1 Problem Set I Cournot.

1. (EXCESS ENTRY) (Vives 4.12) Consider a Cournot market with linear

demand P = a bQ; a > 0; b > 0 and zero marginal costs for all the …rms.
Find the free-entry equilibrium if there is a …xed cost of entry F . Compare
the free-entry equilibrium with the optimal number of …rms to maximize social
welfare.
2. (UNICITY) Prove that equilibrium is unique in a Cournot market whose
demand is given by P (Q) and the cost function of …rms is given by Ci (qi )
i = 1:::n: if the following
_
conditions hold: _
- P 0 (Q) < 0 and Q exists such that for Q Q we have that P (Q) = 0
- P (Q) is log-concave. (A function f is log-concave (log-convex) if log f is
concave (convex)).
- Ci "(qi ) P 0 (Q) > 0 for all i.
(Hint: check …rst of all that the pro…t function of a …rm is quasi-concave on
its output).
3. (Corchón (1994)) Assume that market demand is given by P = a bQ
and cost of …rms by C(qi ) = cqi + d2 qi2 with a > c; d < 0; d + 2b > 0 and
a c c
d + b < 0. (Assume that < to avoid negative costs). Show that total
2b + d d
output decreases with entry.
4. Suppose that costs are zero and that market demand is given by:
aQ
P (Q) = 100E

Show that the reaction function of …rms have zero slope and …nd the equilibrium
with n …rms. Check that demand is log-convex and log-concave.
5. (Martin (2001) p. 351) Assume we have n …rms competing à la Cournot
with demand given by P (Q) satisfying:

P 0 (Q) + QP 00 (Q) < 0

Assume that all …rms have zero costs except …rm 1 that has unit cost c(k), where
c0 (k) < 0: Obtain the marginal change in the equilibrium outputs of …rms due
to a marginal increase in k.
6. (Moulin (1986) p.118) Suppose that n …rms compete à la Cournot in a
market whose concave demand is given by P (Q). Each …rms has an increasing
and convex cost function ci (qi ). Show that in the Cournot equilibrium industry
pro…ts are not maximized.
7. Suppose we have two symmetric …rms competing à la Cournot in a market
with demand P = 1 Q. Production costs of …rms are given by C(q) = F if
q > 0 and C(0) = 0. Find the Cournot equilibria depending on the value of F .
8. Assume that market demand P (X) is concave and strictly decreasing. We
have n …rms, denoted by a natural number from 1 to n. Firm i’s cost function
Ci (xi ) is strictly convex and Ci0 (0) = 0. Assume that in the unique Cournot
equilibrium and in the allocation that maximizes social welfare all …rms produce
a positive quantity. Then show that the quantity sold at the (unique) Cournot

1
equilibrium is lower than the quantity sold in the allocation that maximizes
social welfare.

2 Solution to Problem Set I:

r
FE 1
1. the number of …rms with free entry n =a 1.
bF
a2 1=3
the number of …rms that maximizes social welfare: n = ( ) 1
bF
2 2
a a
If a monopoly is pro…table ( > F ), then > 1 and therefore nF E > n .
4b bF
2. The fact that P (Q) is log-concave implies that:

P "P (P 0 )2 0 (1)

We assume furthermore that:

C" P 0 > 0: (2)

The FOC of …rm i is given by:
@ i
= P (Q) + qi P 0 (Q) Ci0 (qi ) = 0 (3)
@xi
Second order conditions are satis…ed, because whenever (7) holds, the func-
tion is strictly concave, which implies that i is strictly quasi-concave with
respect to qi . To see this we subdivide the Second Order Condition in two
parts:
@2 i
= [P 0 (Q) + qi P "(Q)] + P 0 (Q) Ci" (qi )
(@xi )2
The second part is negative. We are going to see that the second part is not
positive. If the function is concave we are done. Then we have to check it for
the case of convex demands. From (3) we have that:
P Ci0
qi = :
P0
Then;
P Ci0
P 0 (Q) + qi P "(Q) = P 0 (Q) + ( )P "(Q) =
P0
1 1
( (P 0 )2 + (P Ci0 )P ") < ( (P 0 )2 + P P ") 0
P0 P0
The …rst inequality cames from increasing costs and demand convex and the
second from (1). Quasi-concavity of pro…ts imply that a Nash equilibrium exists
(Theorem II.1 in Vives).
The fact that i is strictly quasi-concave with respect to qi , implies that there
is a unique pro…t-maximizing strategy of …rm i, de…ned by (3) and denoted by

2
ri (q i ). In other words, the reaction functions are well-de…ned. The proof of
unicity comes from the fact that the slope of the reaction function of …rms is
negative and greater than -1. Applying the implicit function theorem to (3) we
have that:
P 0 (Q) + qi P "(Q)
ri0 (q i ) =
2P 0 (Q) + qi P "(Q) Ci "(qi )
We have checked that both the numerator and the denominator are negative
and given (2), the denominator is bigger in absolute value than the numerator.
Then 1 < ri0 (q i ) < 0.
De…ne the cumulative reaction function i (Q) from:

qi = ri (Q qi )
0 ri0 (x i )
i (Q) = <0
1 + ri0 (x i )
_
i (Q)is de…ned over [ri (0); P
Q]. Then the equilibrium output will be given by
n
the …xed point of (Q) = i=1 i (Q) and the individual outputs by i (Q).
_
(Q) is de…ned over [maxi ri (0); Q]. Existence and unicity follow from the fact
that (Q)_
is continuous and decreasing and that (maxi ri (0)) maxi ri (0)
and (Q) = 0:
n(a c) @Q
3. Total output in equilibrium is given by: Q = : =
(n + 1)b + d @n
(a c)(b + d)
< 0.
((n + 1)b + d)2
1 1
4. Ri (q i ) = . In equilibrium each …rm produces
a a
log(P (Q)) = aQ is a linear function. Then, P (Q) is log-concave and log-
convex.
5. The Cournot equilibrium is given by the solution of the system of the
F.O.C.

P (Q) + q1 P 0 (Q) c(k) = 0 (4)

P (Q) + qi P 0 (Q) = 0 i = 2; :::n

Taking the the total di¤erential of (4) we have:

dq1 = 1 dQ + 1 dk (5)
dqi = i dQ
i = 2:::n
P 0 (Q) + qi P 00 (Q) c0 (k)
i = and 1 =
P 0 (Q) P 0 (Q)

3
Adding (5) for all i we have:

dQ = dQ + 1 dk (6)
dQ 1
=
dk 1+
n
X
where = i
i=1

Using (5) and (6) we have:

dq1 1
= 1 (1 )
dk 1+
dqi i 1
=
dk 1+
6. The Cournot equilibrium is determined by the …rst-order conditions:
0
qi P 0 (Q ) + P (Q ) = ci (qi )
i = 1:::n

Joint pro…t is given by:

n
X
(q1; :::qn ) = QP (Q) ci (qi )
i=1

@
(q1; :::qn ) = P (Q) + QP 0 (Q) c0i (qi )
@qi
Evaluated at the Cournot equilibrium we have:
@
(q ; :::qn ) = (Q qi )P 0 (Q ) < 0
@qi 1
So reducing any one of the individual productions increases joint pro…t.
7. Call one …rm, …rm 1 and the other …rm 2.
1 qj p
The reaction function of …rms is ri (qj ) = if qj 1 2 F and ri (qj ) =
2
0 otherwise. (i; j = 1; 2 i 6= j). We have a Cournot equilibria wherever reaction
functions cross. Then we distinguish di¤erent cases. They can be understood
taking into account the following. In the standard Cournot equilibrium (F = 0),
1 1
pro…ts of …rms are . It will still be an equilibrium if F . The other
9 9
1
equilibrium that may arise is that one …rm produces the monopoly output ( )
2
1
and the other …rm does not want to produce. This will be the case if F,
16
because
M ax 1 1
q (1 q )q =
2 16

4
 1
. Finally when the …xed costs are greater that the monopoly pro…ts , no …rm
4
wants to produce.
p 1 1
Case 1: If 1 2 F > or 0 F < :
2 16
1
Reaction functions cross only once at q1 = q2 = .
3
1 p 1 1 1
Case 2: If 1 2 F or F .
3 2 16 9
1 1
Reaction functions cross three times at (q1 = q2 = ), (q1 = ; q2 = 0) and
3 2
1
(q1 = 0; q2 = ).
2
p 1 1 1
Case 3: 0 1 2 F < or < F :
3 9 4
1 1
Reaction functions cross twice at (q1 = ; q2 = 0) and (q1 = 0; q2 = ).
2 2
p 1
Case 4: 1 2 F < 0 or F > .
4
Reaction functions cross at (q1 = 0; q2 = 0): Pn
8. Let (x1 ::::xn ) be the Cournot equilibrium where x = i=1 xi .
It satis…es the F.O.C. of pro…t maximization:

P (x ) + P 0 (x )xi Ci0 (xi ) = 0

o o o
PnLet o(x1 ::::xn ) be the allocation that maximizes social welfare where x =
i=1 xi . They satisfy the F.O.C. of maximizing social welfare given by
Z Pn
xi n
X
i=1
P (x)dx Ci (xi )
0 i=1

P (xo ) Ci0 (xoi ) = 0

As demand is downward sloping we have that

P (x ) Ci0 (xi ) > P (xo ) Ci0 (xoi )

Ci0 (xoi ) Ci0 (xi ) > P (xo ) P (x ) (7)

o
Assume that x x . Then (7) implies

Ci0 (xoi ) Ci0 (xi ) > 0

Given that marginal cost is strictly increasing we have that xoi > xi what is a
contradiction with x xo . Then we must have that x < xo .

5
3 Problem Set II Prices.
1. Assume that demand is given by D(p) = a bP and …rms are symmetric with
cost function given by C(q) = 2c q 2 . Find the Bertrand equilibria as a function
of the number of …rms n.
2. (Tirole (1988) p. 319) Obtain the equilibrium of the game where two
…rms choose in a …rst stage capacity and compete à la Bertrand-Edgeworth in
the second stage when demand is given by p = 4 Q and capacity costs 3
per unit. Assume that demand is rationed according to the surplus-maximizing
rule.
3. Vives 5.3/5.4/5.7

6
4 Solution to Problem Set II
1. In class, we checked that no asymmetric equilibrium exists. Then in equilib-
rium all …rms set the same P . It will be an equilibrium if no …rm has incentive
a
to deviate by setting a higher or a lower price. ( P b can not be an equilib-
rium, because we have no sales and …rms have incentives to deviate and set the
monopoly price. Then from now on we only consider P < ab .)
Observe that, given that all …rms set P , no …rms wants to set a higher price
if the following condition holds:
a bP c a bP 2
( )P ( ) 0
n 2 n
a bP c a bP
( ) P ( ) 0
n 2 n

c a bP
P ( ) 0
2 n
ac
P
2n + bc
Observe that, given that all …rms set P , no …rm wants to undercut if the
following condition holds:
a bP c a bP 2 c
( )P ( ) (a bP )P (a bP )2
n 2 n 2
P c c
(a bP ) P (a bP )
n 2n2 2
ac(1 + n 1 )
P
2 + bc(1 + n 1 )

ac ac(1 + n 1 )
Every …rm charging the same price P is an equilibrium if P 2 [ ; ].
2n + bc 2 + bc(1 + n 1 )
2. Monopoly pro…ts are 4. Then no …rm will choose ki such that 4 3ki > 0;
4 4
ki > . But if …rms choose always ki , we will have an equilibrium in pure
3 3
4
strategies in the second stage, because is the Cournot equilibrium with zero
3
costs and we are considering the surplus maximizing rationing rule. In those
equilibria the pro…ts are given by (4 k1 k2 )ki . The payo¤s in the second
stage will be given by (4 k1 k2 )ki 3ki yielding the equilibrium choice of
1
capacity k1 = k2 = .
3
3. (Vives 5.3)
We prove it directly for the asymmetric case. We check that every …rm
charging the competitive price wn and every …rm selling its competitive supply
at
Pnthis price Sj (wn ) is an equilibrium. Recall that by de…nition D(wn ) =
j=1 Sj (wn ): at the competitive price demand equals supply. We are going to

7
check that no …rm has incentives to deviate either by raising or lowering the
price.
a) If it raises the price, it sells nothing and it earns zero pro…ts while in the
competitive equilibrium obtains nonnegative pro…ts.
b) If it lowers the price he will serve the whole market and its pro…ts will
be:
pD(p) C(D(p))
But then as wn > p we have that:

pD(p) C(D(p)) < wn D(p) C(D(p))

But as Sj (wn ) = arg maxq wn q C(q) we have:

pD(p) C(D(p)) < wn D(p) C(D(p)) wn Sj (wn ) C(Sj (wn ))

This shows that by lowering the price, it obtains less pro…ts.

4. (Vives 5.4)
kj P
To prove that DiP (p) DiSM (p), we have to check that D(pi ) j2J
P D(pj )
j2J kj . But given that demand is downward sloping and pj < pi we have that
P kj P kj P
D(pi ) j2J D(pi ) j2J = j2J kj .
D(pj ) D(pi )
5. (Vives 5.7)
Region with pure-strategy equilibrium.
A) Take P (Q) = 1 Q to be the inverse demand. The only candidate
equilibrium in pure strategies is the competitive price P (nk). An equilibrium
at zero price exists if rivals can serve the whole market at zero price.
1
P ((n 1)k) = 0; (n 1)k D(0); k
n 1
B) The following condition guarantees that …rms do not want to deviate to
set a higher price than P (nk).

P (r((n 1)k) + (n 1)k) P (nk); r((n 1)k) + (n 1)k nk

(n 1)k 1 1
r((n 1)k) k; k; k
2 n+1
Region without pure-strategy equilibrium:
1 1
>k>
n 1 n+1
1
We obtain the mixed-strategy equilibrium by construction when k = . As-
n
sume that a_ symmteric equilibrium exists in which …rms randomize over the
support [p ; p] according to the (atomless) distribution function F (). As …rms

8
can guarantee themselves positive expected pro…ts, p = 0 can not belong to
the support and therefore not all capacity can _be sold because D(p) < 1. Fur-
thermore, even when setting the highest price p, sales should be positive, then
_ n 1
D( p) > 0: Both facts imply that for all realizations of prices all …rms sell
n
to capacity except the …rm that charges the highest price. Then the expected
pro…ts are given by:
n 1 p
Fn 1
(1 p )p + (1 Fn 1
)( )
n n
_
If a …rm charges p = p, it will be the one that charges the highest price and
then its expected payo¤ is
n 1
(1 p )p
n
_
p is the price that maximizes this expression:
1 _ n 1
= p = arg max(1 p )p
2n n
Then the expected pro…ts are given by:
_ 1
=
4n2
Then the p should satisfy
_ 1
p=n =
4n
The distribution function satis…es: :
1 p 1
Fn 1
( p)p + (1 Fn 1
)( ) =
n n 4n2
1 1
F (p)n 1
=
pn 4n2 p2
_
F (p ) = 0; F ( p) = 1

9
5 Problem Set III Two-stage games.
1. (STRATEGIC TRADE POLICY) Assume that n …rms of country A compete
with one …rm of country B in a foreign market whose demand is given by
P = a bQ. All …rms have the same unit cost c, where a > c. Obtain the
subgame perfect equilibrium of the two-stage game in which in the …rst stage
the Government of country A sets a unit export subsidy s that reduces the unit
cost of …rms of country A to c s. In the second stage, …rms compete à la
Cournot in the foreign market. The objective of the Government of country
A is to maximize the pro…ts of their national …rms net of the subsidy. Is the
equilibrium export subsidy always positive ? Explain.
2. (STACKELBERG) Assume that two …rms (…rm 1 and 2) compete choos-
ing prices in a di¤erentiated goods market. The demand of the good that …rm
i sells is given by qi = 1 pi + cpj where 0 < c < 1 (goods are substitutes) and
pi and pj are the prices charged by …rm i and j respectively (i; j = 1; 2). There
are no costs.
(a) Find the the equilibrium of Bertrand.
(b) Find the the subgame perfect equilibrium of the two-stage game where
in the …rst stage …rm 1(leader) chooses its price and …rm 2 (follower) chooses
its price in the second stage.
3. (DIVISIONALIZATION) Assume that we have two …rms. Demand is
given by P = Q and marginal cost by c. Find the equilibrium of the following
two-stage game. In the …rst stage, …rms decide the number of (independent)
divisions, given that the cost per divisions is F . In the second stage, divisions
compete a la Cournot (each division maximizes its individual pro…ts). Assume
that …rms when setting the number of divisions maximize the joint pro…ts of
their divisions, for simplicity, treat the number of divisions as a continuous
variable and take advantage of the fact that equilibrium is symmetric.(See Baye,
M., K. Crocker, and J. Ju, 1996, Divisionalization, Franchising and Divestiture
Incentives in Oligopoly, American Economic Review, 86, 223-236).
4. (Vives 6.2) (STRATEGIC MANAGERIAL INCENTIVES) Consider a
Cournot duopoly with linear demand p = a Q with a > 0. Firms have constant
marginal cost c. Suppose that each owner gives its manager an incentive contract
which is a linear combination of pro…ts ( i ) and sales (Si ):

i i + (1 i )Si

Study the two-stage game in which …rst owners choose i and then managers,
upon having observed the contracts o¤ered to both of them, compete à la
Cournot. Assume that managers are paid their (…xed) opportunity cost. There-
fore when choosing i owners want to maximize the pro…ts of their …rms. (See
Fershtman and Judd (1987) ”Equilibrium Incentives in Oligopoly” The Ameri-
can economic review, vol. 77 no 5 pp. 927-940).
5. (ENTRY DETERRENCE) Find the Subgame Perfect Equilibrium of the
following two-stage game. In the …rst stage, Firm 1 (leader) chooses its output
while Firm 2 (follower) chooses its output in Stage 2. Variable costs are zero

10
but Firm 2 faces a …xed cost of F when producing. The usual interpretation is
that Firm 1 is the incumbent and Firm 2 a potential entrant. Market demand
is given by P = 1 Q.
6. (ENTRY DETERRENCE) Find the Subgame Perfect Equilibrium of the
following two-stage game. In the …rst stage, Firm 1 (leader) chooses its price
while Firm 2 (follower) chooses its price in Stage 2. Variable costs are zero but
1
Firm 2 faces a …xed cost of F < when producing. The usual interpretation is
4
that Firm 1 is the incumbent and Firm 2 a potential entrant. Market demand
is given by D(P ) = 1 P . Assume that when both …rms set the same price all
the demand goes to the follower.
7. Assume that Firm 1 and Firm 2 compete in a homogenous good market
whose demand is given by Q = 1 P , where P is price and Q the quantity.
Unit cost of Firm 2 is given by c and the unit cost of …rm 1 by c x1 , where
x1 is his R+D investment. The R+D investment costs C(x1 ) = x21 . Find the
Subgame Perfect Nash equilibrium of the following two-stage game. Assume
that 1 > c > 34 .
In the …rst stage, Firm 1 chooses the level of its R+D investment x1 .
In the second stage, Firm 1 and 2 compete in prices. Assume that at equal
prices all demand goes to Firm 1.
1
8. Consider a Cournot duopoly with demand P = . Firms have constant
Q
marginal cost equal to 1. Suppose that each owner gives its manager an incentive
contract which is a linear combination of pro…ts ( i ) and sales (Si ):

i i + (1 i )Si
where i 2 (0; 2)

Find the Subgame Perfect Nash Equilibrium of the two-stage game in which
…rst owners choose i and then managers, upon having observed the contracts
o¤ered to both of them, compete à la Cournot. Assume that managers are
paid their (…xed) opportunity cost. Therefore when choosing i owners want to
maximize the pro…ts of their …rms.
9. (Huck et al (2001)) Assume we have m + n …rms competing in a market
whose demand is given by P = Q. All …rms have (constant) unit cost c:
Find the Subgame Perfect Nash Equilibrium of the following two-stage game.
In the …rst stage, m …rms (leaders) decide their output. In the second stage,
the remaining n …rms (followers) decide their output. Market price is the one
that equates demand with production of …rms. Study the pro…tability of the
merger of two leaders and the pro…tability of the merger of two followers.

11
6 Solution to Problem Set III
1. The output sold by the …rms of country A in the equilibrium of the second
a c + 2s a c ns
stage is q = and the price margin p c = . Then in the
b(n + 2) n+2
…rst stage the Government of country A maximizes n(p c)q. It is maximized
(a c)(n 2)
when s = . The same result can be obtained by equating the
4n
production of the national …rms to the output of the Stackelberg leader:
a c
= nq
2b
1
2. (a) p1 = p2 = .
2 c
2+c 4 + 2c c2
(b)p1 = 2
and p2 = .
4 2c 4(2
! c2 )
(1=3)
( c)2
3. n = 1 =2
F
4. The quantity sold in the second stage by the manager of …rm i is given
a 2 ic + j c a + ( i 3)c + j c
by: qi = and the price margin by p c = .
3 3
When choosing i , the owner of …rm i maximizes (p c)qi . The FOC is given
then by:
2c(a + ( i 3)c + j c) + c(a 2 i c + j c) = 0
Imposing symmtery i = j = , we obtain the equilibrium:
a c
=1
5
1 q1
5. Given an output of …rm q1 , Firm 2 decides to produce r2 (q1 ) =
2
whenever it yields greater pro…ts than not producing:
2
1 q1
F 0
2
p
q1 1 2 F
1
In the …rst stage, if F 16 , the optimal decision of Firm 1 is easy to derive, it
produces the monopoly
q output and Firm 2 does not enter in the second stage
1 1
because 1 2 16 = 2. In this situation entry is said to be blockaded. For
1
F < 16 , the payo¤ of Firm 1 is given by:
p
(1 q1 r2 (q1 ))q1 if q1 1 2 F (8)

(1 q1 )q1 otherwise (9)

12
 In (8) the optimal decision is q1 = 12 and pro…ts obtained are 81 . In (9) the
optimal output is the lowest output that inducespthat Firm p 2 will not produce
in the second stage. This means a pro…t of: 2 F (1 2 F ). To derive the
optimal output one should obtain the values of F such that:
1 p p
2 F (1 2 F) 0
8 p
3 2 2
F
32
p p
3 2 2 1 3 2 2 1
Then for F ; …rm 1 will produce q1 = 2 and for < F < 16
32 p 32
it will produce q1 = 1 2 F . In this last case, entry is said to be deterred.
6. We solve the model as usual by backward induction.
Second stage.
Firm 1 has set a price equal to p1 in the …rst stage. Then we derive the
optimal strategy of Firm 2.
1 1 1
If p1 2 , then p2 = 2 . Observe that 2 is the monopoly price.
1
If p1 < 2 , then p2 = p1 if p1 (1 p1 ) > F and p2 = 1, otherwise.
First stage.p
1 1 4F
p = < 12 is the lowest price that satis…es p(1 p) = F .
2
If Firm 1 sets p1 > p , he will be undercut in the second stage by Firm 2
and therefore he will obtain zero pro…ts.
If Firm 1 sets p1 p he will obtain pro…ts of p1 (1 p1 ): This function is
strictly concave maximized in p1 = 12 . As p < 12 Firm 1 will optimally choose
p1 = p .
7. We solve the model by backward induction.
In the second stage we have Betrand competition with asymmetric costs
where Firm 1 has unit cost c x1 and Firm 2 has cost c. If Firm 1 were a
1 + c x1
monopolist, it would set pm = .
m
2
Then if p c (x1 1 c) the equilibrium is p1 = pm and p2 pm . Firm
2
1 c + x1
1 obtains pro…ts m 1 = .
2
m
If p > c, the equilibrium is given by p1 = p2 = c and Firm 1 obtains
1 = x1 (1 c):
In the …rst stage the payo¤ of Firm 1 as a function of its level of investment
is given by:

1 x21 if x1 < 1 c
m
1 x21 if x1 1 c

It is easy to verify that m

1 x21 0 if x1 1 c. Then the optimal investment
will be obtained from maximizing 1 x21 .

13
 @( 1 x21 )
= 1 c 2x1 = 0
@x1
1 c
x1 =
2
8. In the second stage the manager of …rm i maximizes:

1 qi
Oi = i ( 1)qi + (1 i) =
q1 + q2 q1 + q2
qi
= i qi
q1 + q2
@Oi qj
= i =0 (10)
@qi (q1 + q2 )2
@O1 @O2 q1 + q2
+ = 1 2 =0
@q1 @q2 (q1 + q2 )2
1
q1 + q2 = (11)
1 + 2

Using (11) in (10) we obtain the second stage equilibria:

j
qi = 2
( 1+ 2)

Then the pro…ts of …rm i as a function of the incentive parameters is:

!
j
i = 2 ( 1 + 2 1)
( 1 + 2)

@ i 1 2 +2
= 3 =0 (12)
@ i ( 1 + 2)

@2 i
It is easy to check that < 0 when (12) holds. Then the pro…t function
(@xi )2
is quasi-concave in i . We have a multiplicity of equilibria that satisfy:

1 + 2 = 2
for i 2 (0; 2)

In the unique symmetric equilibrium we have that 1 = 2 = 1: The incen-

tives are not distorted from pro…t maximization. The reason for this is that
when …rms are symmetric, the reaction function has zero slope in equilibrium.
Applying the implicit function theorem to (10), we have:
0 qi qj
Ri (qj ) =
2qj

14
When the slope of the reaction function of the competitor is zero, the strategic
e¤ect of incentives disappears and the optimal contract is to tell managers to
maximize pro…ts.
Pm9. We proceed by backward induction. Given that the leaders have produced
i=1 qi , the gameP
followers play is like a standard Cournot game with (residual)
m
demand P = i=1 qi Q. Then, each follower will produce
Pm
i=1 qi c
qf = (13)
n+1
:The pro…t of each leader in the …rst stage as a function of its output is given
by:
m
X Pm
i=1 qi c
i = ( qi n c)qi =
i=1
n+1
m
!
1 X
= qi c qi
n+1 i=1

Except for a multiplicative constant, this is the same pro…t function as in the
standard Cournot case with demand P = Q and m …rms. Then both games
c
will have the same equilibria. In equilibrium each leader produces ql =
m+1
c
and then using (13) each follwer produces qf = . Then pro…ts
(m + 1)(n + 1)
2
1 c
of leaders and followers are given respectively by l = and
n+1 m+1
2 2
1 c
f = .
m+1 n+1
Except for a multiplicative constant l is like the standard Cournot pro…ts
with m …rms and n …rms respectively. Then we can use what we know about
merger pro…tability in the Cournot model to derive the pro…tabilty of mergers of
leaders. In the linear Cournot model (Salant et al (1983)) mergers of two …rms
are never pro…table except in the case of duopoly. The merger of two leaders
will only be pro…table if m = 2. For the same argument we can conclude that
the merger of two followers will only be pro…table if n = 2.

15
7 Problem Set IV. Integration.
1. Assume that market demand is given by Q = 1 P . We have n …rms with
zero costs and one …rm with constant unit cost c that compete à la Cournot.
Suppose that one e¢ cient …rm and the ine¢ cient …rm merge and the new …rm
produces at zero cost.
a) Find the values of c such that the merger increases welfare. Observe
that the merger may increase welfare even though both price and concentration
increase. Show that the result can be stated as the merger increases welfare if

2(n + 1)
s<
2n2 + 6n + 5
where s is the premerger market share of the ine¢ cient …rm (see Faulí-Oller
(2002)).
b) Find the values of c such that the merger increases pro…ts. Show that the
result can be stated as the merger increases pro…ts if

se n(n + 2)
>
s 2(n + 1)

where se is the premerger market share of the e¢ cient …rm (see Faulí-Oller
(2002)).
2. (DOUBLE MARGINALIZATION) Assume that an upstream …rm, de-
noted by U , produces an input at cost cu . n downstream …rms transform the
input into an homogeneous …nal one on a one-for-one basis and at zero marginal
cost. Downstream …rms compete à la Cournot. Find the (observable) price that
…rm U will set for the input (no …xed part is allowed). Show that its pro…ts tend
to the monopoly pro…ts when n tends to in…nity i.e. the margin of downstream
…rms vanishes. Market demand is given by P = Q.
3. Assume that an upstream …rm, denoted by U , produces an input at cost
cu . n downstream …rms transform the input into an homogeneous …nal one on
a one-for-one basis and at zero marginal cost. There also exists a competitive
market where the input can be obtained at price c where c cu . The timing
of the game is the following: in the …rst stage Firm U o¤ers two-part tari¤s
(wi q + Fi ) supply contracts to downstream …rms. In the second stage they
decide whether to accept or reject the contracts. Finally, downstream …rms
compete à la Cournot in the …nal market. Obtain the equilibrium payo¤s of
downstream …rms as a function of cu , n and c if contracts are secret and …rms
have passive conjectures (See Bru et al. IVIE WP-AD 2001-28). Market demand
is given by P = Q.
4. Assume that two …rms (A and B) compete à la Cournot in a market.
Market demand is given by P = Q, unit production cost is c and distribution
costs are zero. Firm B sells directly to consumers whereas Firm A creates
independent franchises to distribute the good. Creating a franchise suppose a
positive but small …xed cost ". Analyze the following game. In the …rst stage,
Firm A decides the number of franchises. In the second stage, it o¤ers two

16
part tari¤ supply contracts (wi q + Fi ) to the franchises. In the third stage, the
franchises decide whether to accept the contract or not and in the fourth stage,
independent franchises and Firm B compete a la Cournot. Obtain the Subgame
Perfect Nash Equilibrium of the previous game for the following situations.
a) Supply contracts are observabe both to the franchises and Firm B.
b) Supply contracts are observable to all franchises but not to Firm B.
c) Each franchise observes only the supply contract that is o¤ered by the
franchisor, but it does not observe the contracts o¤ered to the other franchises.
Firm B does not observe any contract. Furthermore, assume that franchisees
have passive conjectures.
5. Assume that an upstream …rm, denoted by U , produces an input at cost
C(q) = dq 2 . One downstream …rm (D) transforms the input into an homoge-
neous …nal one on a one-for-one basis and at zero marginal cost. Obtain the
supply contract that U o¤ers to D in the following situations:
a) It can o¤er a two-part tari¤ contract.
b) The supply contract can only include a linear price.
6. Assume that an upstream …rm, denoted by U , produces an input at zero
cost. 2 downstream …rms (say Firm 1 and Firm 2) transform the input into
an homogeneous …nal one on a one-for-one basis and at zero marginal cost.
Downstream …rms compete à la Cournot. In the …rst place, …rm U sets the
price for the input it charges to Firm 1 (w1 ) and to Firm 2 (w2 ). In the second
place, …rms compete a la Cournot. Find the Perfect Bayesian equilibrium if
prices for the input are not observed by competitors assuming that downstream
…rms have passive conjectures. Market demand is given by P = Q.

17
8 Solution to Problem Set IV.
1. Using the results obtained in class for a Cournot market with linear demand
and asymmetric constant marginal costs we have that, before the merger, total
(n + 1)(1 c=(n + 1))
output is given by QB = , the output of the ine¢ cient
n+2
1 (n + 1)c 1+c
…rm is qcB = , the output of e¢ cient …rms is qeB = and pro…ts
n+2 n+2
B 2 1
of …rms is B i = qi i = c; e. Observe that c < for the ine¢ cient …rm
n+1
to produce. Otherwise, the ine¢ cient …rm does not produce and the merger has
n
no e¤ect. After the merger, total output is given by QA = , individual
n+1
A 2
output by q A = Qn and individual pro…ts by A = q A .
a) Gross consumer sursplus as a function of sales is given by CS(Q) =
Q2
Q . Social Welfare before the merger is lower than after the merger when:
2
CS(QA ) > CS(QB ) cqeB (14)
2n + 3
or c >
2n3 + 8n2 + 9n + 3
The market share of the ine¢ cient …rm as a function of c is:
qcB 1 (n + 1)c
s(c) = =
QB n+1 c

It decreases with c. Then when (14) holds

2n + 3 2(n + 1)
s(c) < s( )= 2
2n3 2
+ 8n + 9n + 3 2n + 6n + 5
b) The merger is pro…table if the …rms that merge obtain more (joint) pro…ts
after the merger i.e.:
A B B
> e + c (15)
2
n 2
or c >
n3 + 3n2 + 4n + 2
The ratio between premerger output of the e¢ cient and ine¤cient …rms is given
as a function of c by:
qB 1+c
f (c) = eB =
qc 1 c(n + 1)
As it is increasing in c, when (15) holds

n2 2 n(n + 2)
f (c) > f ( )=
n3 + 3n2 + 4n + 2 2(n + 1)

18
2. If the price set by the upstream …rm (the wholesale price) is w, downstream
n( w)
…rms will sell . Then the pro…ts of the upstream …rm as a function of
n+1
w are:
n( w)
(w cu ) (16)
n+1
+ cu
They are maximized at w = . Evaluating (16) at w we have:
2
2
cu n
(17)
2 n+1
2
cu
Given that are the monopoly pro…ts, (17) tends to the monopoly
2
pro…ts when n tends to in…nity.
3. To derive the equilibrium contracts is good to think in the optimal con-
tract between the upstream …rm and one downstream …rm i holding constant
the output of the other …rms to q i . In this case, Firm i will accept the contract
if
2 2
q i wi q i c
F
2 2
and will produce:
i wi q
qi =
2
Then the upstream …rm will set the contract to maximize:

M ax q i wi
wi ,F (wi cu ) +F
2
2 2
q i wi q i c
s:t: F
2 2

Given that he optimally will choose the hightest …xed part that satis…es the
inequality, the previous maximization programm can be written as:
2 2
M ax q i wi q i wi q i c
wi (wi cu ) +
2 2 2

The FOC is given by

q i wi 1 1
( )(wi cu ) ( )( q i wi ) = 0
2 2 2
wi = cu

The …xed fee will be equal to:

2 2
q i cu q i c
F =
2 2

19
Then the downstream …rm will obtain:
2
q i c
(18)
2

As in equilibrium for all …rms the wholesale price will be equal to marginal cost,
they will sell the Cournot outputs with these costs:
cu
qc =
n+1
Then replacing in (18) q i by (n 1)qc , we obtain the payo¤s of downstream
…rms:
2
2( c) (n 1)(c cu )
2(n + 1)
4. a) Denote franchises by subindex i, i = 1:::n In the fourth stage n independent
franchisees, each one with marginal cost wi , and Firm B, with cost c, compete
in the market. Given the results in class the output sold in equilibrium by each
franchise and its pro…ts are given respectively by:
P
(n + 1)wi + c + j6=i wj
qi =
n+2
2
i = q i

In the third stage, franchisees will accept any supply contract whose Fi
is lower than the pro…ts they are going to get in the fourth stage. In the
second stage, Firm A will set the …xed part exactly equal to the pro…ts that
each franchisee is going to obtain in the fourth stage. Then when setting the
wholesale prices, Firm A will maximize:
n P P 2
X (n + 1)wi + c + j6=i wj (n + 1)wi + c + j6=i wj
(wi c) +
i=1
n+2 n+2
n P P
X (n + 1)wi + c + j6=i wj (n + 1)wi + c + j6=i wj
(wi c) +
i=1
n+2 n+2
Pn n P
+ i=1 wi (n + 1)c X (n + 1)wi + c + j6=i wj
n+2 i=1
n+2
Pn Pn
+ i=1 wi (n + 1)c n 2 i=1 wi + nc
(19)
n+2 n+2
The maximization proces will not determine the individual values wi but only
its sum. The optimal sum amounts to:
n
X (n 2) + (3n + 2)c
wi = (20)
i=1
4

20
 Plugging (20) into (19) we obtain the pro…ts Firm A gets by setting the
optimal contract:
2
( c)
8
They amount to the pro…ts the Stacklerberg leader will obtain in this market.
The role of the observable contracts is to turn Firm A into a Stackelberg leader.
Then the objective in the First Stage is:
2
( c)
n"
8
It is decreasing in n. Therefore Firm A will set only 1 independent franchise.
Setting franchises or contracts have the equivalent e¤ect of allowing Firm A to
commit to a higher output in the marketplace. Given that franchises are costly
and contracts are free, Firm A chooses the latter option.
b) Firm B does not observe the supply contracts. Then, optimal supply
contracts have to be calculated holding constant the output of …rm B (qB ). In
other words the optimal supply contracts have to be calculated as if Firm A
served demand P = ( qB ) Q. In class we solved this problem and I only
recall the result. Wholesale prices will be set such that output of franchises (qA )
is equal to the monopoly output

qB c
qA = (21)
2

On the other hand, in equilibrium, Firm B should maximize pro…ts given the
output of franchises. This implies

qA c
qB = (22)
2

Solving (21) and (22), we obtain the equilibrium outputs:

c
qA = qB =
3
With the …xed fee, Firm A will extract all the rents from the franchises. He will
obtain the pro…ts generated by the sales of the franchises. They amount to
2
c c c
( c 2 ) =
3 3 3

Then, in the …rst stage, the pro…ts of Firm A as a function of the number of
franchises is:
2
c
n"
3
It is decreasing in n. Therefore, Firm A will set only 1 independent franchise.
Any possible strategic e¤ect of creating competing franchises is o¤set by the

21
subsequent choice of contracts that will induce franchises to choose the joint
pro…t maximization output.
c) A supply contract (wi q + Fi ) is only observable to the franchise i that re-
ceives it. Then, optimal supply contracts have to be calculated holding
P constant
the output of …rm B (qB ) and the output of the other franchises ( j6=i qj ). In
other words the optimal supply P contracts have to be calculated as if Firm A
served demand P = ( qB j6=i qj ) Q. In class we solved this problem and
I only recall the result. The …xed part extracts all the pro…ts the franchisee is
going to obtain in the fourth stage and the wholesale price is set equal to mar-
ginal cost wi = c. Given those contracts, in the fourth stage both the franchises
and Firm B will compete at cost c. Then everyone will produce:
c
n+2
With the …xed fee, Firm A will extract all the rents from the franchises. He
will obtain the pro…ts generated by the sales of the franchises. They amount to:
2
c c c
( c (n + 1) )n =n
n+2 n+2 n+2

Then in the …rst stage the objective of Firm B will be:

2
c
=n n"
n+2

@ n+2
=( c)2 "=0 (23)
@n (n + 2)3
If " = 0, the solution to (23) is n = 2. This should be the optimal option for
" low enough. Then Firm A sets two franchises.
6.)
a) The objective of U when he sets a linear price of w and extracts all the
rents through the …xed fee is given by:

a w 2w + (1 d)(a w)
i =
2 2
@ i
= 2w + 2d(a w) = 0
@w
da
w =
1+d
It replicates the situation with vertical integration. The output with vertical
a a w a
integration is and = .
2(1 + d) 2 2(1 + d)

22
 b)The objective of U is given by:
a w 2w d(a w)
i =
2 2
@ i
= w(2 + d) + da + (a w)(2 + d) = 0
@w
a(1 + d)
w+ =
2+d
We have w < w+ , then welfare is higher with a two-part tari¤ contract.
7. Assume that the equilibrium we are looking for is given by:

(w1 ; w2 ; q1 (w1 ); q2 (w2 )) (24)

Observe that the strategy of downstream …rms specify an output for any contract
they may receive. To be an equilibrium every agent should be maximizing. As
far as downstream …rms are concerned, this implies that they maximize pro…ts
given the ouput chosen by the competitor in equilibrium.
q2 (w2 ) w1
q1 (w1 ) =
2
q1 (w1 ) w2
q2 (w2 ) =
2
For Firm U , it means that the contracts it sets maximize its revenues. As prices
are not observed by competitors, when Firm U changes the price to Firm 1, Firm
2 still behaves as prescribed by the equilibrium and obtains the same revenues
from it. Therefore, we have to check that Firm U maximizes the revenues he
can separately extract from each downstream …rm. If he sets w1 , Firm U will
obtain:

I(w1 ) = w1 q1 (w1 )
@I(w1 ) 1
= q1 (w1 ) w1 = 0
@w1 2
q2 (w2 )
w1 =
2
Then for (24) to be an equilibrium we must have that:
q2 (w2 )
w1 = (25)
2
We have a symmetric condition for Firm 2:
q1 (w1 )
w2 = (26)
2
From
q2 (w2 ) w1
q1 (w1 ) =
2

23
 q1 (w1 ) w2
q2 (w2 ) =
2
we can obtain that
2w1 + w2
q1 (w1 ) = (27)
3
2w2 + w1
q2 (w2 ) = (28)
3
Plugging (27) and (28) into (25) and (26) and solving the equations we
obtain that:
2
w1 = w1 =
5
Then
q1 (w1 ) = q2 (w2 ) =
5

9 Problem Set V. Product Di¤erentiation.

1. (Tirole (1988) p. 282) Consider the model of di¤erentiation on the line. The
two …rms are located at the two extremities of the segment (x = 0 and x = 1).
Transportation costs are linear in distance (C(d) = td) and the distribution
of consumers is uniform along the segment. The …rms have constant marginal
costs, c1 and c2 , which are not necessarily equal (but, for simplicity, assume
that they do not di¤er too much, so that each …rm has a positive market share
in equilibrium).
Compute the Nash Equilibrium in prices.
2. (Eaton and Lipsey (1975) Review of Economic Studies; 42(1) p. 27-49).
Assume that consumers are uniformly distributed in the segment of length 1.
Each consumer wants to buy one and only one unit of a good. The price of the
good is regulated at a level P greater than the unit cost. Firms decide where
to locate (they only can choose one location). Find the equilibrium when there
are 4 or 5 …rms in the market. (With 3 …rms there is no equilibrium).
3. Assume that consumers are uniformly distributed on a circle with a
perimeter equal to 1. Consumers wish to buy one unit of the good and have lin-
ear transportation costs in distance (C(d) = td). Analyze the Subgame Perfect
Nash Equilibrium of the following game.
In the …rst stage, potential …rms decide whether to enter into the market or
not. Firms that enter have to pay a …xed cost F .
Firms that have entered are automatically located equidistant from one an-
other on the circle.
In the second stage, …rms that have entered compete in prices. For all …rms,
marginal cost is constant and equal to c.
Compare the equilibrium number of …rms with the one that would minimize
the transportation costs.
4. (Cabral (1997) p.86-7) (BRAND PROLIFERATION) Assume that con-
sumers are uniformly distributed in the segment of length 1. Each consumer
wants to buy one and only one unit of a good. The price of the good is regulated

24
at a level P greater than the unit cost c. Two …rms, A and B, can produce the
good. Firms are allowed to locate in more than one location. The cost of each
location is F . We have that:
P c P c
<F <
4 2
Analize the Subgame Perfect Nash Equilibrium of the following game.
In the …rst stage, Firm A chooses its locations.
In the second stage, Firm B chooses its locations.
5. (Motta (1993)) Assume that utility of consumers is given by q p, where
is a taste for quality, q is quality and p is price. is uniformly distributed in
[0; 1]. Consumers want to consume either one or zero units of the di¤erentiated
good. No consumption gives an utility of zero. Firm 2 produces a good of
quality q2 = 1. Firm 1 produces a good of quality q1 , where 0 < q1 < q2 = 1.
Production costs are zero. Find the equilibrium in prices. Observe that there
will be consumers that will prefer not to buy the good i.e. the market will not be
covered. Check that contrary to what happened when the market was covered
the pro…t of the low-quality …rm is not always decreasing in its own quality.
6. Suppose that the price of an industry is regulated at a level p > c where
c is marginal cost. Firm i can spend zi to improve its market share

zi
Pn where 0 < 2:
j=1 zj

The pro…t of …rm i is thus

z
i = (p c)S Pn i zi F
j=1 zj

where S is the size of the market and F the …xed cost of entry. Compute the
(symmetric) free entry equilibrium (the SPE of the two-stage entry-choice of zi
game). Study the asymptotics of the equilibrium when F tends to 0:
7. Consumers are uniformly distributed in the segment [0; 1]. Each consumer
wants to buy one unit of the good. Firm A can locate in any point in the
segment [ 21 ; 0], whereas Firm B can locate in any point in [1; 23 ]. The unit
cost of producing the good is c. Solve the following two-stage game: In the
…rst stage, …rms decide whether to locate. In the second stage, they compete in
prices. Explain the result and compare it with other models in the literature.
(Hint: the second stage equilibrium, given that Firm A locates in a and Firm
B locates in b is given by:
(2 + a + b)t
pA = c+
3
(4 a b)t
pB = c+
3
)

25
10 Solution to Problem Set V. Product Di¤er-
entiation.
1.(See Tirole (1993)p. 301)
2ci + cj
pi = t +
3
2. With 4 …rms two locate in 1/4 and the other two in 3/4.
With 5 …rms, two locate in 1/6, one in 1/2 and the rest in 5/6.
3. (See Tirole (1993) pp. 282-4) In the second stage pro…ts given that n
t
…rms have entered are: 2 .
n
From the zero pro…t condition the number of …rms is obtained:
t
= F
n2 r
t
nc =
F
The number of …rms that minimizes transportation cost is:
r
1 t
n = < nc
2 F
F F
4. If …rm A chooses only one location, it will locate it in (or 1 ).
P c P c
F
Then Firm B will locate in + " and Firm A will obtain zero pro…ts. In all
P c
F
other locations obtains negative pro…ts. If it locates in x 2 [0; ), Firm B
P c
will locate in x + " and Firm A will obtain (P c)x F < 0: If Firm A locates
F F
in x 2 ( ;1 ), Firm B will locate both in x " and x + " and Firm
P c P c
A will obtain F .
If Firm A chooses two locations, it will locate them in 1/4 and 3/4 to reduce
the demand that Firm B can capture. Then Firm B will decide not to enter,
because it can only obtain the fourth of total demand and P 4 c < F: Then Firm
A will obtain P c 2F > 0: Therefore, …rm A will choose two locations.
5. The indi¤erent consumer between buying the good of low and high quality
is:
p2 p1
1 =
1 q1
The indi¤erent consumer between buying the good of low quality and not
buying is:
p1
0 =
q1
Then the pro…ts of the low (…rm 1) and high (…rm 2) quality …rms are given
respectively by:
2 = (1 1 )p2 and 1 = ( 1 0 )p1

26
Equilibrium prices are given by:

q1 (1 q1 ) 2(1 q1 )
p1 = and p2 =
4 q1 4 q1
The pro…ts in equilibrium are:

q1 (1 q1 ) 4(1 q1 )
1 = and 2 =
(4 q1 )2 (4 q1 )2

Then,
@ 1
sign[ ] = sign[ 7q1 + 4]
@q1
Contrary to what happened when the market was covered, it is not always
optimal for the low quality …rm to deteriorate its quality in order to reduce the
competition with the high quality …rm. Now this has the cost of inducing some
consumers not to buy the good.
6. Given that n …rms have entered the market, the pro…t of one of them is
given by:

z
i = (P c)A Pn i zi
j=1 zj

The FOC is given by:

1 Pn
zi j=1 zj zi2 1

(P c)A Pn 2 1=0
j=1 zj

To obtain the symmetric equilibrium we impose it in the FOC:

(n 1)z 2 1
(P c)A 1=0
n2 z 2
The equilibrium level of investment (to improve market share) is:

n 1
z = (P c)A
n2

The zero pro…t condition is given by:

A n 1
(P c) (P c)A F =0 (29)
n n2

F
If we let B = , (29) can be rewritten as:
(P c)A

1 n 1
(P c)A B =0
n n2

27
The number of …rms in the free entry equilibrium is given:
p
1 + (1 )2 + 4B
n =
2B
When A > 1, B > 0. Then, if < 1, we have

Lim 2(1 )
B >0 n = =1
0
1
If = 1, n = p . Then
B
Lim
B >0 n =1
If 1 < 2, then

Lim 1 + 1
B >0 n = = Indeterminate
0
Observe that the value of the square root when B > 0 is 1, because we
have to take its positive solution.
We apply L’Hôpital rule and then

Lim 4
B >0 (1=2) p =
2 (1 )2 + 4B 1

7. Demand given prices is obtained by identifying the indi¤erent consumer:

pA + t(x a) = pB + t(b x)
pB pA + t(b + a)
x(pA ; pB ) =
2t
With prices and demand we can obtain the pro…ts of …rms as a function of
locations. Pro…ts of Firm A and Firm B are given respectively by:
1
A = (pA c)x(pA ; pB ) = (2 + a + b)2 t
18
1
B = (pB c)(1 x(pA ; pB )) = (4 a b)2 t
18
@ A @ B
As we have that > 0 and < 0, the …rst stage equilibrium will be
@a @b
a = 0 and b = 1.

28
11 Problem Set VI: Repeated Interaction.
1. Assume that market demand is given by X = S(a P ). Firms have unit cost
c and compete à la Cournot. This market game is repeated in…nite times. Find
the values of the discount factor such that the symmetric monopoly solution
can be sustained in equilibrium using trigger strategies (reversion to the static
Nash equilibrium).
2. Assume that everything is like in question 1 except that there are only
two …rms. For low values of the discount factor the monopoly solution can
not be sustained in equilibrium. In those cases, …nd the (symmetric) subgame-
perfect equilibrium that yields the …rms the greatest pro…ts, called the extremal
equilibrium.
3. (Bernheim and Whinston (1990)) Assume that we have two …rms that
compete à la Bertrand, demand is given by D(p) and the cost function is given
by: C(q) = cq + F if q > 0 and 0 otherwise. This market game is repeated
in…nite times. Obtain the discount factor that sustain the collusive outcome:
both …rms set the monopoly price and share equally demand. Assume that
…rms use “trigger strategies”. (In the static equilibrium, assume that although
both …rms set the same price, all demand goes to one …rm).
4. (Deneckere (1983)) Assume we have two …rms (…rm 1 and …rm 2) that
compete à la Cournot. Firm i (i = 1; 2) produces good i whose demand express-
ing the price of their product, pi , in terms of the production levels qi is given
by:
pi = 1 qi qj
i; j = 1; 2 i 6= j and 1 > >0
There are no production costs. This market game is repeated in…nite times.
Find the values of the discount factor such that the monopoly solution can be
sustained in equilibrium using trigger strategies (reversion to the static Nash
equilibrium). Study the e¤ect of product di¤erentiation on the possibility of
collusion.

12 Solution to Problem Set VI: Repeated In-

teraction.
1. In class we derived that the discount factor should satisfy:
_
i i
_ ^
(30)
i i
_
i represents the one-period deviation pro…ts, i the (symmetric) collusive prof-
^
its and i the pro…ts in the equilibrium of the one-shot game.
In this case, we have
2
S a c
i = (31)
n 2

29
 2
^ a c
i =S (32)
n+1
One period pro…ts of a …rm if he deviates and produces qi and competitors
stick to the agreement are given by:

S a c
i = (a (n 1) qi )qi
n 2
Then
2
_ (a c)(n + 1)
i =S (33)
4n
Plugging (31),(32) and (33) into (30) we have:
1
4n
1+
(n + 1)2

9
2. For < , the monopoly symmetric solution can not be sustained.
17
In those cases the best symmetric equilibrium will be the lowest output that
satis…es the incentive compatibility constraint. Assume that everybody plays q,
2
2q (S(a c) q)
then each …rm obtains (a c )q. If a …rm deviates, it obtains .
S 4S
Then …rms want to produce q (no to deviate) if:
2 2
1 2q (S(a c) q) a c
(a c )q 0
1 S 4S 1 3

The lowest output that satis…es this equation is:

S(a c)(9 5 )
q =
3(9 )
9
It is greater than half the monopoly output if < : It is decreasing in and
17
2q 9
(a c )q is increasing in if < .
S 17
3. In the one-shot equilibrium both …rms set pc such that

(pc c)D(pc ) = F

and all demand goes to one …rm. (Observe that in this case reversion to the
one-shot Nash equilibrium is an optimal punishment because it yields a payo¤
m
of zero). Then the monopoly solution where every …rm obtains: F; where
2
m
= max(p c)D(p)
p

30
can be sustained if:
m m
F
1 2 2
m

2( m F)
If F = 0, we obtain the well-known result. The greater the …xed costs, the
harder to sustain collusion.
4. First of all, we derive the pro…ts in the one-shot Nash equilibrium ( N ),
in the joint pro…t maximization solution (monopoly) and the deviation pro…ts
from the monopoly solution.
One-shot Nash equilibrium:
Pro…ts of …rm i are given by:

i = (1 qi qj )qi

@ i
= 1 2qi qj = 0 (34)
@qi
i; j = 1; 2 i 6= j (35)
1
Solving the system we obtain the (symmteric) equilibrium q N = . Then
2+
2
N 1
pro…ts are given by = .
2+
Joint-pro…t maximization:
Joint pro…ts are given by:

= (1 q1 q2 )q1 + (1 q2 q1 )q2

@
= 1 2qi 2 qj = 0
@qi
i; j = 1; 2 i 6= j

Solving the system we obtain the (symmteric) joint-pro…t maximization outputs

1 1
q = . Then pro…ts (per …rm) are given by = .
2(1 + ) 4(1 + )
Deviation pro…ts:
If …rm j produces q , then the pro…ts of …rm i are given by:

i = (1 qi )qi
2(1 + )

@ i
= 1 2qi = 0
@qi 2(1 + )
2+
qiD =
4(1 + )

31
 2
D 2+
Then, deviation pro…ts are given by: = .
4(1 + )
Producing the joint-pro…t maximization outputs can be sustained as a Sub-
game Perfect Nash Equilibrium of the in…nitely repeated game if the following
condition holds:
N
D
+
1 1
D
= D N
(36)

As we have:
2
D
=
(4(1 + ))2
2
D N (8 + 8 + 2 )
=
(4(1 + ))2 (2 + )2

Then
(2 + )2
= 2
8+8 +
@
We have that > 0. Then when goods become more di¤erentiated (lower
@
) the set of values of the discount factor such that (36) holds becomes larger.
Then, product di¤erentiation helps collusion to take place.

32