Games and Economic Behavior: Amos Fiat, Elias Koutsoupias, Katrina Ligett, Yishay Mansour, Svetlana Olonetsky

JID:YGAME AID:2252 /FLA [m3G; v 1.121; Prn:8/01/2014; 14:33] P.
1 (1-20)
Games and Economic Behavior ••• (••••) •••–•••
Contents lists available at ScienceDirect
Games and Economic Behavior

www.elsevier.com/locate/geb
Beyond myopic best response (in Cournot competition) ✩

Amos Fiat a,1 , Elias Koutsoupias b,c,2 , Katrina Ligett d,3,∗ , Yishay Mansour a,4 ,
Svetlana Olonetsky a,4
a
The Blavatnik School of Computer Science, Tel Aviv University, Israel
b
Department of Informatics and Telecommunications, University of Athens, Panepistimiopolis, Athens 15784, Greece
c
Department of Computer Science, University of Oxford, OX1 3QD, UK
d
Department of Computing and Mathematical Sciences and Division of the Humanities and Social Sciences, California Institute of Technology,
Pasadena, CA 91125, United States
a r t i c l e i n f o a b s t r a c t
Article history: The Nash equilibrium as a prediction myopically ignores the possibility that deviating from
Received 2 May 2012 the equilibrium could lead to an avalanche of beneficial changes by other agents.
Available online xxxx We consider a non-myopic version of Cournot competition, where each firm selects either
profit maximization (as in the classical model) or revenue maximization (by masquerading
JEL classification:
as a firm with zero production costs). We consider many non-identical firms with linear
C72
D40 demand functions and show existence of pure Nash equilibria, that simple dynamics will
produce such an equilibrium, and that some natural dynamics converge within linear time.
Keywords: Furthermore, we compare the outcome of the non-myopic Cournot competition with that
Cournot competition of the standard Cournot competition. Prices in the non-myopic game are lower and the
Bertrand competition firms, in total, produce more and have a lower aggregate utility.
Delegation games We also briefly consider a non-myopic version of Bertrand competition, and find that prices
Dynamics increase relative to the classical model.
© 2013 Elsevier Inc. All rights reserved.
1. Introduction
The Cournot and Bertrand predictions are both famously problematic, empirically and experimentally. To quote Abreu
(1986), “In recent times this model [Cournot–Nash] has been criticized for being too static, and thereby yielding predictions
which are misleadingly competitive.” For example, the Cournot–Nash equilibrium defines a best response to a given strategy
profile of the other agents, a−i , to be the best action possible, under the assumption that the other players will not deviate
✩
An earlier extended abstract appeared in SODA 2012.
*
Corresponding author. Fax: +1 626 792 4257.
E-mail addresses: fiat@tau.ac.il (A. Fiat), elias@di.uoa.gr (E. Koutsoupias), katrina@caltech.edu (K. Ligett), mansour@tau.ac.il (Y. Mansour),
svetlana.olonetsky@gmail.com (S. Olonetsky).
1
This research was supported in part by the Google Inter-university center for Electronic Markets and Auctions and in part by a grant from the Israeli
Science Foundation, and in part by The Israeli Centers of Research Excellence (I-CORE) program (Center No. 4/11).
2
This research was supported in part by the European Union’s Seventh Framework Programme FP7 / grant 284731 (UaESMC), by the ERC advanced grant
No. 321171 (ALGAME), and by the ESF-NSRF research program Thales (AGT).
3
Research performed in part while a postdoctoral associate at Cornell University and while visiting the Hebrew University. Research supported in part
by an NSF Mathematical Sciences Postdoctoral Fellowship (NSF Award DMS-1004416), NSF grants CCF-0910940 and CNS-1254169, the Charles Lee Powell
Foundation, and a Microsoft Faculty Fellowship.
4
This research was supported in part by the Google Inter-university center for Electronic Markets and Auctions, The Israeli Centers of Research Excellence
(I-CORE) program (Center No. 4/11), by a grant from The Israeli Science Foundation, by a grant from United States–Israel Binational Science Foundation (BSF),
and by a grant from the Israeli Ministry of Science (MoS).
0899-8256/$ – see front matter © 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.geb.2013.12.006
JID:YGAME AID:2252 /FLA [m3G; v 1.121; Prn:8/01/2014; 14:33] P.2 (1-20)
2 A. Fiat et al. / Games and Economic Behavior ••• (••••) •••–•••
from a−i . The Cournot competition model highlights some potential problems with treating the Nash equilibrium as the
inevitable outcome of competitive play.
Consider the following example in the Cournot setting: There are two oil producing firms, Wildcat Drillers and
W. Petroleum. Wildcat Drillers has a production cost of $0.5 USD per mega-barrel; W. Petroleum has a production cost
of $0.3 USD per mega-barrel. If the price per mega-barrel decreases linearly, specifically, if price = (1 − total supply in
mega-barrels), then the Cournot competition equilibrium price is $0.6. At this equilibrium price, both firms are producing
and no firm can benefit by unilaterally changing its production quantity, assuming that the other firm does not change its
production quantity. (In our toy example the price drops down to zero when the world supply is one mega-barrel of oil.)
If W. Petroleum were to increase its production such that the price dropped below $0.5, Wildcat Drillers would be pro-
ducing at a loss. The inherent assumption in the Cournot–Nash equilibrium is that if this happened Wildcat Drillers would
indeed continue producing at the same level as before, despite this loss, or that W. Petroleum would never manipulate
the market in this manner. However, W. Petroleum may hypothesize that if the price were driven down, Wildcat Drillers
would in fact cease production, rather than continuing production at a loss. This hypothesis seems rather natural, but its
predictions are not captured by traditional Cournot–Nash equilibria.
The impetus for our work is a sense of unease about the assumption that agents act myopically and ignore responses to
their own actions. In the context of competition, it seems natural that firms should be able to predict something about the
behavior of other firms, as a function of changes in pricing.
Our work follows a direction pioneered by Vickers (1985), Fershtman and Judd (1987), and Sklivas (1987).5 The model
used in many of these papers is a fixed-depth extensive form “delegation” game: the first stage6 is an “owners game,” and
the second stage is the “managers game.” Essentially, the first stage players (sometimes called the owners or principals)
set parameters for the second stage players (sometimes called the managers or the agents). Once the owners give these
parameters to their agents, the agents are expected to compute and play equilibria of an underlying agent game. Alternately
and equivalently, one can view the delegation problem as the question “What incentives should the principal (owner) offer
the agent (manager)?.”7
While the existing literature seeks to optimize incentives for agents so as to maximize profits, our view of this type of
multistage game is quite different. A delegation game can be interpreted as a way to make sense of off-equilibria behavior:
as the principal sets the utility for the agent, this can encode arbitrary rules for best responses.
In the case of Cournot competition, which is the main focus of this paper, we consider two natural strategies for the
principal: tell the agent to maximize revenue (select action RM), or tell the agent to maximize profit (select action PM).
When a principal selects PM the agent simply tries to maximize firm profits (similar to the standard Cournot competition).
However, when a principal selects RM, the agent ignores production costs, and attempts to maximize firm revenue. After
each selecting one of these two strategies, agents for each firm participate in a Cournot competition, where the PM agents
use their true production costs to determine production levels and the RM agents use a production cost of zero to decide
how much to produce. As in the standard Cournot competition, firms experience utilities as determined by their true
production costs. The major difference between the PM/RM game and the underlying Cournot competition is that when a
principal changes its action in the PM/RM game, it results in a change in the production quantities of the other firms (by
converging to an equilibrium of the underlying Cournot competition).
We consider the case of multiple, non-identical firms. Previous work (except that considering only two firms) has treated
only the symmetric case, wherein all firms have the same production cost. Previous work on the delegation game has been
in the continuous case, allowing the principal to select among convex (and even non-convex!) combinations of revenue
maximization and profit maximization; in such settings it is easy to see that an equilibrium exists.
In the case of asymmetric Bertrand competition, we study two firms in the continuous cost setting, where the firms can
act as if they have a different cost than their true cost. Our qualitative results match those of Fershtman and Judd (1987),
Sklivas (1987), predicting higher prices in the meta-game; however, to our knowledge, we present the first quantitative
study of this case.
1.1. Prior related work
Cournot competition assumes a so-called conjectural variation model (Arrow, 1960), i.e., the Cournot conjectured variation
is that if one firm changes its production level then other firms will not adjust their production level accordingly. Under
this assumption, the Cournot competition is a Nash equilibrium; in this setting the Nash equilibrium is sometimes referred
to as a Cournot–Nash equilibrium.
As noted above, this Cournot conjectured variation is a subject of much debate and criticism in the economics literature,
with conflicting conclusions. Abreu (1986) describes how the threat of punishment in an extended game could support
higher prices than the Cournot equilibrium prices. By contrast, Riordan (1985) considers a setting with imperfect information
5
We remark that similar ideas are due to Kurz (1977), who defined a “distortion game,” wherein agents strategically misrepresent their types to a
taxation mechanism (in the context of the Aumann–Kurz income distribution game).
6
Some papers have delegation game models with three steps.
7
By allowing arbitrary (not necessarily implementable) “compensation functions,” Fershtman et al. (1991) give a folk theorem for achieving Pareto
efficiency in delegation games (this directly implies Abreu’s comment, without an extensive form game).
A. Fiat et al. / Games and Economic Behavior ••• (••••) •••–••• 3
where firms only see the prices they receive. In a multistage game, a firm could increase its output to lower the market
clearing price, causing rival firms to think that the demand curve has shifted down, and hence inducing them to lower their
outputs in the future. Thus, the market price will be lower than that projected by the Cournot competition prices.
Without assuming an extensive form game, Schelling (1960) suggested that one could make “a credible threat” (that one
might not act to maximize profit alone) by delegating authority, e.g., using thugs for extortion or sadists for prison guards.
In general, there are a large number of papers dealing with delegation, and not only for making threats. Many of these
papers give examples of market competition between firms.
There is also a strong connection between Stackelberg equilibria and the delegation game, e.g., if only one player strate-
gizes in the owners’ game, and the others don’t, then the strategic owner will become a Stackelberg leader, and the others
Stackelberg followers (see Berr, 2011, for this result and others, and for a large bibliography).
1.2. Summary of results
Previous work considered continuous mixtures of the two (where it is easy to show that equilibria exist). Furthermore,
except for the case of two firms, previous work assumed that the production costs are equal for all firms. We extend this
work in several directions: different production costs, discrete (binary) options rather than a continuous set of delegation
strategies, and the study of dynamic convergence to the equilibria of the game.
In our discrete, binary PM/RM Cournot model, we show that there always exists a pure Nash equilibrium, and that the
resulting equilibrium price of the PM/RM game is at most the Cournot competition market price and at least half of it.
On the other hand, the aggregate utility of the firms participating in the competition might be significantly lower in the
PM/RM game.
Conceptually, we show that in the Cournot meta-game, strategizing about others’ responses increases competition, re-
duces prices, and improves social welfare, all while reducing corporate profits.
We are also interested in the dynamics underlying the Cournot competition and the PM/RM game. (We believe we are
the first to consider dynamics of the delegation game.) Interestingly, a single change of strategy in the PM/RM game may
result in a dynamic cascade of best response moves in the underlying Cournot competition. For example, if W. Petroleum
increases production, then the market price will go down, and if it goes down enough then some firms may drop out of
the market (e.g., Wildcat Drillers might stop production). As firms drop out of the market, the total supply goes down,
and — possibly — firms that previously were not producing anything (say, a new company called Texas Oil) suddenly start
production.8
The dynamics of the Cournot competition itself have been studied at length, and it is well known that best response
dynamics do not necessarily converge (Theocharis, 1960). However, it is known (Even-Dar et al., 2009) that in regret mini-
mization the action frequencies converge to the Cournot–Nash equilibrium.
We show that best response dynamics in the PM/RM game always converge to a pure Nash equilibrium. We also demon-
strate simple dynamics that converge in a linear number of updates, and thus such an equilibrium is polytime-computable.
Finally, we show that a random ordering of best response moves converges in an expected polynomial number of updates.
One could also argue that a combination of best response by principals and regret-minimization by agents would give
dynamics that converge to the unique Nash equilibrium of the PM/RM game.
We also consider two important special cases of the PM/RM game, in which we give a complete characterization of the
pure Nash equilibria: (a) only two firms in the game and (b) all firms have the same production cost (the symmetric case).
In the symmetric case it is interesting to observe that there are non-symmetric pure Nash equilibria. In fact, for any choice
of i firms selecting PM and m − i firms selecting RM, there is a cost c for which this strategy profile is in equilibrium. Except
in the case of two firms, it seems that prior work on the delegation game has been limited to studying the symmetric case.
Finally, in Section 6, we discuss some results for a similar meta-game overlaid on the asymmetric Bertrand model. We
study two firms in the continuous cost setting, where the firms can act as if they have a different cost, which we refer to as
a meta-cost. In equilibrium, the meta-cost would be a function of the two underlying costs and the maximum demand in
the market. As in Fershtman and Judd (1987), Sklivas (1987), we find that players report higher meta-costs than their true
costs, resulting in higher prices than in the underlying Bertrand competition.
2. The Cournot model
2.1. Standard linear Cournot competition
We consider a set of m firms, M = {1, . . . , m}, producing an identical good, where firm i has production cost c i per unit
of production. Every firm chooses a production level xi ∈ [0, 1]. Let x = x1 , x2 , . . . , xm be the joint production levels of
all m firms. The linear Cournot model we consider here assumes the market price is a linearly decreasing function of the
production levels, that is,
8
The dynamics described above are the dynamics of the underlying Cournot competition, and can be inferred as a consequence of actions in the PM/RM
game. In the PM/RM game, there may also be meta-level cascading effects; for example, firms may move from maximizing profit to maximizing revenue,
and then, after other firms respond (in the PM/RM game), they may go back to maximizing profit.

m
p (x) = 1 − b i xi , (1)
i =1
for strictly positive constants b1 , b2 , . . . , bm . The profit (utility) of firm i ∈ M is the profit per unit of production times the
quantity produced, i.e.,

u i (x) = p (x) − c i · xi .
Consider a linear Cournot competition with firms i ∈ M = {1, . . . , m} and production costs c i . A Cournot–Nash equilibrium
eq eq eq eq eq
is a joint production level, xeq = x1 , x2 , . . . , xm , where for each firm i, xi maximizes the utility for firm i, given x−i .9
That is,
eq eq
xi ∈ arg max u i x−i , x for all 1 i m.
x
Note that we could write xeq = f (xeq ) as a fixed point of the vector-valued function

f (x) = arg max u 1 (x−1 ), . . . , arg max um (x−m ) .
x1 xm
As u j depends on the production cost c j , we define xeq (c ) to be the fixed point of f where c = (c 1 , . . . , cm ) is a vector of
costs such that c i appears within player i’s utility function.
The following proposition, and variants thereof, are well known. We give the proof only for the sake of completeness.
eq eq
Proposition 2.1. Given a linear Cournot competition of m firms with production costs c = (c 1 , . . . , cm ), and let xeq (c ) = (x1 , . . . , xm )
be the production levels at some Cournot–Nash equilibrium. Let N ⊆ M = {1, . . . , m} be the set of firms with strictly positive production
eq
levels at equilibrium, i.e., N = {i ∈ M | xi > 0}, and let n = | N |.
The Cournot–Nash equilibrium has the following characteristics:
1. For any firm i ∈ N (with strictly positive production levels), we have
eq p (xeq (c )) − c i
xi = . (2)
bi
2. The market clearing price at equilibrium is

1+ i∈N ci
p xeq (c ) = . (3)
n+1
We define p eq (c ) = p (xeq (c )).
3. The utility of non-producing firms ( j ∈
/ N) is zero, and the utility of producing firms (i ∈ N) is
( p eq (c ) − c i )2
u i xeq (c ) = . (4)
bi
Proof. To compute the Cournot–Nash equilibrium we take the derivative of u i (x) = ( p (x) − c i ) · xi with respect to xi . It
follows from Eq. (1) that
∂
u i (x) = p (x) − c i − b i xi . (5)
∂ xi
It follows from Eq. (5) that
eq
b i xi (c ) = p xeq − c i = p eq (c ) − c i .
eq
Note that in equilibrium a firm i ∈ M has xi (c ) > 0 iff c i < p eq (c ). Taking the sum over all the firms N ⊆ M with strictly
positive production levels we have
eq
| N | p eq (c ) − cj = b j x j (c ) = 1 − p eq (c ),
j∈N j∈N
where the second equality follows from the definition of the market price in a linear Cournot competition (Eq. (1)). This
implies that the market clearing price at equilibrium is
9
We denote by x−i the vector x except for the i-th component, and by (x−i , a) the vector x where the i-th component is replaced by a.

1+ j∈N cj
p xeq = p eq (c ) = .
n+1
eq
Thus, the utility of a firm i ∈ N, at equilibrium, is ( p eq (c ) − c i ) · xi (c ) = ( p eq (c ) − c i )2 /b i . 2
2.2. The PM/RM game
To address the issue that actions of one firm may impact the actions of another, resulting in an outcome other than a
Cournot–Nash equilibrium, we study a binary delegation game, which we refer to as the PM/RM game. In the game, a firm’s
principal selects between two strategies for its agent:
1. PM (profit maximization), and

2. RM (revenue maximization).
These strategies could be enforced by compensation structures for the agent offering a percentage of firm profits, or revenue,
accordingly.
In this PM/RM game, as in the Cournot competition, we have a set of M firms {1, . . . , m}, and each firm i has a production
cost c i . Each firm has a principal that selects an action in {PM, RM}; for simplicity, we will attribute both the action and the
resulting utility to the firm itself. Let g (c , RM) = 0 and g (c , PM) = c. Given a joint action z ∈ {PM, RM}m , we define a virtual
cost vector y ( z) such that y i ( z) = g (c i , zi ).
Effectively, the principal determines a virtual cost, which could be either the true production cost or zero. In both cases,
the agent takes this virtual production cost and chooses a production level corresponding to that production cost in the
standard Cournot competition. When production costs are zero, profit and revenue are identical, and thus we can consider
such an action as revenue maximizing.
We now consider the Cournot–Nash equilibrium of this virtual Cournot competition, played with virtual production costs
y i ( z) = g (c i , zi ) rather than c i . For this Cournot–Nash equilibrium we have production levels xeq ( y ( z)), and price p eq ( y ( z)).
It follows from Eq. (2) that the production levels derived from the virtual Cournot competition are as follows:
eq
1. If firm i chooses profit maximization (PM) then the production level is xi ( y ( z)) = ( p eq ( y ( z)) − c i )/b i .10
eq eq
2. If firm i chooses revenue maximization (RM) then the production level xi ( z) = xi ( y ( z)) = p eq ( z)/b i .
Similar to the state of affairs for a standard Cournot competition, the utility of firm i ∈ M in the PM/RM game is
u i ( z) = ( p eq ( z) − c i )xi ( z). Note that a firm’s utility in the PM/RM game is determined using the true production costs, not
the virtual production costs.
In this model, market prices will always be positive, i.e., p eq ( z) 0. Similarly, the production level of any firm is always
non-negative: xi 0, ∀i. Let N eq ( z) be the set of firms with strictly positive production levels, given the joint action z of
the PM/RM game. Let PM( z) be set of PM players with strictly positive production levels at joint action z, PM( z) = {r: zr =
PM, cr < p eq ( z)}, and let RM( z) be set of RM players at z, RM( z) = {r: zr = RM}.
A principal i that selects zi = PM is guaranteed a non-negative utility for his firm: Either it does not produce (xi ( z) = 0)
or it produces (xi ( z) = ( p eq ( y ( z)) − c i )/b i > 0), and in both cases u i ( z) = b i x2i ( z). A principal that chooses to maximize
revenue always has strictly positive firm production level, and the firm may find itself with negative utility. However, in the
equilibria of the PM/RM game, all firms have non-negative utility (since all principals always have the option of playing
PM).
We define the best response correspondence of a firm i as BRi ( z−i ) to include all the best response actions, given that
the other firms’ actions are z−i . Since we have only two actions, we sometimes abuse the notation and talk about the best
response action, when it is unique. An improving best response for firm i given joint action z is a zi such that zi ∈ BRi ( z−i )
and zi ∈ / BRi (z−i ). An improving best response sequence is a sequence of joint actions z1 , . . . , zk , in which each joint action
z j +1 is derived from the preceding joint action z j by a single firm making an improving best response move.
3. Nash equilibria and dynamics of the Cournot PM/RM game
In this section, we study the properties of the PM/RM game, establish the existence of pure Nash equilibria, and discuss
convergence of best response dynamics.
3.1. Existence and properties of Nash equilibria
The main theorem of this section establishes existence of pure Nash equilibria of the PM/RM game by showing that any
sequence of improving best response moves converges to a pure Nash equilibrium.
10
As y ( z) is a function of z we will use the notations p ( z) and p ( y ( z)) interchangeably, and do likewise for arbitrary other functions of y ( z).
Fig. 1. Consider joint action z with firms i, j such that c i > c j . Fig. 1(a) corresponds to Lemma 3.3, Fig. 1(b) corresponds to Lemma 3.4.
Theorem 3.1. Any sequence of improving best response moves in the PM/RM game converges to a pure Nash equilibrium.
The theorem is proved via a sequence of characterization lemmas. We defer longer proofs, along with some of the
lemmas, to Appendix A.
We first make an observation relating the prices when two players playing PM and RM swap their actions.
Observation 3.1. Consider firms i and j with production costs c i > c j . Fix some joint action z, and let p be the price if j changes to
PM and i changes to RM from z; let p be the price if i changes to PM and j changes to RM from z. Then, p p .
Proof. We argue that p p , by considering a change of stated cost for player i, starting from joint action ( z−i , j , RM, RM).
One can view the cost change of firm i in two stages. In the first stage it increases its cost to c j , thus setting the system
price to p (it can be the case that i does not produce at p ). In the second stage, firm i completes its cost change by
increasing it by the remaining c i − c j (in the case where i does not produce after the first stage, we have p = p ). Since
the price is monotone in the cost, we get p p . 2
The next lemma plays an essential role in understanding the structure of Nash equilibria of the PM/RM game. It states
that when a firm switches from revenue maximization to profit maximization, the price increases (and therefore the number
of producing firms cannot decrease so long as the switching firm continues production).
Lemma 3.2. Let z−i be a joint action of all firms except of some firm i, and consider the two joint actions z pm = ( z−i , PM) and
zrm = ( z−i , RM) in which firm i has action PM and RM, respectively. Let n pm = | N ( z pm )| and nrm = | N ( zrm )| denote the number of
producing firms in the two joint actions and let the corresponding market prices be p pm = p eq ( z pm ) and p rm = p eq ( zrm ). Then
1. p pm > p rm , and
2. if firm i produces at z pm , then n pm nrm .
We next show that if firm j strictly prefers to switch from RM to PM in an joint action z, then any firm i with higher
production cost that plays RM in z would also strictly prefer to switch to PM.
Lemma 3.3. Consider firms i and j with production costs c i > c j . Consider a joint action z where zi = z j = RM. If in z firm j strictly
prefers PM, then firm i also strictly prefers PM. (See Fig. 1(a).)
Similarly, if firm i strictly prefers to switch from PM to RM in the common action z, then any firm j with lower
production cost that plays PM in z would also strictly prefer to switch to RM.
Lemma 3.4. Consider firms i and j with production costs c i > c j . Suppose zi = z j = PM. If in joint action z firm i strictly prefers RM,
then firm j would also strictly prefer to switch to RM from z. (See Fig. 1(b).)
We use the above lemmas to show that certain sequences of joint actions cannot be part of any improving best response
sequence.
Lemma 3.5. Consider joint action z with zi = PM, z j = RM and c i > c j . In addition, consider the following joint actions: z =
, PM). Then the sequence of joint actions z, followed by z , followed by z cannot be an improving best response
( z−i , RM), z = ( z− j
sequence. (See Fig. 2(a).)
Fig. 2. Impossible series of best response moves with firms i, j such that c i > c j . Fig. 2(a) corresponds to Lemma 3.5; Fig. 2(b) corresponds to Lemma 3.6.
Proof. If z is an improving best response to z , then u j ( z ) > u j ( z ). From Lemma 3.3 it follows that u i ( z) > u i ( z ) should
also hold, which contradicts the assumption that z followed by z is an improving best response sequence. 2
Lemma 3.6. Consider joint action z with zi = PM, z j = RM and c i > c j . In addition, consider the following joint actions: z =
, RM). Then the sequence of joint actions z, followed by z , followed by z cannot be a best response sequence.
( z− j , PM) and z = (z− i
(See Fig. 2(b).)
Proof. If z is an improving best response to z , then u i ( z ) > u i ( z ). From Lemma 3.4 it follows that u j ( z) > u j ( z ) should
also hold, which contradicts the assumption that z followed by z is an improving best response move. 2
The following lemma plays a central role in showing that any best response dynamics converges to a pure Nash equilib-
rium. The lemma shows that if there is a sequence of firms switching from RM to PM, then in the initial joint action, the
lowest cost firm among them would have a best response to switch from RM to PM.
Lemma 3.7. Let z be a joint action with both firms i and j playing RM. Consider an improving best response move of firm i followed
by an improving best response move of j, both changing their strategies from RM to PM. If c i > c j , PM is an improving best response
action for j given z.
We can also show a similar lemma for firms switching from PM to RM:
Lemma 3.8. Let z̄ be a joint action with both firms i and j playing PM. Consider an improving best response move of firm j followed
by an improving best response move of i, both changing their strategies from PM to RM. If c i > c j , RM is an improving best response
action for i given z̄.
Combining lemmas, we show existence of pure Nash equilibria by considering the outcome of a sequence of best re-
sponse moves.
Proof of Theorem 3.1. Suppose that the game does not converge to Nash equilibrium, so there is a sequence of best response
moves that cycles. Consider firm j with lowest cost of those changing action in the cycle. Let P be the maximal chain of
RM → PM moves that includes such a move by j.
Consider P of length at most 2. If j is the first best response move of P , then it contradicts Lemma 3.5 (Fig. 2(a)). If j is
the last best response move of P , then it contradicts Lemma 3.6 (Fig. 2(b)).
We are left with P of size at least 3. Let z be the joint action in the beginning of P . Applying Lemma 3.7 recursively, we
get that BR j ( z− j ) = PM. Let i be the firm that made the best response move before P (it has to be BRi ( z−i ) = RM). Using
Lemma 3.5, we arrive at a contradiction, and thus no such cycle can exist. 2
3.2. Fast convergence of best response dynamics
Consider a joint action z in the PM/RM game. If z is not a Nash equilibrium then at least one of the firms prefers to
switch its strategy. We will first show that a particular order of best response moves leads quickly to a Nash equilibrium.
To do this, let us define BRF( z) to be the set of firms that prefer to switch strategy: BRF( z) = {i | zi ∈
/ BRi (z−i )}. Intuitively,
among the firms playing PM at z, the one with the lowest production cost is most likely to prefer to switch strategy.
Similarly, among the firms in RM( z), the one with the maximum production cost seems most likely to switch strategy.
We will consider dynamics that take advantage of this intuition. Let us define minPM( z) to be the firm with minimum
production cost among firms playing PM at z and maxRM( z) to be the firm with maximum production cost among firms in
RM( z).
Consider the following natural best response dynamics:
While BRF( z)
= ∅, perform one of the following actions
1. If minPM( z) ∈ BRF( z), firm minPM( z) changes its strategy.

2. If maxRM( z) ∈ BRF( z), firm maxRM( z) changes its strategy.
In the proof of the following lemma, we show that if BRF( z)

= ∅ then one of the actions is applicable.
Lemma 3.9. The above procedure converges to a Nash equilibrium in at most 2m steps.
Proof. According to Lemma 3.4, if minPM( z) ∈ / BRF(z) then no firm PM(z) is in BRF(z). Similarly if firm maxRM(z) ∈ / BRF(z),
then no firm in RM( z) is in BRF( z). Therefore, one of the two steps can be always performed while BRF( z)
= ∅.
If we order the firms in decreasing order according to their production cost, the current joint action is a vector in
{PM, RM}m . The procedure either replaces the rightmost PM or the leftmost RM. Furthermore, the most recent action cannot
be undone immediately (otherwise it wouldn’t be a best response).
The claim is that the procedure terminates in at most 2m steps. To see this, observe that at the beginning the procedure
changes the leftmost RM or the rightmost PM until the vector consists of a sequence of PM’s followed by a sequence of
RM’s (or reaches Nash equilibrium in less than m steps). From that point on, all the PM’s precede the RM’s in the current
vector. It follows that it takes at most m steps to reach the point where the PM’s precede the RM’s and at most m more
steps to reach the final vector. 2
Starting from a joint action in which all firms play PM, the above dynamics will converge in at most 2m steps to a Nash
equilibrium. It follows that
Corollary 3.10. There is a polynomial time computation to find some Nash equilibrium of the PM/RM game.
Based on the result above that there always exists a Nash equilibria in which all lower production cost firms play RM
and all higher production cost firms play PM. However, not all Nash equilibria have this property:
Example 3.11. Consider two firms that have costs 0.30 and 0.28. It is straightforward to see that (RM, PM) is a Nash
equilibrium (and also (PM, RM) is an equilibrium).
Lemma 3.9 demonstrates that a particular scheduling of best-response moves converges quickly to a Nash equilibrium;
below we show that even a random sequence of best response moves converges in polynomial time.
Theorem 3.12. A random sequence of best response moves takes at most O (m2 ) steps in expectation to converge to a Nash equilibrium
of the PM/RM game.
Proof. Consider a joint action z, followed by a sequence of improving RM-to-PM moves; Lemma 3.7 implies that for the
firm k of lowest cost in the sequence, PM is an improving best response action given z — i.e., these firms could have been
reordered to make their moves in order of their cost (from lowest to highest). Similarly, Lemma 3.8 implies that a sequence
of firms making improving PM-to-RM moves can be reordered by cost, from highest to lowest, and still all the moves would
be best responses.
Now consider some intermediate joint action z in the execution of an arbitrary sequence of best response moves. Suppose
the subsequent best response move is by a firm k such that either
• k makes a RM-to-PM best response move and ck is less than the cost of firm minPM(z), or
• k makes a PM-to-RM best response move and ck is greater than the cost of firm maxRM(z).
This is possible only under two exceptional conditions:
• the only moves before k’s were of the same type (e.g., if k made a RM-to-PM best response move at z and all previous
moves in the entire best response sequence were also RM-to-PM), or
• the most recent prior move of the other type (e.g., the most recent RM-to-PM move, if k’s move was PM-to-RM) was
actually a move by the then-current minPM or maxRM firm.
This is because, as observed above, if firm k’s best response move of RM-to-PM had been preceded by a sequence of
other RM-to-PM moves, those moves could have been reordered so that k’s move was directly preceded by a PM-to-RM best
response move (unless there was no such preceding move). That preceding PM-to-RM best response move could not have
been by a firm of cost greater than ck , by Lemma 3.5. This leaves the possibility that the preceding PM-to-RM best response
move was by a firm of cost less than ck (and thus of cost lower than minPM( z)). The case when k makes a PM-to-RM best
response move follows an analogous argument using an application of Lemma 3.6.
Thus, with the above exceptions, the only way in which a best response sequence can cause the id of either the minPM
or maxRM firm to change is if that firm is actually selected to best respond. The exception involving a sequence of moves
not preceded by any moves of the other type introduces at most m moves to the beginning of the best response sequence,
and then this exception cannot be invoked again. The second exception, which technically could decrease the id of firm
minPM (analogously, increase the id of firm maxRM), only occurs in conjunction with a previous move such that when the
two moves are considered together, the id of firm minPM actually increases (analogously, the id of firm maxRM decreases).
These two moves can be grouped together, so that this second type of exception effectively means that an increase in the
id of firm minPM (analogously, decrease in the id of firm maxRM) by 1 could require two best response steps.
Now, consider an arbitrary sequence of best response moves, drop the prefix corresponding to the first exception, and
focus for the moment only on the moves that either increase the id of firm minPM or decrease the id of firm maxRM
(rearranging and grouping moves as necessary to handle the second exception above). This subsequence, because no inter-
mediate moves may change the id of either minPM or maxRM, is bounded in length as in Lemma 3.9, by 2m. Now notice
that if the best response moves were selected at random, the expected number of other best response moves between any
two in this subsequence is at most m/2. Thus, the expected total number of best response moves in a random sequence
is O (m2 ).
Notice that this same proof also demonstrates that cycling through the firms in any fixed ordering, offering each the
chance to best respond in turn, also converges in at most O (m2 ) best response moves; a bound can be similarly proven
on any sequence of best response moves where every firm is given the opportunity to best respond with at least some
minimum probability (or with at least some minimum frequency). 2
4. Instances of special interest for the Cournot PM/RM game
We first consider special two cases of the game: a complete characterization of all the pure Nash equilibria when only
two firms compete in the market (appears as Theorem A.3 in Appendix A), and the symmetric case when all firms have the
same cost and for each firm i, b i = 1 (which we consider below, and which provides useful tools for the following section).
We consider the case of m symmetric firms with cost c for each, playing the PM/RM game. Namely, each firm selects an
action in {RM, PM}. The firms that select the action RM will act as revenue maximizers (behave as though their production
cost is zero). The firms that select PM will act as profit maximizers. Note that since this is a symmetric case, at equilibrium
all the firms will be producing, i.e., n = m.
Suppose that k firms select the action PM and n − k firms select the action RM. In this case the price is pk = 1n+ kc
+1 . Firms
that select PM will produce xkp = pk − c = 1−(nn++11−k)c , and firms that will select RM will produce xkr = pk = 1n+kc
+1 . The utility
of firms that select PM is
2
1 − (n + 1 − k)c
ukp = xkp pk − c = ,
n+1
while the utility of firms that select RM will be

1 + kc 1 − (n + 1 − k)c
ukr = xkr pk − c = .
n+1 n+1
We will now compute for which costs c is it a Nash equilibrium to have k firms selecting PM and n − k firms selecting
RM.
−1
Theorem 4.1. Let ak = n(n−kn)+ k+n−1
for k ∈ [1, n], a0 = 0, bk = n(nn−−k1)+k for k ∈ [0, n − 1] and bn = 1. If the players’ cost c ∈ [ak , bk ]
then there is a pure Nash equilibrium with k firms selecting PM and n − k firms selecting RM.
1−(n−k)c 2
Proof. If a firm selecting RM deviates to PM (k n − 1), then its new utility would be ukp+1 = ( n+1
) . Action RM will
be an improving best response if
2
1 + kc 1 − (n + 1 − k)c 1 − (n − k)c
>
n+1 n+1 n+1
2
⇒ (1 + kc ) 1 − (n + 1 − k)c > 1 − (n − k)c
⇒ 1 − (n + 1)c + 2kc − k(n + 1 − k)c 2 > 1 − 2(n − k)c + (n − k)2 c 2

⇒ (n − 1)c > (n − k)2 + k(n + 1 − k) c 2
n−1
⇒ > c.
n(n − k) + k
If a firm selecting PM deviates to RM (k 1), then its new utility would be

1 + (k − 1)c 1 − (n + 2 − k)c
ukr −1 = .
n+1 n+1
Action PM will be an improving best response if
2
1 − (n + 1 − k)c 1 + (k − 1)c 1 − (n + 2 − k)c
>
n+1 n+1 n+1
2
⇒ 1 − (n + 1 − k)c > 1 + (k − 1)c 1 − (n + 2 − k)c
⇒ 1 − 2(n + 1 − k)c + (n + 1 − k)2 c 2 > 1 − (n − 2k + 3)c − (k − 1)(n + 2 − k)c 2

⇒ (n + 1 − k)2 + (k − 1)(n + 2 − k) c 2 > (n − 1)c

⇒ n(n − k) + k + n − 1 c > (n − 1)
n−1
⇒ c> .
n(n − k) + k + n − 1
This implies that for ck ∈ [ak , bk ] there is a pure Nash equilibrium with k firms selecting PM and n − k firms selecting
−1
RM, where ak = n(n−kn)+ k+n−1
for k ∈ [1, n] and bk = n(nn−−k1)+k for k ∈ [0, n − 1]. Note that ak = bk−1 , and let a0 = 0 and
b = 1. This covers the entire range of symmetric production costs. 2
n
5. Comparison of the PM/RM game and Cournot competition
In this section we compare the outcome of the PM/RM game versus the standard Cournot competition. We start by
comparing the prices.
Assume z is a pure Nash equilibrium of the PM/RM game, where n firms are producing and m − n are not. Since this is
an equilibrium, all other firms that select PM do not produce and have zero utility. Any producing firm i ∈ M has production
cost c i < p ( z), and any firm which has c i < p ( z) is producing.
Consider the (standard) Cournot competition price, which is equivalent to having all firms play PM. In this case, we can
think of computing the price in two steps: first, we let the firms that selected RM switch to PM, and then we let any firm
that was not producing before (since its cost was higher than p ( z)) produce. After the first stage, the price is at least the
original price, and at most
1 + n · p ( z) 1 − p ( z)
p = p ( z) + .
1+n 1+n
After the second step, since we are adding firms with production cost c i ∈ [ p ( z), p ], the price can only go down (but
1
remains above the initial price of p ( z)). We can also give a lower bound to p ( z), since clearly p ( z) n+ 1
.
We can now bound the difference between the standard Cournot competition price, p c , and the price set by the PM/RM
game, as follows:
pc p 1 − p ( z)
1 =1+
p ( z) p ( z)(1 + n)
p ( z)
1 1 1
=1+ − 2−
p ( z)(1 + n) (1 + n) 1+n
1
2− ,
1+m
1
where the first inequality uses the fact that p ( z) n+ 1
.
We can also show that the above bound is almost tight. Consider the case of symmetric firms with production cost
1
c= m − m12 . The pure Nash equilibrium of the PM/RM game in this case is to have all firms playing RM (see Section 4). The
1
1+mc 2− m
standard Cournot competition price is 1+m
= 1+m
— contrawise, in the pure Nash equilibrium of the PM/RM game, the
price is 1+1m . Thus, the ratio between the prices is 2 − 1
m
. This is summarized in the following theorem:
Theorem 5.1. Let p c be the Cournot competition price and p pr be the PM/RM game price. Then,
pc 1
1 2− ,
p pr 1+m
pc 1
and, for appropriate production costs, we have that p pr
=2− m
.

In our setting the price p defines the total production, since i ∈ M b i xi = 1 − p. Since the price in the PM/RM game is at
least half the price of the Cournot competition, the total revenue is at least half. (Note that the amount produced increases
as the price decreases.)
We now compare the utilities of the firms in the two settings. We show that the utility can be significantly different.
1
Consider the case of symmetric firms with production cost c = m − m12 and bi = 1 for all i. The utility of each firm in the
Cournot competition is
1 2 2
2− m 1 1 2m2 − m − m(m + 1) + (m + 1) 1
− + = =Θ . (6)
1+m m m2 m2 (1 + m) m 2
The utility of each firm in the PM/RM game is

1 1 1 1 m2 − m(1 + m) + (1 + m) 1 1
− + = =Θ . (7)
1+m m m2 1+m m2 (1 + m) 1+m m4
This implies that the ratio between the utilities can be as large as m2 . Since all the utilities are identical, the same ratio
holds for the sum of the utilities.
Theorem 5.2. There are instances where the sum of the firm utilities when playing the (standard) Cournot competition is a factor of
Θ(m2 ) greater than the sum of the firm utilities when playing the PM/RM game.
We now give a not-quite-matching upper bound for the symmetric case (when all firms have the same cost and b i = 1).
Consider a PM/RM game with a Nash equilibrium where k firms select PM and n − k select RM. Above we analyzed the
case of k = 0. From Section 4, in equilibrium, the sum of the utilities of all k firms playing PM is
2
1 − (n + 1 − k)c
kukp = k
n+1
2
1 − (n + 1 − k) n(nn−−k1)+k
k
n+1
k
= ,
(n + 1)2 (n(n − k) + k)2
since by Theorem 4.1 c = n(nn−−k1)+k is the largest possible cost for an equilibrium where k firms select PM.
Similarly, the sum of the utilities of all n − k firms playing RM is

1 + kc 1 − (n + 1 − k)c
(n − k)ukr = (n − k)
n+1 n+1
1 − (n + 1 − k)c
(n − k)
(n + 1)2
1 − (n + 1 − k) n(nn−−k1)+k
= (n − k)
(n + 1)2
n−k
= .
(n + 1)2 (n(n − k) + k)2
Therefore the social welfare in the PM/RM game is at least

n 1
=Ω .
(n + 1)2 (n(n − k) + k)2 n 5
On the other hand, the social welfare of the (standard) Cournot competition is
2
1−c 1
n .
n+1 n
This establishes the following lemma:
Lemma 5.3. When all production costs are equal and b i = 1 for all i (the symmetric case), the sum of firm utilities playing the standard
Cournot competition is at most a factor of O (m4 ) greater than the sum of firm utilities playing the PM/RM game.
6. Bertrand model
To demonstrate the generality of our methodology, in this section we study a two firm asymmetric Bertrand competition.
Here, each firm has a parameter which represents its production cost; another model parameter describes the maximum
demand in the market. The quantity that a firm sells is decreasing in its cost and increasing in the other firm’s cost.
However, the sum of demands is decreasing in the sum of the costs.
Our main goal is to study a meta-game, where the firms can submit a meta-cost, which can be different from the actual
cost. The prices and demand are set using the meta-costs, and the firms optimize their profit with respect to the true
cost. We exhibit the resulting equilibrium for the case of two firms, and show, under some technical assumptions on the
parameters, that the meta-costs are at least the true costs and hence the prices are going up in the meta-game compared
to the original Bertrand competition.
We start by introducing the asymmetric Bertrand competition model. We have two firms {1, 2}. Initially, each firm
i ∈ {1, 2} declares a price p i 0. Given the two prices, the ( A , a, b, c 1 , c 2 )-Bertrand competition sets the quantity sold by
firm i to be11
qi = A − bp i + ap 3−i .
Our ( A , a, b, c 1 , c 2 )-Bertrand competition has five parameters: (1) A, the maximal demand, (2) b, the sensitivity of the
quantity to price of the product, (3) a, the inverse sensitivity of the quantity to price of the other product, and (4) a cost of
c i for firm i to produce a unit of output. We will assume that b a 0. We will also make a simplifying assumption that
A is sufficiently large w.r.t. c i , specifically, A b max{c 1 , c 2 }.
A simple observation is that the total quantity sold is decreasing in the sum of prices,
q1 + q2 p1 + p2
= A − (b − a) .
2 2
At price p i firm i has a profit of p i − c i per unit sold, where c i is the cost per item of firm i. Therefore, given the quantity
q i of units sold, the utility of firm i is
u i = qi ( p i − c i ).
2
A
Theorem 6.1. At equilibrium of an ( A , a, b, c 1 , c 2 )-Bertrand competition, firm 1 has price p 1 = 2b−a
+ c 1 4b2b ab
2 −a2 + c 2 4b 2 −a2 , quantity
b 2
ab b(2b −a )
2 2
q1 = 2b−a
A+ c
4b2 −a2 2
− 4b2 −a2
c1 and utility u 1 = q1 ( p 1 − c 1 ). (Firm 2’s price, quantity and utility are analogous.)
We include the proof in Appendix A, for completeness.

We can now present the meta-game, which is the goal of this section. In the meta-game, each firm i selects a meta-cost
c̄ i , rather than its true cost c i . The meta-costs will be used to compute the price and quantity using the equilibrium of an
( A , a, b, c̄ 1 , c̄ 2 )-Bertrand competition (Theorem 6.1). Note that the main difference is that we substitute c̄ i for c i . Naturally,
its true cost remains c i , so its profit at price p i is still p i − c i per unit. This implies that the utility of firm 1 is
u 1 = q1 ( p 1 − c 1 )

b ab2 b(2b2 − a2 ) A 2b2 ab
= A+ c̄ −
2 2
c̄ 1 + c̄ 1 + c̄ 2 − c1 .
2b − a 4b2 − a 4b2 − a2 2b − a 4b2 − a2 4b2 − a2
Theorem 6.2. At the equilibrium of a meta-game ( A , a, b, c 1 , c 2 )-Bertrand competition, firm 1 has meta-cost
2c̄ 1 = α A + (β + γ )c 1 + (β − γ )c 2 ,
2a2 (2b+a) (2b2 −a2 )(4b2 −a2 ) (2b2 −a2 )(4b2 −a2 )
where α = 8b4 −4a2 b2 −a3 b
, β= 8b4 −4a2 b2 −a3 b
and γ= 8b4 −4a2 b2 +a3 b
. (Firm 2’s price and meta-cost are analogous.)
Note that we have that β γ . Also, for A bc 1 we have that α + β + γ α + 2γ 2, which implies that c̄1 c1 .
Proof. We now like to compute the reported meta-cost of firm 1. We need to take the derivative of the utility u 1 w.r.t. the
reported cost c̄ 1 0,
11
While technically this linear function might be negative, we will show that in our setting the quantity at equilibrium is always non-negative, i.e., qi 0.

du 1 2b2 b ab2 b(2b2 − a2 )
= A+ c̄ −
2 2
c̄ 1
dc̄ 1 4b2 − a2 2b − a 4b2 − a 4b2 − a2

b(2b − a )
2 2
A 2b 2
ab
− + c̄ 1 + c̄ 2 − c1 .
4b2 − a2 2b − a 4b2 − a2 4b2 − a2
Setting the derivative to zero we get

0 = 2b b(2b + a) A + ab2 c̄ 2 − b 2b2 − a2 c̄ 1

− 2b2 − a2 (2b + a) A + 2b2 c̄ 1 + abc̄ 2 − 4b2 − a2 c 1

= a2 (2b + a) A + a3 bc̄ 2 − 8b4 − 4a2 b2 c̄ 1 + 2b2 − a2 4b2 − a2 c 1 .
We can get a similar identity for firm 2,

0 = a2 (2b + a) A + a3 bc̄ 1 − 8b4 − 4a2 b2 c̄ 2 + 2b2 − a2 4b2 − a2 c 2 .
Combining them, we get that

2b2 − a2 4b2 − a2 (c 1 − c 2 ) = 8b4 + a3 b − 4a2 b2 (c̄ 1 − c̄ 2 ).
This implies that if the firms have the same true cost they will report the same meta-cost. If the firms have different costs,
the firm with the higher cost will report a higher meta-cost.
For the symmetric case, i.e., c = c 1 = c 2 we have that
(2b2 − a2 )(4b2 − a2 ) a2 (2b + a)

c̄ = c+ 4 A,
8b + a b − 4a b
4 3 2 2 8b + a3 b − 4a2 b2
where c̄ is the identical reported cost.
We now give a closed form solution for the equilibrium of the meta-game. Adding the two identities we get that at
equilibrium,

0 = 2a2 (2b + a) A − 8b4 − 4a2 b2 − a3 b (c̄ 1 + c̄ 2 ) + 2b2 − a2 4b2 − a2 (c 1 + c 2 ),
which implies that
2a2 (2b + a) (2b2 − a2 )(4b2 − a2 )

c̄ 1 + c̄ 2 = A+ (c 1 + c 2 ) = α A + β(c 1 + c 2 ),
8b4 − 4a2 b2 − a3 b 8b4 − 4a2 b2 − a3 b
2a (2b+a)
2
(2b2 −a2 )(4b2 −a2 )
where α= 8b4 −4a2 b2 −a3 b
and β = 8b4 −4a2 b2 −a3 b
. In addition,
(2b2 − a2 )(4b2 − a2 )
c̄ 1 − c̄ 2 = (c 1 − c 2 ) = γ (c 1 − c 2 ),
8b4 − 4a2 b2 + a3 b
(2b2 −a2 )(4b2 −a2 )
where γ= 8b4 −4a2 b2 +a3 b
. Note that β > γ . We conclude that,
2c̄ 1 = α A + (β + γ )c 1 + (β − γ )c 2 . 2
A few remarks on the resulting equilibrium in the meta game. If we set a = 0 (no influence between firms) we get
c̄ i = c i , so, as expected, if the firms do not influence each other, they report their true cost in the meta-game. In the case
that b ≈ a (the influence of both prices is equally influential) we get,
c̄ 1 ≈ 4 A /(3a) + 4c 1 /5 + c 2 /5.
For a = 1 and b = 2 we get α = 1/11, β = 21/22 and γ = 35/38, which gives
1 196 7
c̄ 1 = A+ c1 +
c2 .
22 209
418
Qualitatively, since for A bc i we have that c̄ i c i , this implies that the price goes up in the meta game, matching the
non-quantitative intuition of Fershtman and Judd (1987), Sklivas (1987).
7. Discussion and open problems
We consider the PM/RM game described here a way of thinking about off-equilibria behavior, where one does not
assume that the other participants are stuck in a rut and will not switch strategies. It would be interesting to study similar
delegation games in other settings — for example, resource allocation is one area where a natural delegation game appears.
It may also be interesting to consider “compromises” between traditional equilibrium analysis and analysis of meta-games
such as the PM/RM game, and whether it is possible to approximate equilibria of both types of games simultaneously.
Acknowledgments
We are very grateful to Andrzej (Andy) Skrzypacz who very kindly helped us overcome our ignorance and gave us critical
references regarding strategic delegation.
Appendix A. Omitted lemmas and proofs
Proof of Lemma 3.2. For Claim 1, we can derive p pm from p rm by doing the computation in two stages. In the first stage,
we consider the increase in the price as firm i changes its action from RM to PM while the other firms do not react; in the
second stage, the other firms react to the price change and the price drops. We will argue that the price will stay above the
original level.
In the first stage, after firm i changes from RM to PM, the price increases regardless of whether firm i keeps producing
c
or stops producing. Specifically, if it keeps producing, the price increases by 1+ni , and if it stops producing, the number of
rm
producers decreases by 1, and the price increases by a factor of 1+ nrm
nrm
.
In the second phase, some firms that were not producing at price p rm start producing. This affects the price by increasing
the numerator by the sum of the production costs of the new producers; the denominator increases by the number of
new producers. The crucial observation is that the new producers have production cost at least p rm (since they were not
producing at this price). It follows that the changes in the numerator and the denominator of the price will leave the price
above p rm .
Claim 2 follows directly from Claim 1: Since the price goes up, every firm who produces before the change keeps
producing after the price increase; the only exception may be firm i which changed its strategy to PM, but the premise is
that firm i produces. 2
The next lemma bounds the effect on the price when a firm switches from PM to RM.
Lemma A.1. With the same premises of Lemma 3.2 and the additional assumption that firm i produces at z pm , we have
ci ci
p rm + p pm < prm + .
1 + n pm 1 + nrm

Proof. Let C = y ∈PM( zrm ) c y and C = y ∈PM( z pm ) c y . By the premise of the lemma, i ∈ PM( z pm ), hence c i is one of the
terms in C . The difference C − C − c i is the sum of the production costs of the firms that start producing when i switches
from RM to PM. There are (n pm − nrm ) such firms, which by the previous lemma is non-negative. Since each of these firms
has production cost between p rm and p pm , we have
(n pm − nrm ) prm C − C − c i < (n pm − nrm ) p pm .

According to the definition of p eq ( z) we have:
1+C 1 + C
p rm = ; p pm = .
1 + nrm 1 + n pm
Combining these two equations, we have
p pm (1 + n pm ) − p rm (1 + nrm ) = C − C ,
which implies that
c i + (n pm − nrm ) p rm p pm (1 + n pm ) − p rm (1 + nrm ) < c i + (n pm − nrm ) p pm ,

and the lemma follows. 2
Lemma A.2. With the premises of Lemma A.1 and the extra assumption that c i p rm :
1. If firm i strictly prefers PM to RM, then

1
c i > p rm 1 − 2
.
nrm
2. If firm i strictly prefers RM to PM, then

1
c i < p rm 1 − .
n2pm
1 1
Proof. The utilities of firm i in z pm and zrm are u i ( z pm ) = bi
( p pm − c i )2 and u i (zrm ) = bi
p rm ( p rm − c i ). If firm i strictly
prefers PM to RM, we have u i ( z pm ) > u i ( zrm ), and so
2
2 ci
p rm ( p rm − c i ) < ( p pm − c i ) < p rm + − ci , (A.1)
1 + nrm
where inequality (A.1) follows using Lemma A.1 and the fact that the terms in the right-hand side inside the square are
non-negative; this follows immediately from the premise of the lemma that firm i produces at z pm . By simplifying the last
inequality, we get the first part of the lemma.
The second part is similar. Since firm i strictly prefers RM to PM, we have u i ( z pm ) < u i ( zrm ). Therefore
2
ci
( prm − c i ) prm > ( p pm − c i )2 prm + − ci , (A.2)
1 + n pm
where inequality (A.2) follows using Lemma A.1. Again, the right-hand side terms inside the squares are positive, and
this is guaranteed by the extra assumption that c i p rm . The last inequality is equivalent to the second inequality of the
lemma. 2
Proof of Lemma 3.3. Let p = p eq ( z). If c i > p, then clearly i prefers PM (since it has a negative utility when playing RM).
For the rest of the proof we assume that c i p. Consider joint actions z = ( z− j , PM), z = ( z−i , PM) with market prices p
and p , respectively. The utility of firm j in joint action z is u j ( z) = p ( p − c j )/b j , and the utility of firm j in joint action z
is u j ( z ) = ( p − c j )2 /b j . The utility of firm i in joint action z is u i ( z) = p ( p − c i )/b i and the utility of firm i in joint action
z is u i ( z ) = ( p − c i )2 /b i . By assumption, PM is a best response for j when the joint action is z, so u j ( z) < u j ( z ), i.e.,
2
p ( p − c j )/b j < p − c j /b j , (A.3)
and we wish to show that u i ( z) < u i ( z ), i.e.,

2
p ( p − c i )/b i < p − c i /bi . (A.4)
Let n be the number of firms with non-zero production levels in z.

For fixed p and p , define f (r ) = ( p − r )2 − p ( p − r ). We will argue that f (r ) is a non-decreasing function in the range
r c j.
Note that f (c j ) > 0. The derivative of f is

f (r ) = 2 r − p + p

cj
> 2r − 2 p + +p (A.5)
1+n

r
2r − 2 p + +p (A.6)
1+n
n
= 2r −p
n+1

1 n
> 2p 1 − − p, (A.7)
n2 n+1
where line (A.5) follows by Lemma A.1, line (A.6) follows from the assumption that r c j , and line (A.7) follows by
2
Lemma A.2. This last line may be rewritten as p ( n+ 1
(n − 1/n) − 1), which is greater than or equal to zero for n 2,
as we have here.
This gives us p ( p − c i ) < ( p − c i )2 . Finally, from Observation 3.1 we have p p and hence p ( p − c i ) < ( p − c i )2 , which
completes the proof. 2
Proof of Lemma 3.4. Let p = p eq ( z). We have zi = z j = PM. Consider joint actions z = ( z− j , RM), z = ( z−i , RM) with market
prices p and p . The utility of firm j in the joint action z is u j ( z) = ( p − c j )2 /b j , and the utility of firm j in the joint
action z is u j ( z ) = p ( p − c j )/b j . The utilities of firm i are u i ( z) = ( p − c i )2 /b i and u i ( z ) = p ( p − c i )/b i , respectively.
Note that we can assume firms j and i are both producing at z; if j were not producing (p c j ), neither would i (since
c j < c i ), and thus i would not prefer RM (since by Lemma 3.2, switching to RM would only lower the price even further
below c i ).
Further, note that by Observation 3.1, p p , and we need only prove the lemma in the case where i might have
non-negative utility at RM, namely, where c i p and thus c j < c i p p .
By assumption, i prefers RM, so u i ( z) < u i ( z ). Assume by way of contradiction that firm j weakly prefers PM, i.e.,

( p − c j )2 p p − c j . (A.8)
We will show that in this case firm i would also weakly prefer PM, i.e.,

( p − c i )2 p p − c i . (A.9)
For fixed p and p , define f (r ) = ( p − r )2 − p ( p

− r ). Rearranging Eq. (A.8), we get f (c j ) 0. We will show that f (r ) is a
non-decreasing function in the range r c j . Given that, since f (c j ) 0 and c i > c j , we will conclude ( p − c i )2 p ( p − c i ).
Finally, by appeal to Observation 3.1, we obtain ( p − c i )2 p ( p − c i ), the desired contradiction.
We now show that f is non-decreasing in the desired range. Let n be the number of firms with non-zero production
levels in z . The derivative of f is
f (r ) = 2(r − p ) + p
2c j − 2p + p (A.10)
cj
> 2c j − 2p − 2 + p (A.11)
1 + n

1
= 2c j 1 − − p
1 + n

n
= 2c j − p
1 + n

1 n
> 2p 1 − 2 − p (A.12)
n 1 + n

2n − 2
= p − p
n

2n − 2 − n
= p
n

n − 2
=p
,
n
where line (A.10) follows by the assumption r c j , line (A.11) follows from Lemma A.1, and line (A.12) follows by
Lemma A.2. Therefore f (r ) is a non-decreasing function for n 2, yielding the desired contradiction for n 2. The
case of n = 1 is straightforward, as it corresponds to a z that consists entirely of non-producing players playing PM; in this
case, p = p . 2
Proof of Lemma 3.7. Consider joint actions z (where both firms i and j play RM), ẑ = ( z− j , PM), ž = ( z−i , PM), and let joint
action z̄ differ from z by the actions of both firms i and j, i.e., z̄−i ,− j = z−i ,− j and z̄i = z̄ j = PM. Let p, p̂, p̌ and p̄ be the
corresponding market prices, and let the corresponding numbers of producers be n, n̂, ň and n̄, respectively.
If c j > p, then firm j’s utility u j ( z) < 0, thus it prefers ẑ where its utility is non-negative. For the remainder of the proof
we consider c j p.
We need to show that firm j prefers ẑ to z, assume the contrary (that j weakly prefers z to ẑ). Using Lemma A.2,
this gives c j p (1 − n̂12 ). By the assumption of the lemma, j prefers z̄ to ž. Using Lemma A.2, this gives c j > p̌ (1 − ň12 ).
Combining these together, we have

1 1
p̌ 1 − < cj p 1− 2 . (A.13)
ň2 n̂
1 1
According to Lemma 3.2 we have p̌ > p. For inequality (A.13) to hold, we thus at least need 1 − ň2
<1− n̂2
, and thus
we need ň < n̂.
By definition of z, ẑ, and ž, the set of revenue maximizers in these configurations obeys:
RM( z) = RM(ẑ) ∪ { j } = RM(ž) ∪ {i }.

The number of producing firms in every configuration is the number of revenue maximizing firms, plus the number of
profit maximizers whose production cost is less than the price. Configurations ž and ẑ have the same number of revenue
maximizers, and the same profit maximizers, except that in ž firm i is a profit maximizer and in ẑ firm j is a profit
maximizer. From Observation 3.1 we have p̌ p̂, thus, any profit maximizer that produces in ẑ will also produce in ž
(this is also true for j which is a profit maximizer in ẑ and a revenue maximizer in ž). Thus, the only way one can have
ň < n̂ is if firm i stops producing when switching from RM in z to PM in ž. Thus, n̂ ň + 1. Combining these ideas, we
get
ň < n̂ ň + 1,
which implies that n̂ = ň + 1; therefore PM(ẑ) = PM(ž) ∪ { j }, and hence i does not produce in configuration ž.
Ergo, c i p̌ > p, so i’s utility at z is negative, and she must prefer to be a profit maximizer at ž, a contradic-
tion. 2
Proof of Lemma 3.8. Consider joint actions ẑ = (z̄−i , RM), ž = (z̄− j , RM). Let joint action z differ from z̄ by the actions of
both firms i and j, i.e., z−i ,− j = z̄−i ,− j and zi = z j = RM. Let p, p̂, p̌ and p̄ be the corresponding market prices, and let the
corresponding numbers of producers be n, n̂, ň and n̄, respectively.
Notice that because i prefers z to ž, c i < p, and thus
c j < c i < p < p̂ p̌ < p̄ ,
by assumption of Lemma 3.8 and by applications of Observation 3.1 and Lemma 3.2. Hence, i and j both produce at each
of z, ẑ, ž, and z̄.
Assume by way of contradiction that i prefers z̄ to ẑ. Then by Lemma A.2, c i > p̂ (1 − n̂12 ). By the assumption of
1
Lemma 3.8, i prefers z to ž, so by another application of Lemma A.2, c i < p (1 − ň2
). Combining these, we have

1 1
p̂ 1 − < p 1− 2 .
n̂2 ň
Since p̂ > p, we must have n̂ < ň, meaning that at least one firm with cost at least p̂ and less than p̌ must play PM and
produce at ž, but not at ẑ. We have

1+ k∈PM(ẑ) ck
p̂ = ,
1 + n̂
and

1+ k∈PM(ž) ck 1+ k∈PM(ẑ) ck + k∈{PM(ž)\PM(ẑ)} ck 1+ k∈PM(ẑ) ck + p̌
p̌ = = < ,
1 + ň 1 + n̂ + |{PM(ž) \ PM(ẑ)}| 1 + n̂ + 1
where the inequality comes from the fact that every firm in PM(ž) (meaning, firms producing and playing PM at ž) has cost
less than p̌, and the ratio on the right hand side only decreases in the number of firms in {PM(ž) \ PM(ẑ)}. This gives

(1 + n̂) p̌ < 1 + ck ,
k∈PM(ẑ)
and thus

1+ k∈PM(ẑ) ck
p̌ < = p̂ ,
1 + n̂
a contradiction. 2
Theorem A.3. For two firms in the PM/RM game, there is always at least one pure Nash equilibrium, and the characterization when a
joint action is a pure Nash equilibrium is as follows:
1
(RM, RM) when c 1 < 4
and c 2 < 14 .
√
(PM, RM) when either: 1. c 1 ∈ [ 14 , 12 ] and c 2 < 1+4c1 , or 2. c 1 < 32 1 − 2c 2 − (1 − 2c 2 ) and c 2 12 .
√
(RM, PM) when either: 1. c 1 < 1+4c2 and c 2 ∈ [ 14 , 12 ], or 2. c 1 1
2
and c 2 < 3
2
1 − 2c 1 − (1 − 2c 1 ).
(PM, PM) when one of the cases below holds:
1. c 1 ∈ [ 1+4c2 , 12 ] and c 2 ∈ [ 1+4c1 , 12 ].
√
2. c 1 1
2 √
and c 2 ∈ [ 32 1 − 2c 1 − (1 − 2c 1 ), 1+2c1 ].
3. c 1 ∈ [ 32 1 − 2c 2 − (1 − 2c 2 ), 1+2c2 ] and c 2 12 .
4. c 1 > 12 , and c 2 > 12 .
Proof. We consider four cases, depending on the firms’ costs, compared to 1/2.
Case 1 ({c 1 1/2 and c 2 1/2}). This is the most interesting case, in which the two firms are producing, as we will see
later. We first define the price as a function of the action of the firms.12
RM PM
1 1+c 2
RM 3 3
1 +c 1 1+c 1 +c 2
PM 3 3
Next we derive the production level xi ( z), at each joint action.
RM PM
1+c2
RM 1
3b
, 3b1 3b
, 1−
3b
2c 2
1 −2c11 1+2 c1 1+c2 −2c1 1 1+c21 −2c2

PM 3b1
, 3b2 3b1
, 3b2
Each entry of the production level matrix is non-negative for c i 1/2. Therefore, firm i that plays PM will produce. For
two firms we have the following payoff matrix:
RM ⎛ PM ⎞

1 +c 2

1 1
− c1 , − c 1 1+c2 ,
RM 3b1 3
1 1
⎝ 3
1−2c2 2 13b1 ⎠
3b2 3
− c2 3 b2

1 −2c1 2

1 1+c 2 −2c 1 2 1
3
, ,
PM 1+c1 b11+c1 3
1+c1 −2c2 2b11
3
− c 2 3b2 3 b2
To compute the pure Nash equilibria, we compute the preference of the firms. We start with firm 1.
• Firm 1 prefers (PM, RM) to (RM, RM) when ( − c 1 ) < ( 1−32c1 )2 b11 , which
1 1
3b1 3
holds for c 1 > 14 .
• Firm 1 prefers (PM, PM) to (RM, PM) when ( 1+3c2 − c 1 ) 13b
+c 2
< ( 1+c23−2c1 )2 b11 , which holds for c 1 > 1+4c2 .
1
1
• Firm 2 prefers (RM, PM) to (RM, RM) for c 2 > 4 .
• Firm 2 prefers (PM, PM) to (RM, PM) for c 2 > 1+4c1 .
The conditions for each joint actions to be a pure Nash equilibrium, are as follows:
RM PM
RM c1 < 1
4
, c2 < 1
4
c 1 < 1+4c2 , c 2 > 1
4
1+c 1
PM c 1 > 1
4
, c2 < 4
c 1 > 1+4c2 , c 2 > 1+c 1
4
1
Case 2 ({c 1 > 2
and c 2 > 12 }). If both firms select RM then the price is 1/3 and they both have negative utility. Assume firm
1 selects RM. If firm 2 selects PM then the price is p = 1+3c2 . Since c 2 > 1/2, then c 2 > p, and firm 2 is not producing. If the
firm 2 selects PM and is not producing then the price is 1/2 < c 1 , thus firm 1 has negative utility. Therefore, in this case
both firms have PM as a dominating action.
Case 3 ({c 1 12 and c 2 > 12 }). If firm 2 selects RM, then the price is at most 1+3c1 12 . Therefore, in this case, action PM will
be a strictly dominating action for firm 2.
Consider joint action (RM, PM). We have production level x2 = 1+3c2 − c 2 < 0, thus firm 2 is not producing. For joint
action (PM, PM), firm 1 always produces (production level x1 min { 1+c31 +c2 , 1+2c1 } − c 1 > 0), so we have firm 2 produces if
p = 1+c31 +c2 > c 2 .
In the case that c 2 > 1+2c1 firm 1 produces alone, so her dominating action is PM, and her utility is ( 1−2c1 )2 b1 .
1
For c 2 < 1+2c1 , since firm 2 dominating action is PM, the payoff values are:

1 1
u 1 (RM, PM) = − c1 ;
2 2b1
2
1 + c 2 − 2c 1 1
u 2 (PM, PM) = .
3 b1
12
In all matrices that we use, row firm is firm 1 and column firm is firm 2.
Fig. 3. Characterization of two firms’ pure Nash equilibria.
Firm 1 will prefer action PM if

3 1 + c1
c2 ∈ 1 − 2c 1 − (1 − 2c 1 ), ;
2 2
otherwise, when
3
c2 < 1 − 2c 1 − (1 − 2c 1 ),
2
firm 1 prefers RM.
1 1
Case 4 ({c 1 > and c 2 }). Similar to the previous case, firm 1 will select PM. Firm 2 will prefer action PM if c 1 ∈
√ 2 2
1+c 2 √
[ 32 1 − 2c 2 − (1 − 2c 2 ), 2
]; otherwise, when c 1 < 3
2
1 − 2c 2 − (1 − 2c 2 ), firm 2 prefers RM. 2
In each of the four regions, we showed that for any value of the production cost, there exists a pure Nash equilibrium.
(For some values there exist two pure Nash equilibria; see Example 3.11.) A diagram of the pure Nash equilibria appears in
Fig. 3.
Proof of Theorem 6.1. We solve for the equilibrium in the asymmetric Bertrand competition by maximizing the utility w.r.t.
the price. We first take the derivative w.r.t. p i ,
du i dqi
= qi + ( p i − c i ) = qi − b( p i − c i ) = A − 2bp i + ap 3−i + bc i .
dp i dp i
At equilibrium we have
A + ap 3−i ci
pi = + ,
2b 2
and for A bc i we have p i c i . Summing both prices we have,
2 A + a( p 1 + p 2 ) c1 + c2
p1 + p2 = + ,
2b 2
which implies that
2 A + b(c 1 + c 2 )
p1 + p2 = .
2b − a
In addition,
c1 − c2 a( p 1 − p 2 )
p1 − p2 = − ,
2 2b
which implies that
b(c 1 − c 2 )
p1 − p2 = .
2b + a
Therefore, at equilibrium we have
2A 4b 2a
2p 1 = + bc 1 + bc 2 .
2b − a 4b2 − a2 4b2 − a2
Hence,
A 2b2 ab
p1 = + c1 + c2
2b − a 4b2 − a2 4b2 − a2
and
A 2b2 ab
p2 = + c2 + c1 .
2b − a 4b2 − a2 4b2 − a2
Next we compute get the quantities produced at equilibrium:
q1 = A − bp 1 + ap 2

A 2b2 ab A 2b2 ab
= A −b + c1 + c2 +a + c2 + c1
2b − a 4b2 − a2 4b2 − a2 2b − a 4b2 − a2 4b2 − a2
b ab 2
b(2b − a )
2 2
= A+ c2 − c1 .
2b − a 4b2 − a2 4b2 − a2
Note that for A bc 1 we have that q1 0. The utility is at equilibrium is,
u 1 = q1 ( p 1 − c 1 )

b ab2 b(2b2 − a2 ) A 2b a
= A+ c −
2 2
c1 + bc 1 + bc 2 − c1 . 2
2b − a 4b2 − a 4b2 − a2 2b − a 4b2 − a2 4b2 − a2
References
Abreu, D., 1986. Extremal equilibria of oligopolistic supergames. J. Econ. Theory 39, 191–225.
Arrow, K.J., 1960. The work of Ragnar Frisch, econometrician. Econometrica 28, 175–192.
Berr, F., 2011. Stackelberg equilibria in managerial delegation games. Eur. J. Oper. Res. 212, 251–262.
Even-Dar, E., Mansour, Y., Nadav, U., 2009. On the convergence of regret minimization dynamics in concave games. In: STOC, pp. 523–532.
Fershtman, C., Judd, K.L., 1987. Equilibrium incentives in oligopoly. Amer. Econ. Rev. 77, 927–940.
Fershtman, C., Judd, K., Kalai, E., 1991. Observable contracts: Strategic delegation and cooperation. Int. Econ. Rev. 32, 551–559.
Kurz, M., 1977. Distortion of preferences, income distribution, and the case for a linear income tax. J. Econ. Theory 14, 291–298.
Riordan, M.H., 1985. Imperfect information and dynamic conjectural variations. RAND J. Econ. 16, 41–50.
Schelling, T.C., 1960. The Strategy of Conflict. Harvard University Press.
Sklivas, S.D., 1987. The strategic choice of managerial incentives. RAND J. Econ. 18, 452–458.
Theocharis, R.D., 1960. On the stability of the Cournot solution on the oligopoly problem. Rev. Econ. Stud. 27 (2), 133–134.
Vickers, J., 1985. Delegation and the theory of the firm. Econ. J. 95 (Suppl.), 138–147.

Games and Economic Behavior: Amos Fiat, Elias Koutsoupias, Katrina Ligett, Yishay Mansour, Svetlana Olonetsky

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Games and Economic Behavior: Amos Fiat, Elias Koutsoupias, Katrina Ligett, Yishay Mansour, Svetlana Olonetsky

Uploaded by

Copyright:

Available Formats

JID:YGAME AID:2252 /FLA [m3G; v 1.121; Prn:8/01/2014; 14:33] P.

Contents lists available at ScienceDirect

Games and Economic Behavior

Beyond myopic best response (in Cournot competition) ✩

1.1. Prior related work

1.2. Summary of results

2. The Cournot model

2.1. Standard linear Cournot competition

1. For any ﬁrm i ∈ N (with strictly positive production levels), we have

2.2. The PM/RM game

1. PM (proﬁt maximization), and

3. Nash equilibria and dynamics of the Cournot PM/RM game

3.1. Existence and properties of Nash equilibria

3.2. Fast convergence of best response dynamics

1. If minPM( z) ∈ BRF( z), ﬁrm minPM( z) changes its strategy.

In the proof of the following lemma, we show that if BRF( z)

This is possible only under two exceptional conditions:

4. Instances of special interest for the Cournot PM/RM game

5. Comparison of the PM/RM game and Cournot competition

The utility of each ﬁrm in the PM/RM game is

We include the proof in Appendix A, for completeness.

(2b2 − a2 )(4b2 − a2 ) a2 (2b + a)

2a2 (2b + a) (2b2 − a2 )(4b2 − a2 )

7. Discussion and open problems

Appendix A. Omitted lemmas and proofs

(n pm − nrm ) prm C  − C − c i < (n pm − nrm ) p pm .

c i + (n pm − nrm ) p rm p pm (1 + n pm ) − p rm (1 + nrm ) < c i + (n pm − nrm ) p pm ,

1. If ﬁrm i strictly prefers PM to RM, then

and we wish to show that u i ( z) < u i ( z ), i.e.,

Let n be the number of ﬁrms with non-zero production levels in z.

For ﬁxed p and p  , deﬁne f (r ) = ( p − r )2 − p  ( p 

RM( z) = RM(ẑ) ∪ { j } = RM(ž) ∪ {i }.

c j < c i < p < p̂ p̌ < p̄ ,

Next we derive the production level xi ( z), at each joint action.

1 −2c11 1+2 c1 1+c2 −2c1 1 1+c21 −2c2

Fig. 3. Characterization of two ﬁrms’ pure Nash equilibria.

Firm 1 will prefer action PM if

Therefore, at equilibrium we have

You might also like

(n pm − nrm ) prm C − C − c i < (n pm − nrm ) p pm .

and we wish to show that u i ( z) < u i ( z ), i.e.,

For ﬁxed p and p , deﬁne f (r ) = ( p − r )2 − p ( p