Upload
byong-duk-rhee
View
212
Download
0
Embed Size (px)
Citation preview
Regional Science and Urban Economics 36 (2006) 99–117
www.elsevier.com/locate/regec
First-mover disadvantages with idiosyncratic
consumer tastes along unobservable characteristics
Byong-Duk Rhee *
Whitman School of Management, Syracuse University, 721 University Avenue, Syracuse,
NY 13244-2450, United States
Available online 24 August 2005
Abstract
Upon entering a market, a first-mover often preempts the best market location and outperforms
followers in competition. Conversely, late entrants are in better positions to learn more about demand
by adopting a bwait-and-seeQ strategy and then enjoy superior performance. This paper demonstrates
first-mover disadvantage, even in the absence of follower’s informational advantage, by using a
standard spatial competition framework. Specifically, a first-mover gains a substantial advantage
over followers when firms correctly predict individual consumer choices with their market locations.
However, the first-mover obtains lower sales and profit when consumers exhibit large idiosyncratic
tastes along unobservable characteristics.
D 2005 Elsevier B.V. All rights reserved.
JEL classification: L11; L13; L15; M31; R39
Keywords: Location decisions; First-mover disadvantage; Consumer heterogeneity; Unobservable characteristics;
Logit model
1. Introduction
Should you enter a new market before or after your rivals? This is one of the most
fundamental decisions encountered whenever a firm establishes its retail store in a
geographic market segment or when it positions a new product in the characteristics space.
0166-0462/$ -
doi:10.1016/j.
* Tel.: +1 31
E-mail add
see front matter D 2005 Elsevier B.V. All rights reserved.
regsciurbeco.2005.06.004
5 443 1874; fax: +1 315 443 5457.
ress: [email protected].
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117100
Entry order takes precedence over production and marketing decisions and has significant
consequences for the long-term market performance of a firm. In the last two decades,
numerous empirical studies in marketing have shown that first-movers achieve substantial
competitive advantages and are likely to be market leaders in their product categories
(Robinson and Fornell, 1985; Urban et al., 1986).
Most theoretical literature has addressed the issue of first-mover advantage from the
perspective of firms’ sequential market entry in a spatial competition framework (Prescott
and Visscher, 1977; Lane, 1980; Anderson, 1987; Tabuchi and Thisse, 1995, to mention a
few). When entering a market, a Stackelberg leader anticipates following firms’ reactions
to its strategies and incorporates them into its own decision making. Using such strategic
foresight, a first-mover preempts later entrants by adopting the best market locations
(Tabuchi and Thisse, 1995). Preemptive positioning then enables the first-mover to charge
a premium price, leading to higher sales and profit. In less attractive segments, on the other
hand, later entrants should offer bargain prices in order to compete with pioneering
incumbents and attain lower sales and profits. First-movers may also preempt superior
strategic resources in the factor market (Rao and Rutenberg, 1979). Thus, comparative
advantages in production and distribution enable the pioneer to outperform later entrants in
competition and lead to above normal returns for the first-mover.
Schmalensee (1982) shows that imperfect information on the part of consumers leads to
first-mover advantage as well. If consumers are satisfied with the first brand (or store) in a
new product category, they will favor it over later entrants because they are uncertain that
the followers’ brands satisfy their needs. First-movers will then sustain a high market share
and profits so long as consumer experience remains as a crucial information source of
product quality. In addition, first-movers can increase switching costs by developing
brand-specific user skills and influencing consumers’ evaluation of products (Stigler and
Becker, 1977).
A growing body of recent empirical evidence, however, questions this entry order effect
on the market performance of a product and presents the cases of first-mover disadvantage.
Golder and Tellis (1993), for example, report that only 53% of first-movers survive in the
market. Kalyanaram et al.’s (1995) survey of various industries also shows mixed results
and suggests that the order of market entry does not appear to be related to long-term
market performance. First-movers bear a higher risk of new product failure since it is
unusually hard to forecast sales for a pioneering brand. And, while we have observed a
remarkable progress in research methodology, we still cannot eliminate uncertain
consumer responses to a pioneering innovation (Hamel and Prahalad, 1994). In contrast,
late entrants are in better positions to learn more about consumer demand by adopting a
bwait-and-seeQ strategy. Later entrants may achieve higher sales and profits when their
informational advantage prevails over any first-mover’s advantage arising from
preemption and switching costs.
Gal-Or (1985, 1987) analytically proves the disadvantages of moving first, even when
both pioneer and later entrant are equally able to assess demand through market research.
Gal-Or (1987) models a duopoly where firms choose output quantities under the
assumption of incomplete information regarding stochastic demand. The follower has an
informational advantage not only because it directly observes market conditions, but also
because it makes inferences about market conditions based on the first-mover’s quantity
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 101
choice. Thus, informational advantage enables the follower to attain higher market share
and profit.1
This study shows first-mover disadvantage even in the absence of the follower’s
informational advantage. Specifically, we demonstrate consumer heterogeneity as a
determinant of first-mover advantage/disadvantage by using a Hotelling (1929)
framework. This paper, however, differs from previous literature cast in a Hotelling
framework (e.g., Neven, 1987; Tabuchi and Thisse, 1995). The prior research assumes that
consumers differ only in their locations on a market segment along which firms position
strategically, but not in their tastes along characteristics that firms cannot manage due to
their unobservability. In other words, consumer preferences are assumed to be completely
accounted for by firms’ market locations. Thus, consumers patronize the product that is
closer to their locations. Consumers who share the same location always purchase identical
products across various choice occasions.
In contrast, we explicitly assume that consumers differ both in their bobservableQlocations and in their tastes along unobservable characteristics. Such unobservable
characteristics might include latent attributes of firm-specific associations and unspecified
situational influences (Manski, 1977). For example, consumers may hold different brand
reputations and images, the underlying attributes of which are not fully captured in a
strategy (Broniarczyk and Alba, 1994). Product choice may also be affected by physical
and social surroundings in a specific choice situation such as decor, sounds, aromas,
lighting, or other persons present and their characteristics and roles (Belk, 1975). In this
setting, even consumers with the same location may choose different products due to their
idiosyncratic tastes along the unobservable characteristics. This study also differs from
Gal-Or (1987) because it does not limit the investigation to the case of a single strategic
variable such as output quantity. Instead, we assume each firm strategically chooses its
own location as well as price. Further, we assume inelastic demand and firms’ identical
information sets of consumer tastes in order to eliminate asymmetric market conditions
and informational advantage as possible sources of either first-mover advantage or
disadvantage. Using these assumptions, we isolate the effect of consumer heterogeneity on
first-mover advantage.
The findings of this study are as follows. When consumer heterogeneity in tastes along
unobservable characteristics is small, a first-mover preempts the market center in order to
capture the largest number of consumers. The follower then positions away from the
pioneer in order to avoid cutthroat price competition. Consequently, the first-mover earns
higher sales and profit due to its positional advantage in competition. However, as the
heterogeneity increases along the unobservable characteristics, consumers do not
necessarily patronize the product that is closer to them along the market segment. Hence,
preemptive positioning becomes ineffective in competition. Furthermore, as the increase in
heterogeneity along the unobservable characteristics generates brand switching over the
1 This result occurs when the firms’ reaction functions are upwards sloping (Gal-Or, 1985). Gal-Or (1987)
assumes that firms are exogenously endowed with private information of given precision. Though firms are
allowed to determine endogenously how precisely to conduct their market research, Gal-Or conjectures that the
follower always earns higher profits than the first-mover and that the first-mover may refrain from any market
research in order to eliminate the follower’s indirect inference.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117102
entire market, the follower aggressively approaches the market center in order to capture
more switching customers. Anticipating such an action, the first-mover then reacts
passively by moving away from the center in order to avoid destructive price competition.
In equilibrium, the follower positions itself closer to the market center and attains higher
sales and profits. When consumers exhibit sufficiently large idiosyncratic tastes along the
unobservable characteristics, product differentiation becomes insignificant in competition
and both firms position at the market center.
The rest of the paper is organized as follows. Section 2 clarifies the idea of
unobservable characteristics in consumer choice and their implications for demand and
profit in a formal model. Section 3 derives a pure strategy equilibrium and analyzes the
effect of unobservable characteristics on first-mover advantage. The last section draws
implications and suggests areas for future research.
2. The model
There are two firms considering market entry. Consumers will purchase one unit of
product per period. Assume a continuum of consumers uniformly distributed on a market
segment with unit length [0,1], as in the standard Hotelling paradigm (1929). The location
of each consumer is denoted by xa [0,1]. Further, assume a quadratic transportation cost
as in d’Aspremont et al. (1979).2 When a consumer at x purchases a product from firm i
located at xi on the market segment, the consumer incurs transportation costs d(xi�x)2,
where the transportation coefficient d N0 is the importance weight associated with
shopping distance in product choice.
We also allow consumers to differ in their tastes along characteristics that are not
observable to firms but still affect consumer choices. This reflects the fact that firms may
not have perfect information on an individual consumer’s decisions in each choice
occasion (Manski, 1977). For example, even though products are identical functionally,
they may differ in latent firm- or brand-specific associations, the underlying characteristics
of which are not fully known to the firms. Product choice is also affected by unspecified
situational influences. Consumers exhibit different choices when they have idiosyncratic
preferences along these characteristics. Moreover, the manner in which consumers
evaluate the large number of different latent attributes may not be systematic either
(Anderson and de Palma, 1992). For example, consumer choices may vary due to
temporary changes in their value of time (Becker, 1976). Consumer variety-seeking could
be a source of idiosyncratic preferences, following Pessemier’s (1978) view that
consumers could better handle their uncertain tastes by purchasing a portfolio of products.
Consumer heterogeneity along unobservable characteristics may be viewed as an error
in judgment on the part of consumers (Manski, 1977). When search costs are relatively
high compared to the potential benefit of making a correct decision or when consumers
have difficulty assessing product characteristics, errors in accurately measuring the utility
2 Alternatively, we may assume a linear transportation cost. However, the nature of the problem remains the
same even in the linear case and derives fundamentally identical results. A detailed formal investigation in the
linear case is available from the author upon request.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 103
of a product will generate diverse product choices that cannot be accounted for by firms’
strategic attributes (Hawkins and Hoch, 1992). In addition, such heterogeneity may occur
because current analytical tools cannot predict consumer choices perfectly due to
functional misspecification or measurement error problems. Product choices, therefore,
vary across individuals and purchase occasions due to idiosyncratic tastes along such
unobservable characteristics, even when consumers share identical locations on the market
segment.
Attribute Y represents unobservable characteristics to the firms and captures the
remaining consumer heterogeneity that is unexplained by location X. Though the firms
observe the distribution of differing tastes along the attribute Y, they cannot manage it
strategically because of the intractability of its underlying characteristics. This approach
has been employed in research on price discrimination (Anderson and de Palma, 1988) and
on product differentiation (de Palma et al., 1985; Rhee et al., 1992). de Palma et al. (1985)
use this approach in a simultaneous entry game framework. Firms compete on price and
location in a one-stage game. Rhee et al. (1992) examine the same issue in a two-stage
game of location and price. See Anderson et al. (1992) for a detailed discussion of this
approach and its application to spatial competition.
A consumer of type yaR on the attribute Y has valuation ei ( y) of product i. Assume
that both firms equally assess information regarding consumer tastes and attempt to model
choices in the same manner. This assumption eliminates informational advantage as a
possible source of first-mover disadvantage. Therefore, a consumer of type (x, y) obtains
the following (indirect) utility in consuming product i:
Ui x;yð Þ ¼ V � d x� xið Þ2 � pi þ ei yð Þ; ð1Þ
where V is a positive constant and may be viewed as an average gross benefit that
consumers at x can obtain from consuming a unit of the product, and pi is the price
charged by firm i. Assume that V is large enough for the consumer to buy either product.
Consequently, the consumer decision is over which product to buy, not over whether to
buy. The assumption of inelastic category demand eliminates asymmetric market
conditions as a possible source of first-mover advantage. If we allow the existence of
consumers who are served by neither firm, some consumers may not purchase either
product if consumption generates negative utility. The resulting asymmetric market
conditions would confound the effect of consumer heterogeneity on first-mover advantage,
which is what we wish to investigate.
Because the firms cannot observe the underlying characteristics of attribute Y, they do
not know the exact value of ei( y) on any given purchase occasion. Thus, from the firms’
perspective, consumer valuations are random along the unobservable attribute. When
estimating sales and profits, the firms treat the valuation, ei( y), as a random disturbance ei.As a result, a firm’s prediction of consumer choice will be probabilistic in nature. The
probability Pri(x) that a consumer at x will purchase product i over j is
Pri xð Þ ¼ Pr ezpi � pj þ d xi � xj� �
xi þ xj � 2x� �� �
; ð2Þ
where e = ei� ej. The normal distribution is an obvious candidate for the distribution of eidue to the central limit theorem. However, it does not provide a tractable form for the
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117104
choice probability. It is well known that the logistic distribution not only closely
approximates the normal but provides a closed-form expression for the choice probability
as well. Then, assume that e has the logistic distribution with mean zero and standard
deviation pr=ffiffiffi3
p. Anderson et al. (1988) show that the logistic choice model can be
derived from an entropic utility function in the representative consumer approach of
Spence (1976) and Dixit and Stiglitz (1977). Note that parameter rz0 measures the
degree of consumer heterogeneity along the unobservable attribute. It also represents the
importance weight associated with the unobservable attribute.
Each firm offers one product and competes non-cooperatively on price and market
location. Without loss of generality, assume that firm 1 positions to the left of firm 2: i.e.,
x1Vx2. Further, assume that the firms have identical cost structures and that marginal costs
are constant and normalized to zero. Fixed costs are not considered in this study.
Hence, a consumer at x will purchase a product from firm 1 with the probability
Pr1 xð Þ ¼ 1þ exp aþ bxð Þ½ ��1; ð3Þ
where
aup1 � p2 þ d x21 � x22
� �r
and bu2d x2 � x1ð Þ
r:
Fig. 1 illustrates the choice probabilities. Notice that Pr2(x)=1�Pr1(x) under the
assumption of inelastic category demand. The probability monotonically decreases for
firm 1, whereas it increases for firm 2. Furthermore, the inflection point of Pr1(x) occurs at
xx ¼ � a
b¼ x1 þ x2
2þ p2 � p1
2d x2 � x1ð Þ : ð4Þ
Pr1(x) is strictly concave over 0VxV x and strictly convex over xVxV1. Consumers to
the left of x have a higher probability of purchasing a product from firm 1, whereas
consumers to the right have a higher probability of purchasing from firm 2. A consumer at
the inflection point has identical probabilities of purchasing a product from either firm
(Pr1(x)=1 /2).
Fig. 1. The choice probability that a consumer will purchase a product either from firm 1 or 2.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 105
When r =0, the probabilistic choice becomes deterministic as in d’Aspremont et al.
(1979),
Pr1 xð Þ ¼ 1 if 0VxVxx0 if xxVxV1:
�ð5Þ
As shown in Fig. 1, all consumers to the left of x patronize firm 1, whereas all
consumers to the right patronize firm 2. As consumers are increasingly heterogeneous
along the unobservable attribute, their choice probabilities become flatter. When rYl,
consumers will show equal choice probability of either product (1/2), regardless of their
locations on the market segment.
Given the uniform consumer distribution on the market segment, firm 1’s expected
demand is
Q1 p1;p2; x1;x2ð Þ ¼Z 1
0
Pr1 xð Þdx ¼ 1� 1
bln
1þ exp aþ bð Þ1þ exp að Þ
� : ð6Þ
Note that Q2=1�Q1. The expected demands are non-negative for r =0 and strictly
positive for r N0. Under the assumption of constant (zero) marginal costs, the expected
profits are P1=p1Q1 and P2=p2Q2, respectively. The expected profits are continuous
functions of location and price over the strategy set, R2 R2, for any rz0. Although the
firms are uncertain as to a particular consumer’s choice due to heterogeneity along
unobservable characteristics, they know the distribution of e and are then able to compute
their expected sales and profits with certainty. Therefore, the firms have perfect
information regarding expected market demand.
3. Sequential entry under price competition
We model competition in a two-stage sequential entry game as in the prior literature
(Neven, 1987; Tabuchi and Thisse, 1995). In the first stage, a Stackelberg leader chooses a
location first. After observing the leader’s market position, the follower chooses its
location. As in Tabuchi and Thisse (1995), we do not limit the location space to the market
space. For example, a firm may open a retail store or shopping mall away from the
residential area. In the characteristics space, a firm may launch a new product that is quite
different from consumers’ ideals.3 In the second stage, each firm determines its own price
competitively. Consequently, when choosing a location, the firm anticipates the impact of
this decision on price competition. Given any pair of firm locations, we first obtain price
equilibrium in pure strategies.
3 Alternatively, firms’ locations can be restricted to lie within the consumer market, [0,1], as in d’Aspremont et
al. (1979). However, the nature of the problem remains the same even in the restricted case and derives
fundamentally identical results. A detailed formal investigation in the restricted case is available from the author
upon request.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117106
3.1. Price competition
Under the assumption of quadratic transportation cost, consumer utility is linear in eand x for given x1 and x2. The joint density of e and x in the utility function (i.e., the
logistic times the uniform densities) is log-concave. Thus, a unique price equilibrium
exists for any rz0 given any pair of locations x1 and x2 (Caplin and Nalebuff, 1991).
Though it is difficult to find closed-form solutions for the equilibrium prices in the general
case, we can obtain the closed-form solutions in the two cases of r =0 and symmetric
locations (i.e., x2=1�x1).
When consumer tastes are completely homogenous along unobservable characteristics
(r =0), the firms’ demands defined in Eq. (6) become
Q1 ¼x2 þ x1
2þ p2 � p1
2d x2 � x1ð Þ and Q2 ¼ 1� Q1: ð7Þ
Given Pi =piQi, the first-order conditions yield the following closed-form solutions:
p14 ¼ d x2 � x1ð Þ x2 þ x1 þ 2ð Þ3
and p24 ¼ d x2 � x1ð Þ 4� x2 � x1ð Þ3
: ð8Þ
If x2=1�x1, the first-order conditions provide a symmetric price solution p1*=p2*=p*
satisfying
1þ 1
2hln
1þ exp hð Þ1þ exp � hð Þ
� � p
2hrexp hð Þ � 1
exp hð Þ þ 1
� ¼ 0; ð9Þ
where h =d(1�2x1)/r. Hence, the symmetric equilibrium prices are
p4 ¼ p14 ¼ p24 ¼ rhexp hð Þ þ 1
exp hð Þ � 1
� ¼ rh coth
h2
�: ð10Þ
The comparative statics in these two cases are consistent with intuition. We obtain
Bpi* /Bxd N0, where xd =x2�x1, in the case of r =0 and Bp* /Bx1b0 in the case of
symmetric locations. Firms lower their prices in equilibrium due to intense price
competition as their locations get closer along the market segment. Moreover, Bpi* /Bd N0in the case of r =0 and Bp* /Bd N0 and Bp* /Br N0 in the case of x2=1�x1. Hence, firms
have more leeway to post higher prices in equilibrium as either d or r increases. This is so
because price competition is alleviated as consumers assign more weight in product choice
to shopping distance (i.e., increasing d) or their tastes along the unobservable
characteristics (i.e., increasing r). The reduced intensity in price competition leads to a
price increase.
As r increases, consumers are more concerned about their preferences along the
unobservable characteristics in making their choices. Thus, pricing becomes increasingly
dependent on idiosyncratic consumer tastes along the unobservable characteristics, but
relatively less on the firms’ locations. When rYl, equilibrium prices are invariant,
regardless of locations. The firms charge identical prices ( p*c2r) and earn identical
profit margins for any pair of locations x1 and x2 (see Appendix for a formal proof).
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 107
3.2. Positioning competition
The previous section shows that a unique price equilibrium exists given any pair of
locations x1 and x2. In that equilibrium, p1*(x1, x2) and p2*(x1, x2) are functions of these
locations. When we substitute prices in the profit functions with the equilibrium prices,
we obtain profits that depend solely on each firm’s location, x1 and x2: P1(x1, x2) and
P2(x1, x2). Without a loss of generality, it is assumed that firm 1 positions before firm 2 in
the sequential entry game. The Stackelberg equilibrium is, therefore, derived from
following first-order conditions:
dPP1
dx1¼ BPP1
Bx1þ BPP1
Bx2
dx2
dx1V0; ð11Þ
BPP2
Bx2z0; ð12Þ
where dx2/dx1 is equal to �(B2P2/Bx2Bx1)/(B2P2/Bx2
2) using the implicit function theorem
under the first-order condition dP2/dx2=0. Note that the second term in Eq. (11)
represents the effect of the first-mover’s location on its profit via the follower’s positioning
strategies.
The complexity of the problem, however, makes it difficult to find closed-form
solutions for the equilibrium locations in the general case. Thus, we obtain closed-form
solutions in the two extreme cases of r =0 and rYl. By using numerical computations,
we show changes in equilibrium locations as r increases from zero to a sufficiently large
number that yields the same solutions as in the case of rYl.
When consumer tastes are identical in the unobservable characteristics (r =0), the
model is the same as Tabuchi and Thisse (1995). We obtain the following profits by
substituting prices in P1 and P2 with p1* and p2*, given in Eq. (8):
PP1 ¼d x2 � x1ð Þ 2þ x2 þ x1ð Þ2
18and PP2 ¼
d x2 � x1ð Þ 4� x2 � x1ð Þ2
18: ð13Þ
Firm 2’s reaction function, x2*=(4+x1) /3, is derived from the first-order condition,
BP2 /Bx2=0. With perfect foresight, the first-mover takes into account the follower’s
reactions and maximizes the profit:
PP1 ¼4d243
50þ 15x1 � 12x21 � 4x31� �
: ð14Þ
As shown in Tabuchi and Thisse (1995), the first-order condition, BP1 /Bx1=0,
produces the following Stackelberg equilibrium locations and prices:
x14; x24� �
¼ 1
2;3
2
�and p14; p24
� �¼ 4d
3;2d3
�:
Given r =0, the first-mover takes a substantial advantage. The first-mover preempts the
market center and attracts the largest number of consumers. The follower then positions
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117108
itself away from the first-mover in order to avoid cutthroat price competition.
Consequently, the first-mover attains higher market share and profit, Q1*=2/3 and
P1*=8d / 9, than the follower’s Q2*=1 /3 and P2*=2d / 9.
Proposition 1. When rYl, there exists an agglomerated equilibrium at the market
center, x1*=x2*=1/2, and corresponding equilibrium prices are p1*=p2*=2r.
Proof. See Appendix. 5
When consumer tastes are very heterogeneous along unobservable characteristics, their
product choices are almost completely determined by their idiosyncratic preferences along
the characteristics. Hence, as shown in the previous section, both firms earn identical profit
margins in the price equilibrium given any pair of locations x1 and x2. Consequently, entry
order and competitive positioning along the observable dimension decline in importance.
Firms merely search for the location which appeals to the largest number of consumers.
Therefore, as in the case of simultaneous market entry (Rhee et al., 1992), both firms
position at the market center in equilibrium and charge identical prices, 2r. Neither firmhas an advantage based on order of entry. Both firms have identical sales, Q1*=Q2*=1 /2,
and profits, P1*=P2*=r.We resort to a numerical method in order to track changes in equilibrium locations if r
is greater than zero, but less than a sufficiently large number that yields an agglomerated
equilibrium at the market center. The profits are computed with a grid size of 10�2 for
0Vr/dV1.4 Note that r/d represents the relative importance of taste along the
unobservable characteristics compared to shopping distance in product choice. We obtain
the following results through numerical computations:
Proposition 2.
(a) When 0Vr /db0.38, there are dispersed equilibria,5 in which x1* is closer to the
market center than x2*.
(b) When 0.38Vr /db0.76, there are dispersed equilibria, in which x2* is closer to the
market center than x1*.
(c) When r /dz0.76, there exists an agglomerated equilibrium at the market center.
Fig. 2 illustrates changes in equilibrium locations with respect to r/d. Figs. 3 and 4
describe the effect of r/d on first-mover advantage in terms of market share and profit.
4 For each case of r/d, equilibrium prices are computed with a grid size of 10�3 for �2Vx1Vx2 and x1Vx2V2.Specifically, given x1 and x2, price equilibrium is obtained from the simultaneous first-order conditions using the
Newton–Raphson method. In order to check the second-order conditions, we fix p1 ( p2) at the equilibrium level
and change p2 ( p1). We confirm that C2 (C1) is decreasing as p2 ( p1) moves away from the equilibrium level.
Taking the equilibrium prices into account, we obtain x2* that generates the largest profit for each given x1. Then,
we choose x1* that provides firm 1 with the largest profit.5 There are two dispersed equilibria for firm 2’s position to the right or to the left of firm 1. We assume x1Vx2 in
the analysis. The equilibrium in the case of x2Vx1 is the mirror image of the one in the case of x1Vx2 with respectto the market center.
Fig. 2. Equilibrium locations with respect to r/d.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 109
Note that the firm that positions itself closer to the market center charges a higher price and
obtains higher market share and profit.
As r/d increases in the interval of 0Vr /d b0.38, x1* moves away from the market
center, whereas x2* moves toward the center. This is consistent with intuition because we
view a large r as assigning more weight to preference along unobservable characteristics
in product choice. As consumer choices increasingly depend upon their idiosyncratic
preferences along unobservable characteristics, they become less loyal to the product
closer to their locations and may switch to the competing product. As shown in Fig. 1,
the inflection point x is on the right segment of the market at a lower level of r, givenx1Vx2. The increase in r enables the follower to take more customers away from the
Fig. 3. Equilibrium market shares with respect to r/d.
Fig. 4. Equilibrium profits with respect to r/d.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117110
first-mover. Since the increase in r also makes preemptive positioning less effective in
competition, the follower (which positions itself far away from the center) moves
aggressively toward the market center in order to capture more switching customers over
the entire market. Anticipating the follower’s approach, the first-mover insulates itself
from intense price competition by moving away from the market center. By doing so,
however, the first-mover erodes its initial strategic advantage. Thus, its sales and profit
decrease. Nevertheless, since x1* is closer to the center than x2* in the interval of 0Vr /
d b0.38, the first-mover still obtains higher sales and profit: i.e., Q1*zQ2* and P1*zP2*.
When r/d=0.38, these two firms, which have evolved in opposite directions, are
equidistant from the center. The symmetric positions then produce identical sales and
profits.
The choice variation arising from idiosyncratic tastes along the unobservable
characteristics leads not only the follower, but also the first-mover, toward the market
center where demand is greatest. As shown in the second term of Eq. (11), however, the
first-mover should take the follower’s approach into account in its positioning, whereas the
follower takes the first-mover’s location as given. Consequently, the follower is more
aggressive in capturing switching customers. When 0.38br /d b0.76, anticipating the
follower’s aggressive positioning, the first-mover realizes that its positioning for a higher
market share would generate destructive price competition. Hence, the first-mover
positions farther away from the center and leaves more consumers for the follower to
capture. The follower, therefore, obtains greater sales and profits under its positional
advantage: i.e., Q2*zQ1* and P2*zP1*.
Note that x1* moves back to the market center as r/d increases over 0.47. As consumer
choices depend increasingly on their different tastes along the unobservable character-
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 111
istics, proximity in distance becomes ineffective in attracting consumers and the firm’s
incentive for product differentiation decreases. When r/d N0.47, the incentive for greater
demand prevails over the incentive for product differentiation. Thus, the first-mover
gravitates toward the market center. When r/d exceeds 0.76, both first-mover and follower
position themselves at the market center as shown in Proposition 1.
4. Conclusion
The late mover’s informational advantage has been advanced as a source of first-mover
disadvantage in previous game-theoretic literature (Gal-Or, 1985, 1987). This paper,
however, presents consumers’ idiosyncratic tastes along unobservable characteristics as
another source and shows the disadvantage of entering a new market first even in the
absence of informational advantage.
The spatial location model in this paper can be generalized to a situation in which two
product attributes form the characteristics space: a strategic attribute, which firms observe
and control such as market location or product features, and an unobservable attribute,
which firms cannot manage strategically due to its intractability. Consumers are
heterogeneous along both attributes, whereas each firm sequentially chooses its position
on the strategic attribute only.
Given the sequential entry, this paper shows that market preemption is not always a
consequence of moving first. The degree of first-mover advantage is determined by how
much consumer heterogeneity is accounted for by dimensions on which a first-mover can
preempt strategically. When their tastes vary mainly along the strategic attribute,
consumers invoke their judgments along the strategic attribute in making their choices.
Thus, a later entrant is sensitive to the pioneering incumbent’s strategic positioning and
strives to move away from the incumbent in order to avoid cutthroat price competition.
Anticipating the later entrant’s defensive positioning, the first-mover preempts the best
market position and earns higher sales and profit.
However, as consumer choices increasingly depend on their idiosyncratic tastes along
the unobservable attribute, the later entrant moves aggressively to the center in order to
capture increasing choice variations over the entire market. Moreover, preemptive
positioning on the strategic attribute becomes an ineffective means of competition. Thus,
anticipating the later entrant’s aggressive approach, the first-mover passively positions
farther away from the center in order to alleviate destructive price competition and attain
lower sales and profit. In other words, when a large portion of consumer heterogeneity
cannot be identified along strategic dimensions, this study recommends a wait-and-see
strategy because a later entrant obtains greater demand and profit, even in the absence of
informational advantage.
First-mover disadvantage arises when consumer heterogeneity along unobservable
characteristics is large (but not so large as to lead to the agglomeration at the market
center). It results from a random component of consumer utility from firms’ perspectives:
e.g., consumers’ idiosyncratic valuations of diverse firm-specific characteristics and brand
associations, etc. When consumers show different preferences for a product along these
firm- or brand-specific attributes, the intractability of the underlying characteristics enables
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117112
late movers to attain a comparative advantage in competition. This finding prescribes that
a first-mover should not only preempt the best market location, but should also maintain
consistent and positively evaluated brand associations across consumers in order to sustain
their pioneering advantages.
The random component may also be attributed to errors in judgment on the part of
consumers. Such errors may occur when the effort associated with an information search is
relatively costly compared to the potential benefit of making a correct decision. It is well
known in marketing that consumer purchases in many product categories are not very
involved, either situationally or on an enduring basis (Hawkins and Hoch, 1992).
Judgment errors will be prevalent in such low involvement decision making. We would,
therefore, expect a higher possibility of later entrants outperforming first-movers in such
consumer markets.
Alternatively, the random component may be the result of not being able to conduct
market research that uncovers all factors contributing to consumer preference. During the
last several decades, much progress has been made in research methodology. As the
sophistication of research methods and related information technology improves, firms
will become more proficient in countering consumer heterogeneity through a more
accurate understanding of consumers’ true preferences. This leads us to expect that firms
will have more leeway in adopting a pioneering strategy.
We employ the logistic distribution of e because it produces a tractable form for the
choice probability. We conjecture that we would obtain the fundamentally consistent
results and implications with any symmetric unimodal distribution of log-concave
density.6 Given the uniform consumer distribution on the market segment X, a symmetric
unimodal distribution of e will lead to the similar choice probability presented in Fig. 1.
Sufficiently high probability of choosing a remote product due to the idiosyncratic
preferences along unobservable characteristics results in the disadvantage of positioning
first on observable dimensions. For example, when we assume a triangular distribution of
e with density f (e)= (1/a)� (jej/a2), where �aV eVa, we obtain the similar choice
probabilities and findings that the first-mover earns greater sales and profit if a is less than
1.3, but lower sales and profit if a is greater than 1.3. In an extreme case of the uniform
distribution, the choice probability becomes linear and both pioneer and follower always
locate at the market center.7
This study derives the results under somewhat restrictive assumptions of inelastic
demand and uniform taste distribution in order to focus on the causality between
idiosyncratic tastes along unobservable characteristics and first-mover disadvantage. An
extension would be to relax these assumptions in order to check whether the causal
relationship would be generalized to other settings. We conjecture that idiosyncratic tastes
along unobservable characteristics will lead to first-mover disadvantage even under the
relaxation. However, the first-mover disadvantage will occur at higher levels of
heterogeneity along unobservable characteristics than in the current model because of
6 Function f(x) is log-concave in x if ln[ f(xE)]zE ln[ f(x0)]+ (1�E) ln[ f(x1)], where xE=Ex0+ (1�E) x1 and0VEV1. A group of log-concave densities includes the normal, uniform, logistic, Laplace, Dirichlet, exponential,
gamma, and beta distributions.7 A detailed formal proof and numerical computations are available from the author upon request.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 113
local monopoly under finite reservation prices and a concentrated mass of consumers with
similar locations (Tabuchi and Thisse, 1995). In addition, this study assumes a duopoly.
We would extend this research to examine the effect of a potential entrant’s threat on the
findings by increasing the number of competing firms in the market.
Another extension would be the case of a quality-type attribute. Rhee (1996)
examines the effect of consumer heterogeneity along unobservable characteristics on
firms’ quality decisions in a simultaneous entry framework and derives the same results
as in the case of spatial competition. Drawing upon Rhee’s (1996) findings, we
conjecture that a sufficiently large heterogeneity along unobservable characteristics
would lead to first-mover disadvantage even when each firm chooses a level of product
quality in a sequential entry game. Multi-dimensional competition would be an
interesting extension as well. Furthermore, as the results generate testable hypotheses,
careful empirical studies should be called for in order to gain real-life insights about the
critical values of r/d.
Acknowledgement
The author wishes to thank Eunkyu Lee, Jae-Hyeon Pae, Andre de Palma, Julie Ruth,
and Jacques-Francois Thisse for their helpful discussions and comments. The author also
gratefully acknowledges funding from the Hong Kong Polytechnic University Research
Grant G-T439.
Appendix A
Proof of Proposition 1. This proof shows an equilibrium when r approaches. Given the
firms’ demands in Eq. (6), we obtain following first-order conditions:
BP1
Bp1¼ Q1 þ p1
BQ1
Bp1¼ 0 and
BP2
Bp2¼ Q2 þ p2
BQ2
Bp2¼ 0: ðA:1Þ
From Eq. (A.1), we obtain
p24 ¼ � Q2
BQ2
Bp2
� �1
¼ r1þ exp að Þ½ � 1þ exp aþ bð Þ½ �
exp að Þ � exp aþ bð Þ½ � ln1þ exp að Þ
1þ exp aþ bð Þ
� :
ðA:2Þ
Note that
limrYl
p24 ¼ limrYl
r1þ exp að Þ½ � 1þ exp aþ bð Þ½ �
exp að Þ � exp aþ bð Þ½ � ln1þ exp að Þ
1þ exp aþ bð Þ
�
¼ r limrYl
1þ exp að Þ½ � 1þ exp aþ bð Þ½ � limrYl
ln1þ exp að Þ
1þ exp aþ bð Þ
�
1
exp að Þ � exp aþ bð Þ½ � ¼ 2r; ðA:3Þ
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117114
because
limrYl
1þ exp að Þ½ � 1þ exp aþ bð Þ½ � ¼ 4; ðA:4Þ
limrYl
ln1þ exp að Þ
1þ exp aþ bð Þ
� 1
exp að Þ � exp aþ bð Þ½ � by l;Hospital
;s rule
¼ limrYl
exp aþ bð Þ � exp að Þ½ �aþ exp aþ bð Þ 1þ exp að Þ½ �b1þexp að Þ½ � 1þexp aþbð Þ½ � exp aþbð Þ�exp að Þf gaþexp aþ bð Þb½ � ¼
1
2:
ðA:5Þ
Hence, p2*c2r when r is sufficiently large. With the same procedure, we obtain
p1*c2r.At p1* and p2*, the second derivatives of P1 and P2 with respect to prices are
non-positive,
B2P1
Bp21¼ exp að Þ � exp aþ bð Þ½ � 2þ exp að Þ þ exp aþ bð Þ½ �
d x2 � x1ð Þ 1þ exp að Þ½ �2 1þ exp aþ bð Þ½ �2V0 ðA:6Þ
B2P2
Bp22¼ exp að Þ � exp aþ bð Þ½ � 2exp að Þexp aþ bð Þ þ exp að Þ þ exp aþ bð Þ½ �
d x2 � x1ð Þ 1þ exp að Þ½ �2 1þ exp aþ bð Þ½ �2V0
ðA:7Þ
because exp(a +b)z (a). Thus, the second-order conditions are satisfied.
As rYl, equilibrium prices approach 2r, regardless of x1 and x2. Thus, the first-ordercondition of P2 with respect to x2 (given x1) becomes
BPP2
Bx2c2r
BQQ2
Bx2¼ 2r
1
b x2 � x1ð Þ ln1þ exp aVð Þ
1þ exp aVþ bð Þ
� �þ 1
x2 � x1ð Þ
�
x2 exp aVð Þ� exp aVþ bð Þ½ � þ exp aVþ bð Þ 1þ exp aVð Þ½ �1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ �
� �¼ 0;
ðA:8Þ
where aVud(x12�x2
2)/ r. From the above first-order condition, we obtain
x2 � x1 ¼ � r2d
ln1þ exp aVð Þ
1þ exp aVþ bð Þ
�
1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ �x2 exp aVð Þ � exp aVþ bð Þ½ � þ exp aVþ bð Þ 1þ exp aVð Þ½ � : ðA:9Þ
The right-hand side of Eq. (A.9) approaches zero as rYl because
limrYl
ln1þ exp aVð Þ
1þ exp aVþ bð Þ
� ¼ 0: ðA:10Þ
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 115
This ensures that, given x1Vx2, firm 2’s best reaction is positioning as close to x1 as
possible.
When r is sufficiently large, the first-order condition of P1 with respect to x1becomes
BPP1
Bx1c2r
BQQ1
Bx1¼ 2r
1
b x2 � x1ð Þ ln1þ exp aVð Þ
1þ exp aVþ bð Þ
� �þ 1
x2 � x1ð Þ
�
x1 exp aVð Þ�exp aVþ bð Þ½ �þ exp aVþ bð Þ 1þ exp aVð Þ½ �1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ �
�� ¼ 0:
ðA:11Þ
From the above first-order condition, we obtain
1
b x2 � x1ð Þ ln1þ exp aVð Þ
1þ exp aVþ bð Þ
� þ exp aVþ bð Þ
x2 � x1ð Þ 1þ exp aVþ bð Þ½ �
þ x1 exp aVð Þ � exp aVþ bð Þ½ �x2 � x1ð Þ 1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ � ¼ 0: ðA:12Þ
As x2Yx1, the first term in Eq. (A.12) becomes
limx2Yx1
1
b x2 � x1ð Þ ln1þ exp aVð Þ
1þ exp aVþ bð Þ
�
¼ limx2Yx1
r2d x2 � x1ð Þ lim
x2Yx1
1
x2 � x1ð Þ ln1þ exp aVð Þ
1þ exp aVþ bð Þ
�
by l;Hopital
;s rule
¼ limx2Yx1
r2d x2 � x1ð Þ
limx2Yx1
� 2dr
��x2 exp aVð Þ � exp aVþ bð Þ½ � þ exp aVþ bð Þ 1þ exp aVð Þ½ �
1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ �
�
¼ limx2Yx1
r2d x2 � x1ð Þ � d
r
�¼ � lim
x2Yx1
1
2 x2 � x1ð Þ : ðA:13Þ
The last term in Eq. (A.12) becomes
limx2Yx1
x1 exp aVð Þ � exp aVþ bð Þ½ �x2 � x1ð Þ 1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ �
¼ limx2Yx1
x1
1þexp aVð Þ½ � 1þexp aVþbð Þ½ � limx2Yx1
exp aVð Þ�exp aVþbð Þx2�x1
by l;Hopital
;s rule
¼ limx2Yx1
x1
1þexp aVð Þ½ � 1þ exp aVþbð Þ½ � limx2Yx1
2dr
exp aVþbð Þ x2�1ð Þ�exp aVð Þx2½ ��
¼ limx2Yx1
2dx1r 1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ � : ðA:14Þ
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117116
Hence, as x2 approaches x1, Eq. (A.12) becomes
limx2Yx1
exp aVþ bð Þ � 1
2 x2 � x1ð Þ 1þ exp aVþ bð Þ½ � � limx2Yx1
2dx1r 1þ exp aVð Þ½ � 1þ exp aVþ bð Þ½ �
¼ 1
4limx2Yx1
exp aVþ bð Þ � 1
x2 � x1
� � dx1
2rby l
;Hopital
;s rule
¼ 1
4limx2Yx1
2d 1� x2ð Þexp aVþ bð Þr
� � dx1
2r¼ d 1� x1ð Þ
2r� dx1
2r¼ d 1� 2x1ð Þ
2r¼ 0:
ðA:15Þ
Therefore, Eq. (A.15) leads to the following Stackelberg equilibrium locations:
x14;x24� �
¼ 1
2;1
2
�:
Consequently, under sufficiently large s, two firms position themselves at the market
center with an infinitesimal distance between them. In equilibrium, they charge identical
prices, p1*=p2*=2r, and obtain identical demands and profits: Q1*=Q2*=1 /2 and
P1*=P2*=r. 5
References
Anderson, S.P., 1987. Spatial competition and price leadership. International Journal of Industrial Organization
5, 367–398.
Anderson, S.P., de Palma, A., 1988. Spatial price discrimination with heterogeneous products. Review of
Economic Studies 55, 573–597.
Anderson, S.P., de Palma, A., 1992. The logit as a model of product differentiation. Oxford Economic Papers
44, 51–67.
Anderson, S.P., de Palma, A., Thisse, J.-F., 1988. A representative consumer theory of the logit model.
International Economic Review 29, 461–466.
Anderson, S.P., de Palma, A., Thisse, J.-F., 1992. Discrete Choice Theory of Product Differentiation. The MIT
Press, Cambridge.
Becker, G.S., 1976. The Economic Approach to Human Behavior. University of Chicago Press, Chicago.
Belk, R.W., 1975. Situational variables and consumer behavior. Journal of Consumer Research 2, 157–164.
Broniarczyk, S.M., Alba, J.W., 1994. The importance of the brand in brand extension. Journal of Marketing
Research 31, 214–228.
Caplin, A., Nalebuff, B., 1991. Aggregation and imperfect competition: on the existence of equilibrium.
Econometrica 59, 25–59.
d’Aspremont, C., Gabszewicz, J.J., Thisse, J.-F., 1979. On Hotelling’s stability in competition. Econometrica 47,
1145–1150.
de Palma, A., Ginsburgh, V., Papageorgiou, Y.Y., Thisse, J.-F., 1985. The principle of minimum differentiation
holds under sufficient heterogeneity. Econometrica 53, 767–781.
Dixit, A.K., Stiglitz, J.E., 1977. Monopolistic competition and optimum product diversity. American Economic
Review 67, 297–308.
Gal-Or, E., 1985. First mover and second mover advantages. International Economic Review 26, 649–653.
Gal-Or, E., 1987. First-mover disadvantages with private information. Review of Economic Studies 54, 279–292.
Golder, P.N., Tellis, G.J., 1993. Pioneer advantage: marketing logic or marketing legend? Journal of Marketing
Research 30, 158–170.
B.-D. Rhee / Regional Science and Urban Economics 36 (2006) 99–117 117
Hamel, G., Prahalad, C.K., 1994. Competing for the Future. Harvard Business School Press, Boston.
Hawkins, S.A., Hoch, S.J., 1992. Low involvement learning: memory without evaluation. Journal of Consumer
Research 19, 212–225.
Hotelling, H., 1929. Stability of competition. The Economic Journal 39, 41–57.
Kalyanaram, G., Robinson, W.T., Urban, G.L., 1995. Order of market entry: established empirical generalizations,
emerging generalizations, and future research. Marketing Science 14 (Part 2), G212–G221.
Lane, W.J., 1980. Product differentiation in a market with endogenous sequential entry. The Bell Journal of
Economics 33, 237–260.
Manski, C.F., 1977. The structure of random utility models. Theory and Decision 8, 229–254.
Neven, D.J., 1987. Endogenous sequential entry in a spatial model. International Journal of Industrial
Organization 5, 419–434.
Pessemier, E.A., 1978. Stochastic properties of changing preferences. American Economic Review 68, 380–385.
Prescott, E.C., Visscher, M., 1977. Sequential location among firms with foresight. The Bell Journal of
Economics 8, 378–393.
Rao, R.C., Rutenberg, D.P., 1979. Preempting an alert rival: strategic timing of the first plant by analysis of
sophisticated rivalry. The Bell Journal of Economics 10, 412–428.
Rhee, B.-D., 1996. Consumer heterogeneity and strategic quality decisions. Management Science 42, 157–172.
Rhee, B.-D., de Palma, A., Fornell, C., Thisse, J.-F., 1992. Restoring the principle of minimum differentiation in
product positioning. Journal of Economics and Management Strategy 1, 475–505.
Robinson, W.T., Fornell, C., 1985. Sources of market pioneer advantages in consumer goods industries. Journal
of Marketing Research 22, 305–317.
Schmalensee, R., 1982. Product differentiation advantages of pioneering brands. American Economic Review 72,
349–365.
Spence, M., 1976. Product selection, fixed costs, and monopolistic competition. Review of Economic Studies 43,
217–235.
Stigler, G.J., Becker, G.S., 1977. De gustibus non est disputandum. American Economic Review 67, 76–90.
Tabuchi, T., Thisse, J.-F., 1995. Asymmetric equilibria in spatial competition. International Journal of Industrial
Organization 13, 213–227.
Urban, G.L., Carter, T., Gaskin, S., Mucha, Z., 1986. Market share rewards to pioneering brands: an empirical
analysis and strategic implications. Management Science 32, 645–659.