You are on page 1of 28

Exit Deterrence∗

Martin C. Byford†
Joshua S. Gans‡

This paper is the first to provide a general context whereby potential entry
can lead incumbent firms to permanently reduce the intensity of competition
in a market. All previous results found that potential entry would lead to lower
prices and greater competition. Examining markets where entry occurs by the
acquisition of access rights from an existing incumbent, we demonstrate that,
where competitive choices are strategic complements, a more efficient entrant
may be unable to acquire those rights from a less efficient incumbent due to the
accommodating behavior of the efficient incumbent. Similarly, such accommo-
dating behavior may deter efficient investment by an incumbent. These results
have implications as to how economists view potential entry and its benefits.

This Version: June 2009

J.E.L. Classification: L13
Keywords: Entry, Dynamic Price Competition, Markov Perfect Equilibrium

We are grateful to Yongmin Chen, Stephen King, Mark Armstrong (the editor)
and two anonymous referees for comments on an earlier draft of this paper. The latest
version of this paper is available at

Department of Economics, University of Colorado at Boulder.

Melbourne Business School, University of Melbourne.
Economic models of the impact of potential entry on competition give rise to two

distinct predictions. First, theories of contestable markets and limit pricing predict
that credible entry threats will induce incumbents to price lower so as to deter that
entry. Second, originating with Selten’s chain store paradox, potential entry will

not induce a prior change in incumbent behavior as it either does not commit them
to maintaining that behavior in the post-entry environment or their own forecasted
behavior in that environment is sufficient to deter entry in of itself. Resolving the

theoretical tension between the two predictions has relied on examination of the
actions incumbents can take to commit themselves to tough post-entry competition
or the possibility that pre-entry behavior might signal relevant information regarding

anticipated post-entry behavior. Regardless, together, all of these theories predict
that potential entry will not lead to less competition in a market.1
The existing models all have in common one feature: that entry, if it occurs, will be

de novo; that is, the entrant adds to the pool of competitors in a market. In contrast,
there are many situations where entry arises by acquisition of an incumbent. While,
in some situations, this may be a choice aimed at eliminating incumbent competitors,

in others, there may be market constraints on de novo entry. For instance, in a
market with high sunk entry costs, a natural oligopoly might arise with a ceiling on
the number of suppliers (Gilbert and Newbery, 1992). In this case, entry may only

be credible if it is by acquisition.
Consider a market in which there exists an asset that is required in order for a
firm to gain access to the market. We term this asset as access right which has char-

acterized by being (a) necessary for a firm to participate in a market; (b) potentially
transferable and (c) rival in that it cannot be utilised by more than one firm at any
point in time. The owner of such an asset may sell, but can never share, its access to a

market. Transferability means that either the asset itself can be traded between firms
See Davis, Murphy and Topel (2004) for a recent treatment and Wilson (1992) for a review.

or, alternatively, it can be traded as a pool of assets through a merger, acquisition or

management buyout.2
Access rights can come in the form of government licensing arrangements — in-
cluding broadcasting licenses for television and radio, and spectrum used by wireless

carriers — as well as complementary assets that are uneconomic to replicate but criti-
cal for competition. For example, sea and air ports require significant amounts of land
that satisfy the very specific requirements. Likewise, an asset such as “eyeballs” is

necessary to compete in the the market for internet advertising; particularly targeted
advertising connected to web searches and internet portals.3
The focus of this paper is on markets with restricted access: that is, where there is

a strictly limited number of access rights. Entry into a restricted access market may
only take place via acquisition of an incumbent’s access right: forcing the incumbent
to exit. Simple intuition would suggest that, if an entrant emerges who is capable of

utilizing the access right more efficiently than an incumbent, there would exist gains
from trade in an exchange in which the access right is transferred to the entrant.
Hence, it would normally be presumed that the emergence of a potential entrant

would (a) not result in any change in pre-entry behaviour of incumbents, and (b)
eventually result in entry.
However, this simple intuition neglects the role that rival incumbents might play

in influencing the terms of trade in the acquisition market for the access rights.
Using an infinite horizon, dynamic model of competition, we show that, in a Markov
perfect equilibrium, the threat of entry may result in reduced competition. In the

model, an inefficient firm competes with an efficient firm in a restricted access market.
The rivalry assumption may seem questionable in a world in which regulators frequently require
firms to grant their rivals access to privately owned infrastructure. However, this issue can be resolved
by distinguishing between the market for infrastructure services, and the downstream market for
which the infrastructure is a necessary input.
Moreover, the term access right could apply to situations in which firms need to access a con-
strained distribution system (Dana and Spier, 2007) or where limitations on physical capital restricts

With the emergence of a potential entrant, the inefficient firm has an incentive to

sell its rights. Nonetheless, we demonstrate that, where the actions of firms are
strategic complements, when firms are sufficiently patient, there exists a Markov
perfect equilibrium in which the efficient incumbent relaxes competition, directing a

stream of profits to the inefficient firm sufficient to eliminate the gains from trade in
the acquisition market. The efficient incumbent takes this action anticipating that the
inefficient incumbent will respond by remaining in the market. In turn, the inefficient

incumbent remains in the market anticipating that this accommodating behavior will
continue into the future.
Because the threat of entry allows the efficient firm to commit to this less aggres-

sive stance, even in the absence of collusion, prices are higher. The higher prices, in
turn, deter exit. Interestingly, compared to the situation where no potential entrant
existed, it is possible for both incumbent firms enjoy higher profits. Hence, incum-

bents may be in favor of policies (e.g., spectrum or broadcast re-sale rights) that
free-up entry into such markets.
This result is significant because, to our knowledge, it is the first paper that pro-

vides a situation whereby potential entry could result in reduced competition in a
market.4 Stiglitz (1981) develops a model whereby potential entry might temporarily
cause prices to increase for a monopolist. He considers a situation whereby a mo-

nopolist chooses to extract a resource at a slower rate so as to forstall the entry of
a competing substitute. Strictly speaking, in his model, it is the availability of the
resource (akin to the incumbent holding additional capacity) that is entry deterring

and the ability to sustain higher prices is, at best, temporary.
The paper proceeds as follows: The general form of the model and the central
There are numerous examples of papers that show that increasing actual competition in a
market can lead to less competition and increased prices (Satterthwaite, 1979; Perloff and Salop,
1985; Stiglitz, 1987; Schulz and Stahl, 1996; Janssen and Moraga-González, 2004; Gabaix, Laibson,
and Li. 2005; Perloff, Suslow, and Seguin, 2006; and Chen and Riordan, 2008). These results are
based on the interplay of product differentiation and/or consumer search. In contrast, in the model
developed in this paper competition is greater post-entry than the pre-entry benchmark.

results are set out in section 1. Section 2 presents a simple example on the Hotelling

line, demonstrating that the emergence of a potential entrant can leave both incum-
bent firms better off. Section 3 presents a generalisation demonstrating that the
equilibrium does not rely on knife edge assumptions. The paper concludes with a

discussion of the results.

1 The Model

This section develops an infinite horizon, dynamic model of competition in a restricted
access market. For the purposes of this paper it is assumed that there exist two access

rights, and as a consequence there can be at most two firms competing in the market
at any point in time. Initially, the access rights are owned by firms 1 and 2, with

firm 1 assumed to be more efficient than firm 2. Firms 1 and 2 interact in a repeated
game of duopoly competition.
At some point an exogenous policy break occurs, permitting resale of access rights.

For instance, the policy change may allow broadcast or spectrum license re-sale after
some earlier allocation to firms 1 and 2. From this point forward firms external to
the market become potential buyers for the access rights, and consequently potential


Assumption 1. It is assumed that: (a) regardless of the history, at least one potential
entrant is present in each period; (b) each potential entrant is as efficient as firm 1;
(c) anti-trust regulators prevent any one firm acquiring both access rights.

Intuitively, (a) can be understood as the analogue of free entry in a restricted

access market. Capital seeks a profit and as a consequence there always exists a
potential buyer for any under performing asset. Where trading of access rights has
occurred in a previous period, a former incumbent may also be a potential entrant.

(b) Can be interpreted as implying that both firm 1, and the potential entrant,
are operating at the technological frontier. The efficiency of the potential entrants

suggests that the policy break permitting resale of access rights is desirable from a

social perspective.5
Potential entrants are indexed in order of arrival. The first potential entrant to
emerge is denoted as firm 3. If firm 3 ever succeeds in acquiring an access right and

entering the market, firm 4 emerges as a potential entrant, and so on.
After a potential entrant arrives, the game consists of two distinct phases. Each
phase is itself a repeated game with the number of periods in each phase endogenously


1. In the entry phase an efficient firm (initially firm 1) competes with firm 2,
under the threat of entry by an efficient potential entrant (initially firm 3).
Each period begins with the incumbent firms selecting actions and realizing

instantaneous profits. The incumbent firms then have the opportunity to sell
their access rights to an entrant. If no transaction takes place, or if the efficient
incumbent sells out, then the game enters a new period of the entry phase. If

the inefficient firm sells out, the game proceeds to the post-entry phase.6

2. In the post-entry phase, two efficient firms compete. In each period of that
phase, the two firms simultaneously select actions before realizing instantaneous

profits. As in the entry phase, either incumbent may trade its access rights to
an equally efficient entrant. This game is repeated infinitely.

All firms discount the future by the common discount factor δ ∈ (0, 1). Because
the transition from the entry to post-entry phase only occurs with firm 2’s consent,

the game may never reach the post-entry phase. Figures 1 and 2 illustrate the intra-
and inter-period timing of the game.
See Cramton (2000) on the effect of the FCC’s ”unjust enrichment” requirements on the ability
to resell spectrum.
If the efficient incumbent exits the market, it is replaced by the entrant, and a new potential
entrant emerges immediately. It is assumed that there always remains a potential entrant to take
its place. Consequently, the replacement of an efficient incumbent, with an equally efficient entrant,
does not alter the structure of the game.

Period t Competition Stage Exit Stage Period t + 1
s s s s s -
firms select profits incumbents
actions (ai , aj ) realised may exit

Figure 1: Intra-Period Timing
Exogenous Policy Change Firm 2 Exits

yes yes
- -
Duopoly Phase Entry Phase Post-Entry Phase

6 6 6
no no

Figure 2: Inter-Period Timing

It will be useful to benchmark the outcomes here with a baseline duopoly phase

involving competition between firms 1 and 2. Each period of the duopoly phase sees
firms 1 and 2 simultaneously selecting actions and realising the consequent instanta-
neous profits. Exit is not possible by selling access rights to eachother.

As will be demonstrated below, the duopoly and post-entry phases share the
characteristic that, in a Markov perfect equilibrium, market outcomes involve firms
considering only instantaneous profits. In contrast, as behavior in the entry phase

might trigger a transition the post-entry phase, firms active in the entry phase consider
outcomes beyond the current period.

1.1 Primitives

All firms in the game share the instantaneous profit function π(ai , aj , θ); where for a
closed bounded interval A ⊂ R, ai ∈ A is the firm’s own action, aj ∈ A is the rival

firm’s action, and θ is a parameter describing the firm’s inefficiency or cost level. It is
assumed that π is twice continuously differentiable in all its arguments and bounded
from above. Moreover, π is assumed to be strictly concave in the firm’s own action

(∂ 2 π/∂a2i < 0), while for all aj and θ there exists a function R(aj , θ) ∈ A such that

∂π R(aj , θ), aj , θ /∂ai = 0. It follows from the continuity and differentiability of π

that R is likewise continuously differentiable. Moreover, if ∂ 2 π/∂ai ∂aj > 0 (< 0) the
actions of firms are strategic complements (substitutes) and ∂R/∂aj < 0 (> 0). It is
assumed that this derivative is bounded such that ∂R/∂aj ∈ (−1, 1).7

For the purposes of this paper, aggressive behavior by a firm is characterised by
that firm selecting a high value of ai , while a low value of ai is interpreted as the firm
adopting a passive position. Formally, this characterization requires the firm’s profit

to be strictly decreasing in its rival’s action (∂π/∂aj < 0).
The inefficiency parameter is assumed to have a negative impact on profits hence
∂π/∂θ < 0; as well as reducing the incentive for a firm to behave aggressively,

∂ 2 π/∂ai ∂θ < 0. This second condition implies ∂R/∂θ < 0. In words, increasing
the inefficiency parameter causes the best response function to shift inward.
In the baseline model it is assumed that firm 2 has an inefficiency of θ2 = c > 0,

while θi = 0 for all remaining firms i ∈ {1, 3, 4, . . . }.

1.2 States

For the firms competing in this market, the state of the world is captured by the
efficiencies of the two firms currently in possession of the access rights, as well as the
existence and inefficiency of a potential entrant. Formally, the state space can be

written as Ω = R2 × (R ∪ ∅) where elements in R represent the inefficiency of the
three respective firms, while a value of ∅ for the third co-ordinate indicates that no
potential entrant exists.

Of course, at most three elements of this space may arise in any given specification
of the model. In the duopoly phase, two firms are active with inefficiencies zero and c;
in the entry phase a potential entrant emerges with inefficiency zero; while in the post-

entry phase both the two active firms, and the potential entrant share inefficiencies of
These bounds ensure the existence and uniqueness of an equilibrium in the static game.

zero. It follows that we can utilize the discrete variables D = (0, c, ∅), E = (0, c, 0)

and P = (0, 0, 0) to represent the state of the world in the duopoly, entry and post-
entry phases respectively. Note that while one efficient firm selling its access right to
another efficient firm changes the identities of the firms active in the market, it does

not alter the state.

1.3 Refining the Set of Equilibria

In accordance with the folk theorem, the model set out above possesses a large number

sub-game perfect Nash equilibria. The purpose of this paper is to investigate the
impact of potential entry on the oligopoly behaviour of the incumbent firms. In other
words, the impact of the (potential) transition between the states D, E and P . The

obvious refinement to employ is Markov perfection.
The set of Markov perfect equilibria (MPE) is produced by refining the set of
all sub-game perfect equilibria to only admit those sub-game perfect equilibria in

which all firms play Markov strategies; that is, strategies that depend only on payoff
relevant information in the game. However, while constraints are placed on the set of
strategies that a firm can employ in equilibrium, every MPE must be robust against

a unilateral deviation by a firm to any available strategy within the game; including
non-Markov strategies.
Markov perfection is a widely employed refinement where examining oligopoly (as

opposed to collusive) behaviour in infinite horizon, dynamic oligopoly models.8 In
the context of this paper the most significant feature of the MPE refinement is that
in equilibrium firms may not condition their actions on the past behavior of any firm,

where that behavior did not impact upon payoff relevant variables. This prevents
firms from employing super-game punishment strategies in equilibrium; and as such
Maskin and Tirole (1988a&b) introduce MPE as means of examining oligopoly behavior over
an infinite horizon, while Ericson and Pakes (1995) developed the foundation for employing MPE
in empirical work. Maskin and Tirole (2001) provide a comprehensive discussion of the virtues of
the Markov perfect refinement in applied work. See Vives (1999, Chapter 9) for a comprehensive

an MPE cannot be regarded as a collusive outcome.

It must be stated that the Markov perfect refinement has a dramatic effect in the
model developed in this paper. As shown below, there exist at most two types of
MPEs in the entry game. However, we do not regard the exclusion of non-Markov

sub-game perfect equilibria as problematic. Quite the opposite. The Markov perfect
refinement allows us to isolate a critical feature of our model. With super-game
punishment strategies excluded in an MPE, any deviation by a firm from its static

best response must be the consequence of at least one firm being able to induce a
transition between states. Thus, in the present model, any discrepancy between MPE
behavior, and the static oligopoly equilibrium, must be the consequence of firm 2’s

ability to trade away its access rights.

1.4 Markov Strategies

In an MPE, firms respond to the prevailing payoff relevant variables in the mar-
ket. It follows that each firm’s MPE strategy can be expressed as a functions of
these variables. When selecting actions for the competition stage, the payoff rele-

vant information is the prevailing state of the world ω ∈ {D, E, P }. It follows that
the competition stage component of firm 1’s strategy can be written as the triplet
(aD E P
1 , a1 , a1 ), where superscripts denote the state corresponding to the actions. Like-

wise, the competition stage component, of the entrant firm i’s strategy, is the pair
(aE P D E
i , ai ); while for firm 2 we have the pair (a2 , a2 ).

In the exit stage of the entry phase the payoff relevant variable is the magnitude

of the profits that a firm receives by staying the market. It follows that we can define
the set πiω ⊂ R such that firm i will exit if, and only if, firm i’s instantaneous profit
from the competition stage of the current period is an element of πiω . The superscript

ω ∈ {E, P } represents the state to which the set πiω applies.

1.5 Equilibrium Duopoly and Post-Entry Phase Profits

In an MPE, actions, and, therefore, instantaneous profits in the duopoly and post-
entry phases are independent of firm behavior in any other phase.

Lemma 1. In a Markov perfect equilibrium firms play their static Nash equilibrium
actions in both the duopoly phase and post-entry phase.

Proof. Once in the post-entry phase, the exchange of access rights cannot alter the

state, thus the post-entry phase repeats indefinitely. It follows that in a Markov
perfect equilibrium the two firms active in the market both select the action aP that

maximises their instantaneous profits; that is to say aP = R R(aP , 0), 0 .

While the game will eventually proceed beyond the duopoly phase, actions in the
duopoly phase neither effect instantaneous payoffs in later phases nor the probability
of entering the entry phase. It follows that in a Markov perfect equilibrium firms 1

and 2 select actions to maximise instantaneous profits, thus aD D
1 = R R(a1 , c), 0 and

aD D
2 = R R(a2 , 0), c .

Lemma 1 establishes that both prior to the policy break that creates the possibility
of a potential entrant, and following the displacement of the inefficient incumbent,
the market behaves as in a one shot game. It is only during the entry phase that the

strategic effect of the option to exit comes into play.
Lemma 1 also defines the instantaneous profits for the duopoly and post-entry
phases, and therefore both the viability of the potential entrant and the incentive for

the efficient incumbent to resist entry.

Lemma 2. In a Markov perfect equilibrium π(aD D P P
1 , a2 , 0) > π(a , a , 0).

Proof. First note that as R(·, c) lies strictly inside R(·, 0), aD P
2 < a . It follows that,

π(aD D P D P P
1 , a2 , 0) > π(a , a2 , 0) > π(a , a , 0),

where the first inequality results from the strict concavity of π, while the second is

implied by assumption that π is strictly decreasing in a rival’s action.

Lemma 2 unambiguously states that entry has a negative impact on the profits of

firm 1, and therefore that firm 1 has an incentive to prevent entry.

Assumption 2. π(aD D P P
2 , a1 , c) < π(a , a , 0) for all c > 0.

Assumption 2 is necessary for the replacement of firm 2 to be viable. Intuitively,
an entrant and firm 2 will never be able to agree on a price for firm 2’s access rights if
the access rights are more valuable to firm 2 than the entrant. Note that assumption

2 is an assumption over the form of the profit function.

1.6 Entry Phase

A potential entrant can only purchase an access right if there exists a fee that is both
acceptable to the entrant, and sufficient to induce the incumbent to exit the market.
For the fee to be acceptable to an incumbent in a sub-game perfect equilibrium, it

must be greater than the stream of profits the incumbent would receive by remaining
in the market. We make the following simplifying assumption:

Assumption 3. All incumbent firms have a weak preference to remain in the market.

Assumption 3 is necessary to avoid an openness problem. Intuitively, a weak
preference to remain in the market can be understood as representing the existence

of arbitrarily small transaction costs associated with the transfer of access rights.
By purchasing firm 2’s access right, an entrant causes the game to transition to
the post-entry phase. It follows that the return to purchasing firm 2’s access right in

the entry phase, or any incumbent firm’s access right in the post-entry phase, is,

Πi = π(aP , aP , 0).

This is also the return that an incumbent firm receives by remaining in the market

in the post-entry phase. The equality of these returns prevents entry in the post-
entry phase as a potential entrant is not willing to offer a fee that is greater than
π(aP , aP , 0). This result generalizes to the entry phase as well.

To analyze the outcomes that occur, we first rule out acquisition by the entrant
of the efficient incumbent.

Lemma 3. In an MPE, an entrant will never purchase the access rights of an efficient

Proof. The exchange of access rights between two efficient firms does not change the

state of the game, and therefore, in an MPE, does not alter the remaining incumbent’s
behaviour. The entrant and efficient incumbent share identical profits functions and
choice sets. It follows that in an MPE, an entrant cannot expect a stream of profits

that is greater than the stream of profits that the original owner of the rights would
be able to secure.

In the proof, this is a direct consequence of the assumption that following the pol-
icy break, at least one potential entrant is always present. Consequently, acquisition
of the efficient incumbent does not change the phase of the game and hence, there

are no strictly positive gains from such an acquisition.9
The return that firm 2 receives by remaining in the market indefinitely — thereby
preventing the game from proceeding to the post-entry phase — is,

Π2 = π(aE E
2 , a1 , c).
If there were only a single potential entrant, then acquisition of the efficient incumbent would
eliminate the threat of entry and return the game to the duopoly phase. However, as is demonstrated
in the Hotelling example of section 2, there exist parameters whereby the profits of the efficient
incumbent are lower in the duopoly phase than in the subsequent entry phase. In such a case a
variant of the lemma will continue to hold. In section 3 we also consider what happens when the
potential entrants are less efficient than the efficient incumbent.


Let F π(aE E
2 , a1 , c) represent the fee that an entrant pays to firm 2 in exchange for

firm 2’s access right where,
δ δ
π(aE E E E P P

F 2 , a1 , c) ∈ π(a2 , a1 , c), π(a , a , 0) ,
1−δ 1−δ

and F 0 ≥ 0.10 In words, F is in the range that is acceptable to both firms, and is
weakly increasing in the current profitability of firm 2. Given firm 2’s weak preference

to remain in the market, any fee that would induce firm 2 to exit where π(aE E
2 , a1 , c) ≥

π(aP , aP , 0) would not be acceptable to the entrant. The following lemma summarises
firm 2’s strategy in the exit stage of the entry phase.

Lemma 4. In an MPE, firm 2 will exit in the entry phase if, and only if, its in-
stantaneous profit is less than the entrant’s equilibrium instantaneous profit in the

post-entry phase; π2 = −∞, π(aP , aP , 0) .

Firm 2’s response to a potential entrant’s offer of F is purely mechanical; an

optimal response to the offer at the point in time that the offer is made. This must be
the case in a Markov perfect equilibrium as Markov perfect equilibria are sequential.
It turns out that this mechanical behavior extends to the competition stage of the

entry phase. In other words, firm 2’s strategy in the entry phase, does not encompass
any strategic objective beyond maximizing instantaneous profits.

Lemma 5. In an MPE, firm 2 always plays its static best response in the competition
stage of the entry phase; aE E
2 = R(a1 , c).

Proof. There two possibilities to consider. If, given firm 1’s choice of aE
1 , firm 2

can select aE E E P P E
2 such that π(a2 , a1 , c) ≥ π(a , a , 0) then selecting a2 to maximise π

delivers firm 2 an outcome that is strictly preferred to any fee that an entrant is

willing to offer.
This formulation for F is satisfied by both the generalised Nash bargaining solution — a rea-
sonable assumption where firms 2 and 3 are engaged in bilateral bargaining over the price of the
access rights — and the case F = 1−δ π(aP , aP , 0) which would be expected to arise where two or
more equally efficient potential entrants are in competition for the access rights.

Where firm 2’s choice of aE E E P P
2 cannot result in π(a2 , a1 , c) ≥ π(a , a , 0) firm 2 will

exit at the end of the current period of the entry phase. Nevertheless, maximising
π remains in firm 2’s best interests as it maximises firm 2’s instantaneous payoff
in the current period, as well as weakly increasing the fee firm 2 receives from the


1.7 Markov Perfect Equilibria

It follows from lemma 5 that there exist at most two classes of Markov perfect equi-
libria to this game: A competitive equilibrium in which firm 2 exits in the first period

of the entry phase, and an accommodating equilibrium in which firm 2 remains in the
market indefinitely. The following lemmas characterise firm 1’s strategies in each of
these equilibria.

Lemma 6. In an MPE in which firm 2 remains in the market indefinitely — an

accommodating equilibrium — aE
1 solves,

π R(aE E
= π(aP , aP , 0).

1 , c), a1 , c

Proof. From lemmas 4 and 5 it is straightforward to see that aE
1 must satisfy,

π R(aE E P P

1 , c), a1 , c ≥ π(a , a , 0). (1.1)

To exclude the inequality note that π R(aE E E
1 , c), a1 , c is strictly decreasing in a1 .

Given that π(aD D P P E
2 , a1 , c) < π(a , a , 0) it follows that in order to satisfy (1.1) a1 <

1 and thus from the single crossing property of the best response functions a1 <

R R(aE1 , c), 0 .

Suppose to the contrary that π R(aE E P P
1 , c), a1 , c > π(a , a , 0) in a Markov perfect

equilibrium. Given the continuity and concavity of π, firm 1 can increase its instanta-
neous profit in each period of the exit phase by marginally increasing aE
1 . For a small

enough increase (1.1) will continue to hold and firm 2 will remain in the market. A

Lemma 7. In an MPE in which firm 2 exits in the first period of the exit phase — a

competitive equilibrium — firm actions and instantaneous profits in the entry phase
are identical to their actions and instantaneous profits in the duopoly phase.

Proof. It follows from lemmas 2, 4 and 5 that it is only necessary to show that
aE D
1 = a1 . To see that this must be the case note that firm 1’s only concern in the

exit phase is to maximise its instantaneous profits.

It follows from lemma 6 that a necessary condition for the existence of an accom-
modating equilibrium is,

1   δ
π aE E E E
π(aP , aP , 0). (1.2)
1 , R(a 1 , c), 0 ≥ π R R(a 1 , c), 0 , R(a 1 , c), 0 +
1−δ 1−δ

That is, the stream of profits firm 1 receives accommodating firm 2 must be at least
as great as the profit firm 1 receives from deviating to its static best response in the

first period of the entry phase, followed by infinite repetition of the post-entry game.
Note that there exists a δ ∈ (0, 1) such that (1.2) holds if, and only if,

π aE E P P

1 , R(a1 , c), 0 > π(a , a , 0). (1.3)

Similarly, from lemma 7 it follows that a necessary condition for the existence of
a competitive equilibrium is,

δ 1
π(aD D
1 , a2 , 0) + π(aP , aP , 0) ≥ π(a1 , aD
2 , 0), (1.4)
1−δ 1−δ

where a1 solves π(a2 , a1 , c) = π(a1 , aD
2 , 0). Firm 1 will not attempt to accommodate

firm 2 so long as the return from accommodating firm 2 is weakly less than the return
from playing static best responses in every phase of the game. As π(aD D
1 , a2 , 0) >

π(a1 , aD
2 , 0), condition (1.4) is always satisfied for δ sufficiently close to zero.

Proposition 1. Where the actions of firms are strategic complements there exist

c̄ > 0 and δ̄ ∈ (0, 1), such that for all c ∈ (0, c̄) and δ ∈ (δ̄, 1) an accommodating
equilibrium exists.

Proof. Define a(c) as the action that solves,

π R a(c), c , a(c), c = π(aP , aP , 0),

and note that a(c) is continuous, strictly decreasing in c, and that as c → 0, a(c) → aP .

The continuity of all functions in c guarantees that there exists numbers ĉ > 0 and
âE P E P
1 < a such that for all c ∈ (0, ĉ), a(c) ∈ (â1 , a ).

For a given c > 0, there exists δ ∈ (0, 1) such that an accommodating equilibrium

exists if, and only if, (1.3) holds. Note that in the limit, as c → 0, (1.3) holds with
equality. Taking the derivative of the LHS of (1.3) yields,

d  ∂π(ai , aj , 0)
π a(c), R a(c), c , 0 = a0
dc ∂ai
0 ∂R(ai , c) ∂R(ai , c) ∂π(ai , aj , 0)
+ a + . (1.5)
∂ai ∂c ∂aj
(ai ,aj )=[a(c),R(a(c),c)]

The first term on the RHS of (1.5) is unambiguously negative as a(c) < R R(a(c), c), 0 ,
and approaches zero as c → 0. The second term is positive where the actions of firms
are strategic complements.

For arbitrarily small c > 0, the first term on the RHS of (1.5) is likewise arbitrarily
close to zero and therefore (1.5) is positive. It follows that there exists c̃ > 0 such that
(1.3) is satisfied for all c ∈ (0, c̃). Defining c̄ = min{ĉ, c̃} completes the proof.

Proposition 1 proves that strategic complementarity is a sufficient condition for

the existence of an accommodating equilibrium for a non-degenerate interval of in-
efficiency parameters. Moreover, it is clear from the proof that the distortion away
from static equilibrium behaviour, necessary to implement the accommodating equi-

librium, is increasing in c. The following propositions provide a comparison of firm
competition and payoffs under the competing equilibria.

Proposition 2. For sufficiently high δ ∈ (0, 1), an accommodating equilibrium deliv-
ers firms 1 and 2 outcomes that are strictly preferred to a competitive equilibrium.

Proof. It is straightforward to see that firm 2 prefers the accommodating equilibrium.

To see that firm 1 is likewise better off note that (1.3) must hold in an accommodating

Proposition 3. In an accommodating equilibrium competition is at its lowest in the
entry phase.

Proof. Given the necessity of strategic complementarity aE D
1 < a1 < a
and aE
2 <

2 .

It is important to note that an accommodating equilibrium can only exit in an

infinite horizon setting. Firm 1 takes an action that is less aggressive than its in-
stantaneous best response, anticipating that firm 2 will respond by remaining in the
market. In turn, firm 2 decides to remain in the market anticipating that firm 1 will

continue its soft stance next period. The equilibrium depends upon the possibility
of future interaction in every sub-game along the equilibrium path; unravelling in a

finite horizon.
In this way, the equilibrium developed in this paper is similar to the MPE of
infinite horizon sequential moves price setting games developed by Maskin and Ti-

role (1988b) and Eaton and Engers (1990). All these equilibria involve firms taking
actions that are less than competitive, and deliver firms profits that are in excess of
competitive levels.

1.8 Costly Entry

Thus far we have assumed that the only cost of entry is cost of acquiring the access

right. In reality, we would expect there to exist both transaction costs associated ex-
changing the access right and some expense required for firm 3 to establish a presence
in the market. Here we show that the addition of an entry cost improves the ability

of the efficient incumbent to accommodate its inefficient rival, while simultaneously
reducing the impact of accommodating behavior on competition in the market.

Suppose that, in addition, to the fee paid for the access right, firm 3 incurs a cost

C > 0 if it enters the market. The imposition of an entry cost reduces the return to
firm 3 of purchasing the access rights. Going forward, the return to purchasing the
access right in any given period is reduced by the amount of the entry cost to,

Π3 = π(aP , aP , 0) − C.

It follows from lemma 4 that, in the presence of an entry cost, firm 2 will trade away

its access rights if, and only if,

π(aE E P P
2 , a1 , c) < π(a , a , 0) − C.

From lemmas 5 and 6 it follows that in order to accommodate firm 2, firm 1 must
select aE
1 such that,

π R(aE E P P
C < π(aP , aP , 0).

1 , c), a1 , c = π(a , a , 0) − (1.6)

The inequality in (1.6) indicates that it is easier for firm 1 to accommodate firm 2
where entry is costly. Given this structure the condition 1.3 is once again necessary

for the existence of an accommodating equilibrium.

Proposition 4. Suppose that the action of firms are strategic complements and that
c ∈ (0, c̄), where c̄ is defined as in proposition 1. For all C > 0 there exists δ ∈ (0, 1)
such that there exists an accommodating equilibrium to the entry game.

Proof. The case in which,
C≥ π(aD D
2 , a1 , c),
can be dismissed as aE D
1 = a1 is sufficient to accommodate firm 2. For the remaining

cases note that by the monotonicity of π R(aE E E
1 , c), a1 , c in a1 , and the proof to

proposition 1, there exists aE
1 ∈ a(c̄), a1
that satisfies (1.6) and therefore, by the

proof to proposition 1, also satisfies (1.3).

Costly entry reduces firm 3’s willingness to pay for the access right. Consequently,

the stream of profits that the efficient firm must direct to its rival — in order to
prevent the inefficient firm from selling out to the entrant — is reduced, and the
magnitude of the distortion that must be created in order facilitate accommodation is

likewise smaller. In other words, the lower the cost of entry that the entrant faces, the
greater will be the distortion away from a competitive outcome in an accommodating

1.9 Endogenous Efficiency Enhancing Investment

By assumption, the efficient production technology is available to all potential en-

trants. It is therefore reasonable to ask: Why does the inefficient incumbent not
simply implement the cost reduction itself? It turns out that, with or without the
presence of a potential entrant, there exists an accommodating equilibrium in which

firm 2 will not adopt an efficiency-enhancing technology, even where upgrading the
technology is costless.
To see this, suppose that as well as (or instead of) the option to exit the market,

firm 2 has the option to reduce its inefficiency from c to zero. Doing so causes the
game to transition to the post-entry phase. Once again it is firm 2 that controls
the transition between states. The return to implementing the efficient technology is
exactly the stream of post-entry profits 1−δ
π(aP , aP , 0). It follows that if there exists
an accommodating equilibrium to the entry game where firm 2’s inefficiency is c, the
same accommodating equilibrium will prevent firm 2 implementing the more efficient

technology unilaterally.

2 Accommodation in Hotelling Competition

This section utilizes the example of price competition on the Hotelling line to illus-
trate propositions 1–3. Hotelling competition provides a clear example as not only is

competition reduced in the entry phase of an accommodating equilibrium in accor-

dance with proposition 3, but the profits of both firms increase with the emergence of
a potential entrant. The Hotelling example also provides a basis for a social welfare
analysis of accommodating behavior.

2.1 The Example

Firm 1 is locates at the left end of a unit line and faces a marginal cost of zero, while

firm 2 is located at the right end of the line and faces a marginal cost of c ∈ (0, 3t).11
A unit mass of consumers is uniformly distributed along the line with each consumer
having unit demand and linear transport cost t > 0 per unit travelled. The market

is assumed to be covered in each phase. If firm 3 replaces firm 2 in the entry phase
it takes firm 2’s position at the right end of the line.
Each period firms compete by simultaneously setting prices. Given that low prices

represent aggressive behaviour, a firm’s action is interpreted as being the negative of
its price. For a firm with marginal cost θ, instantaneous profit is given by the function

pj − pi + t
π(pi , pj , θ) = (pi − θ) ,

where pi is the firm’s own price and pj is the price set by the rival firm. The corre-
sponding best response is given by the function,

pj + t + θ
R(pj , θ) = .

Employing the results of the previous section, this structure is sufficient to establish

the Markov perfect equilibrium behaviour of firms in both the duopoly and post-entry
The upper-bound on c ensures that firm 2 will always produce a strictly positive quantity in a
static equilibrium.

phases. Lemma 1 implies,
1 1 1
1 = t + c, π(pD D
1 , p2 , 0) = t+ c ,
3 2t 3
2 1 1
2 = t + c, D D
π(p2 , p1 , c) = t− c ,
3 2t 3
pP = t, π(pP , pP , 0) = .

It is straightforward to see that π(pP , pP , 0) > π(pD D D D P P
2 , p1 , c) and π(p1 , p2 , 0) > π(p , p , 0).

The value of the variables in the entry phase depend on the nature of the equilib-

rium. If an accommodating equilibrium exists pE
1 is defined implicitly by,

 pE E
1 − R(p1 , c) + t t
π R(pE E E
= = π(pP , pP , 0).

1 , c), p1 , c = R(p1 , c) − c
2t 2

The unique solution to this equation is pE E E
1 = t + c. Moreover, p2 = R(p1 , c) = t + c

yielding instantaneous profits of,

t+c t
π(t + c, t + c, 0) = , π(t + c, t + c, c) = .
2 2

To prove the existence of an accommodating equilibrium it only remains to show

that there exists δ ∈ (0, 1) such that (1.2) holds. R(t + c, c) = t + c/2 thus (1.2)
1 t+c 1 1 δ t
≥ t+ c + ,
1−δ 2 2t 2 1−δ 2
δ c2
c≥ ,
1−δ 4t 
c 1
δ≥ ∈ 0, .
4t + c 7

In words, an accommodating equilibrium exists for all c ∈ (0, 3t) and δ ∈ (1/7, 1); a
wide range of discount factors.

Surprisingly, both incumbent firms benefit from the emergence of a potential en-
trant in an accommodating equilibrium. Given that c < 3t,
t+c 1 1
> t+ c , (2.1)
2 2t 3

indicating that firm 1 enjoys higher profits in the entry phase than it does in ei-

ther the post-entry or duopoly phases. Intuitively, with prices strategic complements
in Hotelling competition, the inefficient incumbent responds to a relaxation of com-
petition by itself reducing competition. In turn this reduces the cost to firm 1 of

accommodating firm 2. Absent a potential entrant, firm 1 cannot credibly commit to

this reduction in competition as R R(p1 , c), 0 < p1 for p1 > t + c/3, implying that
firm 1 has the incentive to unilaterally lower its price. This incentive is eliminated

by the fact that firm 2 will sell out to an efficient entrant should the flow of profits
to the firm fail to match the value of the access rights to the entrant.

2.2 Welfare in an Accommodating Equilibrium

A welfare analysis of the accommodating equilibrium in the Hotelling case is straight-
forward. The twin assumptions of market coverage and unit demand imply that social
welfare is a function of productive efficiency alone.

Productive efficiency, and therefore welfare, is highest in the post-entry phase
where the efficiency of both firms gives rise to a zero cost of production. In the
duopoly phase the inefficient firm serves a fraction q2D = (3t − c)/6t of the market

at a marginal cost of c. In other words, going forward the welfare gain that results
where the inefficient firm is replaced by an efficient entrant is,
c c2
∆WGain = − .
1 − δ 2 6t
By choosing to accommodate its rival in the entry phase the efficient incumbent
not only prevents the efficiency gain ∆WGain but also concedes market share to its

inefficient rival. In turn, this results in the cost of production rising to c/2 each period;
higher than the cost of production in either the duopoly or post-entry phases. In fact,
in an accommodating equilibrium, the emergence of a potential entrant results in the

welfare loss,
∆WLoss = .
1−δ 6t

The c2 term represents the fact that as the inefficient firm’s cost disadvantage in-

creases, so too does the share of the market that the efficient firm must sacrifice to
the inefficient firm in an accommodating equilibrium. At the same time the higher
value of c increases the welfare cost of any given market share that the efficient firm

The order of prices — implied by proposition 3 and illustrated in this example —
have implications for cases in which demand is downward sloping. In an accommo-

dating equilibrium prices are highest in the entry phase and lowest in the post-entry
phase. This suggest that in addition to the loss of productive efficiency, accommoda-
tion may also increase the magnitude of the dead-weight loss in the market.

3 Asymmetric Entrants

The model developed in section 1 assumes that both firm 1 and the potential entrants
share equally efficient production technologies. Here we verify that the existence and

nature of the MPEs is not an artefact of this knife edge assumption.
This section develops a generalisation in which the efficiency of the potential
entrants is less than that of firm 1 by an arbitrarily small amount ε > 0. Formally,

θi = ε for all firms i ∈ {3, 4, . . . } that may emerge as potential entrants. This
difference may represent that firm 1 possesses some market specific knowledge or

advantage, or access to a patented technology, not available to an entrant.
For the purposes of this section it is assumed that the actions of firms are strategic
complements, and that ε < c and c ∈ (0, c̄), where c̄ is as defined in proposition 1.

3.1 States

Given the asymmetry between the efficiencies of incumbents and potential entrants,

the state space is more complex than in the baseline model. Once the policy break
permits trading in access rights, there are four states in the game may operate.

Initially, firms 1 and 2 compete under threat of entry by firm 3; state (0, c, ε). From

this state, an entrant can replace either of the incumbents giving rise to the states
(ε, c, ε) or (0, ε, ε). If entry eventually results in the replacement of both of the original
incumbents the state becomes (ε, ε, ε).

From lemma 3 it follows that in an MPE there can be no entry in the states
(0, ε, ε) and (ε, ε, ε). Consequently, in an MPE, firms play their instantaneous best
(0,ε,ε) (0,ε,ε)
responses in these states. We write these MPE actions as a1 , a3 and a(ε,ε,ε) ;
(0,ε,ε) (0,ε,ε) (0,ε,ε) (0,ε,ε)
and note that π(a1 , a3 , 0) > π(a(ε,ε,ε) , a(ε,ε,ε) , ε) > π(a3 , a1 , ε).

3.2 Accommodating Equilibrium

For sufficiently small ε > 0, accommodation is possible in the states (0, c, ε) and
(ε, c, ε).
The case in which the entrant purchase access rights from the efficient incumbent,

bringing about the state (ε, c, ε), is equivalent to the case analysed in section 1. Given
the continuity assumptions, for sufficiently small ε > 0 an accommodating equilibrium
(ε,c,ε) (ε,c,ε)
exists and the corresponding MPE instantaneous profits satisfy π(a3 , a2 , ε) >
(ε,c,ε) (ε,c,ε)
π(a2 , a3 , c) = π(a(ε,ε,ε) , a(ε,ε,ε) , ε).
Accommodation in the state (0, c, ε) requires firm 1 to take an action a1 such
 (0,ε,ε) (0,ε,ε)
that π R(a1 , c), a1 , c = π(a3 , a1 , ε). Moreover, for sufficiently small ε > 0,
 (0,ε,ε) (0,ε,ε)
π a1 , R(a1 , c), 0 > π(a1 , a3 , 0) and therefore there exists δ ∈ (0, 1) such that
playing the strategy a1 is rational.
To see that an entrant would not be willing to purchase firm 1’s access right in
(0,ε,ε) (0,ε,ε)
the state (0, c, ε), given that π(a(ε,ε,ε) , a(ε,ε,ε) , ε) > π(a3 , a1 , ε), an entrant must
take a softer action in the states (ε, c, ε), than firm 1 must take in the state (0, c, ε),

in order accommodate firm 2. Combined with the efficiency advantage of firm 1,
 (ε,c,ε) (ε,c,ε)
π a1 , R(a1 , c), 0 > π(a3 , a2 , ε).

4 Conclusion

This paper provides a general context where potential entry can reduce the intensity

of competition. The mechanism for this result is the combination of an environment in
which entry requires the acquisition of critical incumbent assets and key competitive
variables are strategic complements. In this situation, an inefficient incumbent may

be deterred from selling assets to an efficient entrant by the accomodating actions of
a more efficient incumbent. That incumbent sacrifices short-term profits to improve
the profitability its rival to such an extent that there are no gains to trade in selling

out to a more efficient entrant.
As a novel result in economics, the result in this paper should make economists
somewhat more cautious in advocating reforms that will improve potential entry in

markets. In some circumstances, we have shown that those reforms may lead to
reduced welfare. Of course, the extent to which the theoretical conclusions reached

here are of relevance in policy-making is ultimately an empirical question that we
leave for future researchers.


Chen, Y. & M. Riordan (2008), Price-Increasing Competition, RAND Journal of
Economics 39, 1042–1058.
Cramton, P. (2000), Lessons from the United States Spectrum Auctions, Testi-

mony before the US Senate Budget Committee.
Dana, J. & K. Spier (2007), Entry Deterrence in a Duopoly Market, Advances in
Economic Analysis & Policy 7, 1653–1653.

Davis, S., K. Murphy & R. Topel (2004), Entry, Pricing, and Product Design in
an Initially Monopolized Market, Journal of Political Economy 112, 188–225.
Eaton, J. & M. Engers (1990), Intertemporal Price Competition, Econometrica

58, 637–659.

Ericson, R. & A. Pakes (1995), Markov-Perfect Industry Dynamics: A Framework
for Empirical Work, Review of Economic Studies 62, 53–82.
Gabaix, X., D, Laibson & H. Li (2005), Extreme Value Theory and the Effects of

Competition on Profits, mimeo, Harvard University.
Gilbert, R. & D. Newbery (1992), Alternative Entry Paths: The Build or Buy
Decision, Journal of Economics and Management Strategy 1, 129–50.

Janssen, M. & J. Moraga-Gonzlez (2004), Strategic Pricing, Consumer Search and
the Number of Firms, Review of Economic Studies 71, 1089–1118.
Maskin, E. & J. Tirole (1988a), A Theory of Dynamic Oligopoly, I: Overview and

Quantity Competition with Large Fixed Costs, Econometrica 56, 549–569.
Maskin, E. & J. Tirole (1988b), A Theory of Dynamic Oligopoly, II: Price Com-
petition, Kinked Demand Curves, and Edgeworth Cycles, Econometrica 56, 571-599.

Maskin, E. & J. Tirole (2001), Markov Perfect Equilibrium, Journal of Economic
Theory 100, 191–219.
Perloff, J. & S. Salop (1985), Equilibrium with Product Differentiation, Review of

Economic Studies 52, 107–120.
Perloff, J., V. Suslow & P. Seguin (2006), Higher Prices from Entry: Pricing of
Brand-Name Drugs, mimeo., Michigan.

Satterthwaite, M. (1979), Consumer Information, Equilibrium Industry Price, and
the Number of Sellers, Bell Journal of Economics 10, 483–502.
Schulz, N. & K. Stahl (1996), Do Consumers Search for the Highest Price? Oligopoly

Equilibrium and Monopoly Optimum in Differentiated-Products Markets, Rand Jour-
nal of Economics 27, 542–562.
Stiglitz, J. (1981), Potential Competition May Reduce Welfare, American Eco-

nomic Review 71, 184–189.
Stiglitz, J. (1987), Competition and the Number of Firms in a Market: Are

Duopolists More Competitive Than Atomistic Markets?, Journal of Political Econ-

omy 95, 1041–1061.
Vives, X. (1999), Oligopoly Pricing: Old Ideas and New Tools, MIT Press: Cam-
bridge (MA).

Wilson, R. (1992), Strategic Models of Entry Deterrence, Handbook of Game The-
ory (R.J. Aumann and S. Hart, eds) Vol 1, North Holland: Amsterdam, Chapter 10.