[P2 < r2] prob[Pi < ri] > prob[P2 < r2]
firmwith premiumbrandpro- firmwith premiumbrandpro-
motes less often motes more often
Averagediscounts E[Pj P1 < r1] > E[P2 IP2 < r2] E[P IP1 < ri] < E[P2IP2 < r2]
firmwith premiumbrandof- firmwith premiumbrandof-
fers a largeraveragedis- fers a largeraveragedis-
count count
NOTE.-Variablesare identifiedin text. Subscripts1 and 2 referto premium-priced
and regular-
pricedbrands,respectively.

brand,and in the same way we can thinkof brand2 as a regularpriced


low or a high share brand. Once again it is easy to see that there is no
equilibriumin pure strategies. However, a mixed-strategyequilibrium
exists with propertiessimilarto the undifferentiatedcase. It turns out
that the propertiesof the equilibriumdepend on the magnitudesof the
followingparameters:01, a2, r1 and r2.Dependingon whether(1 - al)
is greaterthan or less than (1 - 2), the propertiesof the solutionsare
different. In Appendix C, I sketch the equilibriumfor the case when
(1 - (xl)rl is less than (1 - a2)r2, and in table 2, I present the properties
of the equilibria. (The implications on the average discounts are
difficultto verify analytically.However, they were foundto be true for
a wide range of values of the parameters.)Note that if a2 > al1, then
(1 - aol)r > (1 - a2)r2. This means that the last columnin table 2 can
correspondto a case where the largerbrand(O2) is not the premium-
priced brand. From the table, we see that the firmsrandomizeover a
differentrange of prices. When (1 - oal)rl < (1 - x2)r2,the premium
brand randomizes over (rl, PA),and firm 2 randomizes over (r2, PI - 5),
where Pi is the lowest price that firm i would ever charge (= airi1(a1
+ I)). The existence of the mass point also differs in the two cases.
When (1 - aol)rl < (1 - a2)r2, the premium-pricedlarge share
branddiscounts less often and offers a largeraverage discount. When
(1 - al)rl > (1 - a2)r2, then the premium priced brand offers a
higher average discount and promotes more often.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
Promotional Strategies 441

Implications
In the last section a model of duopoly was developed, and the pricing
strategiesof the rivals were illustrated.It was shown that the behavior
of the switchingsegmentimplieddifferentstrategies.In termsof strate-
gic prescriptions,I can offer the following guidelines.
1. If the switching population is completely indifferent between the
two brands, a large share brand should lower its price less frequently
than the lower share brandand offer the same average discount. The
followingobservationson the depth and frequencyof discounts can be
made.
i) The frequency of discounts by the largerbrand (= (0t2 + 1)(Oi
+ Is)) is an increasingfunction of the size of the smallerbrand'sloyal
marketand a decreasingfunction of the size of its own loyal market.
ii) The average discounts of both brandsare increasingin the reser-
vation price (r).
iii) Holding the switcher segment size constant, if ai increases (and
Ot2 decreases), the expected price of both brands increase, and the
average discounts of both brands consequently decrease.
2. If a firmcan protect its loyal marketfromprice cuts throughsome
form of price discrimination,then its ability to compete on price in-
creases. For example, in the above illustration,if the large sharebrand
can exclude its loyal marketfrom obtainingthe lower price while the
small brand cannot, then the large share brand should be the one to
offer lower prices more frequently.
3. When the switchers are not indifferentbetween the brands, the
strategiesare different.
i) If the inducement to make the switchers switch to the less pre-
ferredbrandis "large" or if the size of the ,3 segmentis "small," then
neither firm may want to promote its product, and we will observe a
constant price as the equilibriumstrategy.
ii) If switchers preferbrand 1 at equal prices, then the optimalstrat-
egy for firm 1 is to promote its product more often and offer a lower
average discount than brand 2.
iii) If switchers prefer brand 2 at equal prices, then the optimal
strategyfor firm 1 (firmwith the largershare)is to promoteits product
less often but offer largeraverage discounts than firm2.
4. If the switchers are willing to pay a price premiumto one of the
brands,then the strategiesare differentfrom the one describedabove.
A premium-pricedbrandin generalwill offer a higheraveragediscount
and also promote more often unless its share of the loyal segment is
very large (i.e., (1 - x1)rl< (1 - X2)r2).
5. In categories where there are many brands with intense rivalry,
the brandthat has the least amountof pullingpower may not want to

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
442 Journal of Business

discount at all if there are costs for lowering price. In this case the
brandmay want to keep a permanentlower price. Genericbrandsand
some store brandsin many grocery productcategoriesadopt this strat-
egy.
EmpiricalAnalysis
In testing the predictions of these models, an interesting question
arises, namely, how do we know firmsfollow the strategiesprescribed
here? The one indirecttest we can propose is to identifythe key com-
parativestatics-namely, the frequencyof price changes and the aver-
age value of the discount-and test the relationshipbetween these and
the magnitudes of the loyal segments and the switching behavior.
Other importantpoints are the following. In the theoretical develop-
ment, the firm sells directly to the users of the product. There is no
intermediarysuch as a retailer. However, in practice, price discounts
are often providedthroughthe retailer,and the retailer'srole in passing
on the discounts was not modeled in this paper.The retailersobjective
functionis not to maximizea particularbrand'sprofitbut to maximize
the overall profitof his store while competingwith otherretailersin his
geographical market. Therefore, unless the retailer's incentives are
aligned with the manufacturer,the observed discount patterns are
likely to be different from the predictions of the model. What this
implies is that it is desirablethat this aspect be modeledexplicitly, and
implicationsfrom a manufacturers-retailer model be derived prior to
empiricaltesting, or this aspect must be taken into account even with-
out explicit modeling. However, one can test the implicationsof this
model with discounts provided directly to the consumers such as
throughcents-off coupons. While I providedevidence on the positive
correlationbetween the average discount and the price of brandsin a
numberof categories examined (Narasimhan1984), I did not test the
implicationson share or the,frequency of discounts.

Summaryand Conclusions
In this article, a model was developed to explore the equilibrium-
pricing strategies of brandedproducts that enjoy a monopoly market
protected from other firms and a common marketin which every one
competes. It was shown that a mixed-strategyequilibriumexists with
the firmsrandomizingover an interval. Dependingon the behaviorof
the switching population, different strategies were derived for the ri-
vals. The two key statistics of interest are the probabilitythat a brand
would have a lower price than the reservationprice of the consumers
(frequencyof deals) and the expected value of the discount (depth of
discount). Empiricallytestable propositionswere derivedby assuming
that time series of observationon prices and sales are generatedby the

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
Promotional Strategies 443

players executing their optimalstrategiesevery period. Some caveats


on empiricaltesting were provided.
Several modelingfeatures that lead to the interpretationof the pric-
ing strategy given in this article are worth emphasizing. First, that
there is some subpopulationof brand switchers who behave homoge-
neously is important.For example, takingthe case when the switchers
are not indifferent,if the d parameteris distributedover the population
of the switchers accordingto a continuousdistribution,then there is a
pure-strategyequilibrium,and thereforethere is no need to changethe
prices. Second, for the randomizationstory to have any meaning, in-
tertemporaldynamics should have very little influence. Translatedto
the context in which this model is set, specifically,the switchersshould
remainas switchers. They should not develop any lasting loyalty to a
brand.Or, even if some of them do, there shouldbe an infusionof new
switchers. By the same token, the loyal consumersshouldnot develop
any "negative" attitude toward a brandjust because its price is re-
duced often.8This may be somewhatmitigatedby targetingadvertising
towardloyal consumers, emphasizingthe qualityof the productrather
than its availabilityon promotions.
There are several key directions for future research. An important
omission of this article is the absence of a model describingthe role of
the retailerand his competitiveenvironment(see the empiricalanalysis
section earlier).The retailer'spolicy would reflectajoint maximization
of the category profitsand thereforecan be differentfrom the policies
espoused here. Further,a retailercompetes with other retailerswithin
a geographicalmarket for a share of consumers' dollars. Possessed
with such knowledge, the manufacturersin turnwill alter their behav-
ior. What the outcome would be of such strategic interactions is a
question that future research should answer.
Next, modeling more than two firms seems to be desirable, given
that often there are more than two brandscompetingin the same sub-
market.Anotherimportantextension would be to introducea dynamic
element in the demand function. When many new products are in-
troduced, firmshave to informpotential consumersof their very exis-
tence and also have to induce consumers to try their products so that
each firmcan develop a groupof loyal consumersover time. Modeling

7. Thereis an obvious dangerin extendingthe equilibriumdevelopedin this articleto


marketswhere the game is played for many periods. The equilibriumsolution from a
static modelwill representone of the manypossible equilibriathatexist in a multiperiod
context. For example, over time firms may reach a collusive outcome by developing
appropriatestrategies, such as punishinga rival for low pricing,etc. I ignorethese as-
pects here.
8. Thereis a perceptionamongmarketersthatconsumersconsidera brand"cheap"if
its price is reducedtoo often. For some items that are mainlyboughtfor conspicuous
consumption,this is probably true. How far this influences behavior in nondurable
purchasesis not known.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
"44 Journal of Business

this aspect will have important implications on the promotional strate-


gies for new (vs. established) products. Further, how advertising,
which serves to inform potential consumers of the existence of the
product, should be used in conjunction with the pricing strategy will be
of immense interest to marketers. For example, whether a new brand
(small share) should try to penetrate a market using advertising or price
promotions is an important question that managers worry about.

Appendix A
Proof of Propositions 1-5
1.
PROPOSITION There is no Nash equilibriumin pure strategies.
Proof. Suppose (Pt, P*) is an equilibriumpairof Nash strategies.Then, by
definition,there is no such Pi (i = 1, 2) such that I1(Pi,Pj*)> Hf(P*,Pj). The
proof of the propositionproceeds by provingthe existence of such a Pi.
Case (i).

P*= Py,
Hii(P*,PJ) = otiP* + 1/2 UP* (Al)
Let Pi = P* - E, E > 0. Then
Hli(P-, Pj) = oti(P* - E) + I(P -E). (A2)
From (Al) and (A2),

HII(Pi,Pr) > Hi(P, Pj) if E 2(< + mu)

Since P > 0, such an E always exists. Q.E.D.


Case (ii). Withoutloss of generality,let P* < PJ;

fli(PTE PJ ) = otiPi + UP S.

Choose Pi = PE + E< PJ

His(Pi, Pj) = Oti(Pi + E) + (PE + E).

Then
fli(Piq PJ) > fli(P*, PI). (A3)
The proof of the existence of such an E so that (A3) holds is trivial.
2. The strategy sets SVand Spare convex. That is, there are no
PROPOSITION
holes.
Proof. The proof proceeds by first showingthat there are no holes in T =
sV n Sand then showing that there are no holes in T' = V - SVfSn
No Holes in T. Let P = inf(T) and P = sup(T). To show that T is convex,
we show that there are no "holes"^inT. That is, there is no intervalI = (pk,
ph) such that, for p <Pk <ph <j5 and for P E I, P C T. This could happen
when one of the firmshas supportover the intervalI andthe otherone does not

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
PromotionalStrategies 445

or when neither firm has support over the intervalI. I show that neither of
these two is possible.
Firstnote that if the ith firmchargesP E I with probabilityzero, then so does
thejth firm. To see this, let P1 and p2 be definedas
P1 E S and P1 = sup{P IP < pk},

p2 ES and p2 = inf{PIP> Ph}.

Since the profitsfor thejth firm,when it chargesPj = {ctjPj + [1 - Fi(Pj)] PrPj},


are increasingin Pj for Pj E I, firmj is better off chargingp2 with probability
[Fj(p2) - Fj(P)] and has no mass over the set I.
Now consider the case that neitherfirmis randomizingover the set I. The
profitto the jth firmwhen it chargesP1 is
ajP1 + [1 - Fi(P)]rP1.
Next, consider the profits that would accrue if the jth firm chargesph This
equals
(pjph + [1 - Fi(Ph)] hPh.

But, since Fi(Ph) = F1(Pk) = F1(P1), the profits obtainedby chargingph are
strictly greaterthan the profitsobtainedby chargingF1, contradictingthe as-
sumptionof an equilibrium.
No Holes in T'. Once again define P = inf(T') and P = sup(T'). Note that
T' correspondsto the set of prices chargedby i and not by j, and, by virtueof
the above proof, eitherP < inf(Sj) or P > sup(Sj). Note furtherthat T' cannot
containany holes. If it did, firmi could strictlymakeitself betteroff by moving
the mass from the lower end of the hole to the upperend since by doing so it
does not lose any sales but charges a higherprice to its buyers.
PROPOSITION 3. Neither firmcan have a mass point in the interioror at the
lower boundaryof the other's support,nor can eitherfirmhave a mass point at
the upperboundaryof other's supportif that boundaryis a mass point for the
other firm.
Proof. Let Pi = inf(SV)and Pi = sup(S'). We note that Pi > 0 since the
firmshouldmake positive profitsin equilibrium.Assume to t~hecontraryof the
propositionthat thejth firmchargesa price P*, Pi < P* < Pi with probability
w. The proof proceeds by showingthat firmi can increaseits profitsby chang-
ing its strategy.
Fromproposition2, we know thatthereare no "holes" in the strategyset for
firm i. Consider the profits of the ith firm when it charges (P* - E) and
(P* + E), E > 0. These are given respectively by
- E) + [1 - -
(A4)
oi(P* Fj(P* -E)]rP(P* E),

and
Ui(P* + E) + [1 - Fj(P* + E)]P(P* + E). (A5)
Subtract(A5) from (A4) to yield
-2E(cUi + p) + EP[Fj(P* + E) -Fj(P* - E)] (A6)
+ PrP*[Fj(P* + E) - Fj(P* - E)] -2E(oXi + p) + PrW + PP*W.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
446 Journal of Business

For a small enough e > 0, this is strictly positive, suggestingthat firm i, by


shiftingsome mass to the left of P* from the rightof P*, can be made strictly
better off, contradictingthe equilibrium.
Whathappenswhen P* = Pi and firmi has a mass point at Pi equalto wi?In
this case also, firmi can do better by charging(Pi - e) with probabilitywi and
chargingPi with zero density. Thus the proof holds even when the other firm
has a mass pointat its upperboundary.The only time the proofdoes not hold is
when P* = Pi and firmi does not have a mass point at its upperboundary.
PROPOSITION 4. The strategy sets S7 and V are identicalwhen neitherfirm
has a mass point. If firmj has a mass point at P, then firmi will chargeP with
zero density in equilibrium.That is, firm i will randomizein the half open
interval[PAP) when firmj has a mass point at P.
Proof. Considerthe case of no mass points. Assume to the contrarythat
(without loss of generality)SVC SJ;This implies that an intervalexists with
firmi havingno supportover the interval,but firm does. Further,by proposi-
tion 2, this intervalwhere firmi has no supportis either at the lower or at the
upperend. That is, for Pj E I = S - SJnfS* eitherPj < Pi or Pj > Pi. If the
intervalis below Pi, then firm is strictlybetteroff by chargingPi with probabil-
ity Fj(Pi) and not chargingany price below Pi with positive density. However,
if such an intervalexists above Pi, then firmj is strictly better off chargingPA
with probability[1 - Fj(Pi)] and not chargingany price in this interval. In
either case, randomizingover the same set as the rival strictly dominates
randomizingover S.
Assume now thatfirm has a mass pointat Pj with a weightw. The difference
between profitswhen firmi charges(Pj - e) and Pj is given by [wIPjl2 - (ai
+ w13)],whichis strictlypositive for sufficientlysmalle. Therefore,if firm has
a mass point at Pj, firm i is better off not chargingthat price but coming
arbitrarilyclose to it. Notice that by proposition2, holes in the strategy set
cannot exist in equilibrium.
5.
PROPOSITION

sup(S*) = sup(S) = r.

Proof. Suppose, to the contrary, sup(S*) = P < r. The profitfor the ith
firmwhen it chargesP is oxiPsince with probabilityone the other firmis going
to be charginga price less thanP (ties are ruledout by proposition3). Consider
the deviant strategy of charging r with probability one. The expected profit for
the ith firmis air > aiP. Therefore,P = r.

Appendix B
Sketch of Equilibrium When Switchers Are Not Indifferent
In this appendixI sketch the equilibriumwhen the switchersare not indifferent
between the two brands. I consider the two cases identifiedin the text.
Case 1. In this case, switcherspreferbrand1 at equalprices. Their switch-
ing behavioris describedbelow:
buybrandlifPI<P2 + dandPI?r;
buy brand2 ifPI - P2 + d and P2 ' r;
buy neither if PI > r and P2 > (r - d).

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
Promotional Strategies 447

For the pure strategyequilibriumto breakdown, we need (0L2+ P)(r - d) >


cl2r.We will furtherassume that A2 + d> 1. First note that the lowest price
firm2 will charge is given by P2. Given this, firm 1 will not chargeany price
below (P2 + d). Second, note that firm2 will not chargeany price in the range
(r - d, r) since it obtainszero sales fromthe switchersegmentregardlessof the
price of brand 1. It is profitablefor firm 2 to charge r in this case. We can
thereforeassert that firm 1 randomizesover (A2+ d, r) and firm2 randomizes
over (52, r - d, and possibly r). Also note that in equilibriumfirm2's profits
are given by c2r and firml's profitsare given by (a1 + r)(P2 + d). I proceedto
solve for the distributionfunctions below.
ot1P+ [1 - F2(P - d)]P= (ot1+ 1)(A2 + d), P2 + d?<P cr,
a2P + [1 - F1(P + d)]3P= a2r, P2 ' P S r - d.
Solving for F1jandF2, we obtain
O for P AP2+ d,
F1(P) = + ?t2/P - ot2rl(P - d) for 2 + d < P < r,
I for P - r,
ro for P ' P2,
F - Jl+taIl-( t1 + v)(2+d)I(P+d) forA2?P-(r-d),
F2(P) 11 + at/ - (c + )(P2 + d)lr for (r -d) - P c r,
I for P > r.
Fromthese expressions the mass points for firms 1 and 2 at r can be evaluated
as w1 = [1 - F1(r)] and w2 = [1 - F2(r)]. The expressions for w1 and w2 are
providedin table 2 of the text.
Case 2. This case is when switchers prefer brand 2 at equal prices. The
behaviorof the switchers can be describedas:
buy brand I if P1 ' P2 - d and P1 r;
buy brand 2 if P1 > P2 - d and P2 ' r;
buy neither brand if P1 > r - d and P2 > r.
For the pure strategyequilibriumto breakdown, we requirethat (axI+ r)(r -
d) > ujr. This can be simplifiedas d < 13r1(al + Pr).
Further,it is obvious since
P1 > P2,P1 + d > P2. Note that firm 1 will not price below PA,and, conse-
quently,firm2 will not price below PA+ d. Further,firm1 will have no mass in
the interval(r - d, r) since it does not obtain any switcherswhen it prices in
the interval. Thus firm 1 will randomizebetween (P1, r - d, and possibly r),
and firm 2 will randomizebetween (PA+ d, r). By simply changingthe sub-
script, we can write down the distributionfunctions by referringto case 1 as
fo for P _ P1 + d,
F2(P) = + ott/ - otirlp(P - d) forA1 + d 'P ' r,
LI for P - r,
for P P
A1,
F1(P) - 1 + 012/1 - (at2 + p)(A5 + d)I(P + d) for?PI P _ (r-d),
11- ((2 + )(A1 + d)lr
I + t2/ for (r -d) _ P c r,
I for P r.
Fromthese expressions we can write down the mass pointsfor firms1 and 2 as
before: w1 = [1 - F1(r)], and w2 = [1 - F2(r)]. These are provided in table 1.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
448 Journal of Business

Appendix C
Sketch of Equilibrium in a Premium versus Regular Brand Market
In this appendixa sketch of the equilibriumfor the differentiatedproductscase
is providedfor the case (1 - a1)rI < (1 - 2)r2.
Let Pi = (airi)/(ai + >). Then, P1 > P2 + 8. This implies that firm 1 will
never price below PAand firm 2 will never price below (PI - 8). The sets of
prices firm 1 and firm 2 will charge are given by (PI, r1) and (P1 - 8, r2),
respectively. Thus, it can be seen that firm 1 has no supportover the interval
I- 5, ), andfirm2 has no supportover the interval(r2,r1).The distribution
functions are then definedby the following expressions:
a1P + [1 -F2(P-8)]PP= aor, P1 P r,
oU2P + [1 -FI(P + 8)]P= (2 + )PI, P1-8 P r.

These can be solved to yield


for P P1,
F1(P) = + x2/r- (U2 + '(I - 8)/[P(P -8)] for PA P < ri,
1 for P 2 rl,
00 forP?PI -8,
F2(P) = + x1/r - ((xrl)/[lr(P + 8)] for PA P < r2,
I1 for P > r2,
and the mass point for firm 1 at r1is given by
mI = (cxl - U2)P/(U1 + > - 8(U2 + 3)rI/[r2(cxI + A].

References
Blattberg,R. C.; Eppen, G. D.; and Lieberman,J. 1981. A theoreticaland empirical
evaluationof pricedeals for consumernondurables.Journalof Marketing45 (Winter):
116-29.
Blattberg,R. C., and Sen, S. K. 1974.Marketsegmentationusingmodelsof multidimen-
sional purchasingbehavior.Journalof Marketing38:17-28.
Blattberg,R. C., and Sen, S. K. 1976. Marketsegments and stochastic brandchoice
models. Journal of Marketing Research 13:17-28.
Butters, G. 1977. Equilibriumdistributionof sales and advertisingprices. Review of
Economic Studies 44 (October): 465-91.
Dasgupta,P., and Maskin, E. 1986.The existence of equilibriumin discontinuouseco-
nomic games, I: Theory. Review of Economic Studies 53:1-26.
De Palma,A.; Ginsburgh,Y.; Papageorgiou,Y. Y.; andThisse, J. F. 1985.The principle
of minimum differentiation holds under sufficient heterogeneity. Econometrica
53:767-81.
Gal-Or,E. 1982. Hotelling's spatialcompetitionas a model of sales. EconomicLetters
9:1-6.
Guadagni,P. M., and Little, J. D. C. 1983.A logit model of brandchoice calibratedon
scannerdata. MarketingScience 2 (Summer):203-38.
Hotelling,H. 1929. Stabilityin competition.EconomicJournal39:41-57.
Jeuland,A. P., andNarasimhan,C. 1985.Dealing-temporary pricecuts-by selleras a
buyer discriminationmechanism.Journalof Business 58, no. 3: 295-308.
Narasimhan,C. 1984. A price discriminationtheory of coupons. MarketingScience 3
(Spring):128-47.
Narasimhan,C. 1988. A model of discountingfor repeat sales. In T. Devinney (ed.),
Issues in Pricing: Theory and Research. Lexington, Mass.: Lexington Books, in press.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
Promotional Strategies 449

Narasimhan,C., and Rao, R. C. 1985. Models of price promotionunder endogenous


search.Workingpaper.Chicago:Universityof Chicago,GraduateSchoolof Business.
Quelch, J. A. 1987. Influenceson sales promotiondesign. Workingpaper. Cambridge,
Mass.: HarvardUniversity.
Raj, S. P. 1985. Strikinga balance between brand "popularity"and brand loyalty.
Journal of Marketing 49 (Winter): 53-59.
Salop, S., and Stiglitz, J. 1977.Bargainsand ripoffs:A model of monopolisticallycom-
petitive price dispersion. Review of Economic Studies 44 (October): 493-510.
Sobel, J. 1984. The timing of sales. Review of Economic Studies 51 (July): 353-68.
Spatt, C. S. 1981. An intertemporaltheory of sales. Paper presentedat the American
EconomicAssociationmeeting, Washington,D.C., December.
Varian,H. R. 1980.A modelof sales. AmericanEconomicReview70 (September):651-
59.

This content downloaded from 128.252.111.87 on Mon, 24 Mar 2014 15:44:56 PM


All use subject to JSTOR Terms and Conditions
View publication stats

You might also like