Instructor’s Manual for

Industrial Organization
Theory and Applications
by Oz Shy
The MIT Press
Cambridge, Massachusetts
London, England
Copyright c 1995–2004 Oz Shy. All rights reserved.
The manual was typeset by the author using the L
A
T
E
X2
ε
document preparation software by
Leslie Lamport (a special version of Donald Knuth’s T
E
X program) during the months from
September 1995 to November 1995 while I was visiting the Economics Department of the
University of Michigan. All the figures are also drawn in L
A
T
E
X using software called T
E
Xcad
developed by Georg Horn, which can be downloaded from various mainframes.
Version 1.0 [Draft 20], (1995): Prepared for the first printing of the first edition
bkman21.tex first draft to be posted on the Web (2000/07/07)
(parallel to the 5th Printing of the book)
v.24 (2004/04/14): 6.1.d, 6.1.e
This version: bkman24.tex 2004/04/14 17:36
Contents
To the Instructor v
2 Basic Concepts in Noncooperative Game Theory 1
3 Technology, Production Cost, and Demand 7
4 Perfect Competition 9
5 The Monopoly 11
6 Markets for Homogeneous Products 17
7 Markets for Differentiated Products 25
8 Concentration, Mergers, and Entry Barriers 29
9 Research and Development 33
10 The Economics of Compatibility and Standards 35
11 Advertising 37
12 Quality, Durability, and Warranties 39
13 Pricing Tactics: Two-Part Tariff and Peak-Load Pricing 43
14 Marketing Tactics: Bundling, Upgrading, and Dealerships 45
15 Management, Compensation, and Regulation 49
16 Price Dispersion and Search Theory 51
17 Miscellaneous Industries 53
To the Instructor
Before planning the course, I urge the instructor to read carefully the Preface of the book that
suggests different ways of organizing courses for different levels of students and also provides a
list of calculus-free topics.
The goals of this manual are:
• To provide the instructor with my solutions for all the problems listed at the end of each
chapter;
• to convey to the instructor my views of what the important concepts in each topic are;
• to suggest which topics to choose for different types of classes and levels of students.
Finally, please alert me to any errors or incorrect presentations that you detect in the book
and in this manual. (e-mail addresses are given below). Note that the errata files (according
to the printing sequence) are posted on the Web in PDF format. Before reporting errors and
typos, please view the errata files to see whether the error you found was already identified and
corrected.
Haifa, Israel, (April 14, 2004)
E-mail : ozshy@ozshy.com
Backup : OzBackup@yahoo.com
Web-page : www.ozshy.com
Catalog-page : http://mitpress.mit.edu/book-home.tcl?isbn=0262691795
This draft : bkman24.tex 2004/04/14 17:36
Chapter 2
Basic Concepts in Noncooperative Game Theory
An instructor of a short course should limit the discussion of game theory to the four most
important concepts in this chapter that are essential for the understanding most of the analyses
presented in this book:
1. The definition of a game (Definition 2.1): It is important that the student will understand
that a game is not properly defined unless the list of players, the action set of each player,
and the payoff functions are clearly stated. It is important that the student will understand
the meaning of the term outcome as a list of the specific actions chosen by each player
(and not a list of payoffs as commonly assumed by students).
2. Nash equilibrium (Definition 2.4).
3. Welfare comparisons among outcomes (Definition 2.6).
4. Extensive form games, strategies (compare with actions), subgames, and the SPE (Defi-
nition 2.10).
I urge the instructor to discuss the issues of existence, uniqueness, and multiple equilibria in
class. More precisely, it is important that the student will know that in order to prove existence,
it is sufficient to find only one NE outcome; however, to prove nonexistence, the student must
go over all outcomes and show that at least one player benefits from unilateral deviation.
I believe that the above can be covered in 2 lectures, or in three hours of instruction. If
you wish to devote more time to game theory, you can introduce the equilibrium in dominant
actions (Definition 2.3) before teaching the NE equilibrium concept. If you wish to emphasis
more game theory, I advise covering repeated games (section 2.3).
Answers to Exercises
1. (a) It is straightforward to conclude that
R
1
(a
2
) =

WAR if a
2
= WAR
WAR if a
2
= PEACE
and R
2
(a
1
) =

WAR if a
1
= WAR
WAR if a
2
= PEACE.
That is, WAR is each player’s best response to each action taken by the other player
(hence, WAR is a dominant action for each player). Now, (ˆ a
1
, ˆ a
2
) = (WAR, WAR)
is a (unique) NE since this outcome is on the best-response function of each player.
(b)
R
J
(a
R
) =

ω if a
R
= ω
φ if a
R
= φ
and R
R
(a
J
) =

φ if a
J
= ω
ω if a
J
= φ.
There does not exist a NE for this game since there does not exist an outcome the
is on both best-response functions. That is, R
J
(ω) = ω, but R
R
(ω) = φ, but
R
J
(φ) = φ, but R
R
(φ) = ω, so R
J
(ω) = ω, and so on.
2 Basic Concepts in Noncooperative Game Theory
(c)
R
α
(a
β
) =

B if a
β
= L
T if a
β
= R
and R
β
(a
α
) =

L if a
α
= T
R if a
α
= B.
A NE does not exist for this game since R
α
(L) = B, but R
β
(B) = R, but R
α
(R) =
T, so R
β
(T) = L, and so on.
2. (a) If (T, L) is a NE, then
π
α
(T, L) = a ≥ e = π
α
(B, L), and π
β
(T, L) = b ≥ d = π
β
(T, R).
(b) If (T, L) is an equilibrium in dominant actions, then T has to be a dominant action
for player α, that is
π
α
(T, L) = a ≥ e = π
α
(B, L) and π
α
(T, R) = c ≥ g = π
α
(B, R);
and L is a dominant action for player β, that is
π
β
(T, L) = b ≥ d = π
β
(T, R), and π
β
(B, L) = f ≥ h = π
β
(T, R).
Let us observe that the parameter restrictions given in part (a) are also included in
part (b) confirming Proposition 2.1 which states that an equilibrium in dominant
actions is also a NE.
(c) i. (T, L) Pareto dominates (T, R) if (a ≥ c and b > d) or (a > c and b ≥ d);
ii. (T, L) Pareto dominates (B, R) if (a ≥ g and b > h) or (a > g and b ≥ h);
iii. (T, L) Pareto dominates (B, L) if (a ≥ e and b > f) or (a > e and b ≥ f).
(d) Loosely speaking, the outcomes are Pareto noncomparable if one player prefers (T, L)
over (B, R), whereas the other player prefers (B, R) over the outcome (T, L). For-
mally, either
π
α
(T, L) = a > g = π
α
(B, R) but π
β
(T, L) = b < h = π
β
(B, R),
or
π
α
(T, L) = a < g = π
α
(B, R) but π
β
(T, L) = b > h = π
β
(B, R).
3. (a) There are three Pareto optimal outcomes:
i. (n
1
, n
2
) = (100, 100), where π
1
(100, 100) = π
2
(100, 100) = 100;
ii. (n
1
, n
2
) = (100, 99), where π
1
(100, 99) = 98 and π
2
(100, 99) = 101;
iii. (n
1
, n
2
) = (99, 100), where π
1
(99, 100) = 101 and π
2
(99, 100) = 98;
(b) There is no NE for this game.
1
To prove this, we have to show that for every outcome
(n
1
, n
2
), one of the players will benefit from changing his or her declared value. Let
us look at the following outcomes:
1
A typo in the question (first printing) leads to this undesirable result (see Basu, K. 1994. “The Traveler’s
Dilemma: Paradoxes of Rationality in Game Theory.” American Economic Review 84: 391–395; for a mechanism
that generates (2, 2) as a unique NE outcome).
Basic Concepts in Noncooperative Game Theory 3
i. If (n
1
, n
2
) = (100, 100), then player 1 can increase his or her payoff to π
1
= 101
by declaring n
1
= 99.
ii. If (n
1
, n
2
) = (99, 100), then player 2 can increase his or her payoff from π
2
= 98
to π
2
= 99 by declaring n
2
= 99. Hence, players always benefit from declaring
a value of one dollar lower than the other player, and so on.
iii. If (n
1
, n
2
) = (2, 2), then player 1 can increase his or her payoff from π
1
= 2 to
π
1
= 98 by declaring n
1
= 100.
4. (a) There are several NE outcomes for this game. For example, (LG, SM, SM) is a NE
outcome since
π
C
(LG, SM, SM) = α ≥ γ = π
C
(SM, SM, SM)
π
F
(LG, SM, SM) = β = β = π
F
(LG, LG, SM)
π
G
(LG, SM, SM) = β = β = π
G
(LG, SM, LG).
Hence, no firm finds it profitable to unilaterally change the type of cars it produces.
(b) Again, there are several NE. The outcome (LG, SM, SM) is also a NE for this game,
and the proof is identical to the one given in part (a).
5. (a) There are three subgames: the game itself, and two proper subgames labeled JL
(for Jacob left) and JR (for Jacob right). The three subgames are illustrated in
Solution-Figure 2.1.



φ
ω
ω
φ φ
ω
Rachel
Jacob
π
R
= 2
π
J
= 1
π
R
= 0
π
J
= 0
π
R
= 0
π
J
= 0
π
R
= 1
π
J
= 2

ω
φ
π
R
= 0
π
J
= 0
π
R
= 1
π
J
= 2

φ
ω
π
R
= 2
π
J
= 1
π
R
= 0
π
J
= 0
JL JR
Solution-Figure 2.1: Three subgames of the dynamic Battle of the Sexes
(b) One easy way to find NE outcomes for an extensive form game is to construct a
normal-form representation.
2
However, the normal-form representation is already
given in Table 2.2. Also, equation (2.1) proves that the NE outcomes involve the
two players going together either to football or to the opera. Formally, it is easy to
verify that the following three outcomes constitute NE:
s
J
=

φ if s
R
= ω
φ if s
R
= φ
and s
R
= φ,
2
This subquestion may confuse students who are already thinking in terms of backward induction. Since our
analysis is based on SPE you may want to avoid assigning this subquestion (only). Also, an instructor teaching
from the first printing is urged to change this subquestion to finding the NE for the entire game only; hence, to
postpone finding the NE for the subgames to the next subquestion, where the NE of the subgames are used to
find the SPE.
4 Basic Concepts in Noncooperative Game Theory
s
J
=

ω if s
R
= ω
ω if s
R
= φ
and s
R
= ω,
s
J
=

ω if s
R
= ω
φ if s
R
= φ
and s
R
= φ.
(2.1)
(c) Using backward induction, we first construct Jacob’s strategy (which constitutes the
Nash equilibria for the two proper subgames). Looking at the proper subgames in
Solution-Figure 2.1, we conclude that
R
J
(a
R
) =

ω if a
R
= ω (subgame JR)
φ if a
R
= φ (subgame JL).
Now, we look for Rachel’s strategy (given Jacob’s best response). If Rachel plays
a
R
= ω, then her utility is given by π
R
(ω, R
J
(ω)) = 1. In contrast, if Rachel plays
a
R
= φ, then her utility is given by π
R
(φ, R
J
(φ)) = 2 > 1. Altogether, the strategies
s
R
= a
R
= φ and R
J
(a
R
) =

ω if a
R
= ω
φ if a
R
= φ
constitute a unique SPE.
(d) No! Jacob’s best response is to play φ whenever Rachel plays φ. Thus, Jacob’s
announcement constitutes an incredible threat.
3
Clearly, Rachel, who knows Jacob’s
best-response function in the second stage, should ignore Jacob’s announcement,
since Jacob, himself, would ignore it if Rachel plays φ in the first stage.
6. (a) Jacob’s expected payoff is given by:

J
(θ, ρ) = θρ ×2 +θ(1 −ρ) ×0 + (1 −θ)ρ ×0 + (1 −θ)(1 −ρ) ×1
= 2 + 3θρ −ρ −θ.
Rachel’s expected payoff is given by:

R
(θ, ρ) = θρ ×1 +θ(1 −ρ) ×0 + (1 −θ)ρ ×0 + (1 −θ)(1 −ρ) ×2
= 2 + 3θρ −2ρ −2θ.
(b) The players’ best-response functions are given by
R
J
(ρ) =

0 if ρ < 1/3
[0, 1] if ρ = 1/3
1 if λ > 1/3
and R
R
(θ) =

0 if θ < 2/3
[0, 1] if θ = 2/3
1 if θ > 2/3.
The players’ best-response functions are drawn in Solution-Figure 2.2.
(c) Solution-Figure 2.2 demonstrates that (θ, ρ) = (2/3, 1/3) is a NE in mixed actions.
Also, note that the outcomes (0, 0) and (1, 1) are also NE outcomes in mixed actions
and are the same as the pure NE outcomes.
3
Instructors: Here is a good opportunity to discuss the concepts of credible and incredible threats.
Basic Concepts in Noncooperative Game Theory 5


ρ
θ


θ
ρ


ρ
θ
1 1 1
1 1 1
1
3
R
R
(θ) R
J
(ρ)
R
R
(θ)
R
J
(ρ)

2
3
2
3
1
3


Solution-Figure 2.2: Best-response functions for the Battle of the Sexes in mixed actions
(d) Substituting (θ, ρ) = (2/3, 1/3) into the players’ payoff functions (defined above)
yield

J
(2/3, 1/3) = Eπ
R
(2/3, 1/3) =
2
3
.
(e) The best-response functions in Solution-Figure 2.2 intersect three times, meaning
that in the mixed extension game, there are two equilibria: the two pure NE outcomes
and one NE in mixed actions. In contrast, Figure 2.3 has only one intersection since
in that game NE in pure strategies does not exist, and therefore the best-response
functions intersect only once, at the NE in mixed actions.
Chapter 3
Technology, Production Cost, and Demand
This chapter summarizes the basic microeconomic tools the students need to know prior to
taking this class. My advice for the instructor is not to spend time on this chapter but simply
assign this chapter (with or without the exercises) as reading in the first class. However, my
advice is to return (or refer) to this chapter whenever definitions are needed. For example, when
you first encounter a discussion of returns to scale, I urge you to make a formal definition and
refer the students to Definition 3.2. Similarly, when addressing elasticity issues, students should
be referred to Definition 3.3.
Answers to Exercises
1. (a) Let λ > 1. Then, by Definition 3.2, the technology exhibits IRS if
(λl)
α
(λk)
β
= λ
α+β
l
α
k
β
> λl
α
k
β
,
which holds when α + β > 1. Using the same procedure and Definition 3.2, this
technology exhibit CRS if α +β = 1, and DRS if α +β < 1.
(b) In this technology, the factors are supporting since
MP
L
(l, k) ≡
∂Q
∂l
= αl
α−1
k
β
,
hence,
∂MP
L
(l, k)
∂k
= αβl
α−1
k
β−1
,
which is greater than zero under the assumption that α, β > 0.
2. (a) Let λ > 1. Then, by Definition 3.2, the technology exhibits IRS if
(λl)
α
+ (λk)
α
= λ
α
(l
α
+k
α
) > λ(l
α
+k
α
),
which holds if α > 1. Similarly, the technology exhibits DRS if α < 1, and CRS if
α = 1.
(b) MP
L
(l, k) = αl
α−1
. Hence, ∂MP
L
(l, k)/∂k = 0. Therefore, the factors are neither
substitutes nor complements.
3. Let λ > 1. This (quasi-linear) technology exhibits DRS since
λl +

λk < λl +λ

k = λ(l +

k).
4. (a)
AC(Q) =
TC(Q)
Q
=
F
Q
+c, and MC(Q) =
∂TC(Q)
∂Q
= c.
These functions are drawn in Figure 4.2 in the book.
8 Technology, Production Cost, and Demand
(b) Clearly, AC(Q) declines with Q. Hence, AC(Q) is minimized when Q = +∞.
(c) Declining average cost function reflects an increasing returns to scale technology.
5. (a) Using Definition 3.3,
η
p
(Q) ≡
∂Q(p)
∂Q
p
Q
= (−1)
p
Q
= −
99 −Q
Q
.
Hence, η
p
(Q) = −2 when Q = 33.
(b) From the above, η
p
(Q) = −1 when Q = 99/2 = 49.5.
(c) The inverse demand function is given by p(Q) = 99 − Q. Therefore, TR(Q) =
p(Q)Q = (99−Q)Q. Hence, MR(Q) ≡ dTR(Q)/dQ = 99−2Q, which is a special
case of Figure 3.3.
(d) MR(Q) = 0 when Q = 49.5. That is, the marginal revenue is zero at the output
level where the demand elasticity is −1 (unit elasticity).
(e) Using Figure 3.5,
CS(33) =
(99 −33)66
2
= 2178, and CS(66) =
(99 −66)33
2
= 544.5
6. (a)
p

=
A
Q
. Hence, p = A
1

Q

1

.
(b)
η
p
=
∂Q(p)
∂Q
p
Q
= A(−)p
−−1
p
Q
=
A(−)p

Ap

= −.
Thus, the elasticity is constant in the sense that it does not vary with the quantity
consumed.
(c) The demand is elastic if η
p
< −1, hence if > 1. The demand is inelastic if
−1 < η
p
≤ 0, hence if 0 ≤ < 1.
(d) By Proposition 3.3, MR = p[1 + 1/(−)]. Hence,
p
MR
= 1

1 +
1

=

−1
which is independent of Q.
Chapter 4
Perfect Competition
A perfectly competitive market is characterized by nonstrategic firms, where firms take the
market price as given and decide how much to produce (and, whether to enter, if free entry is
allowed). Most students probably had some discussion of perfectly competitive markets in their
intermediate microeconomics class. However, given the importance of this market structure, I
urge the instructor to devote some time in order to make sure that the students understand
what price-taking behavior means.
It is now the right time to emphasize that, in economics, the term competitive refers to
price taking behavior of agents. For example, we generally assume that our consumers are
competitive, which means that they do not bargain over prices and take all prices and their
income as given. In a competitive market structure, we make a similar assumption about the
firms.
You may want to emphasize and discuss the following points:
1. The assumption of price taking behavior has nothing to do with the number of firms in the
industry. For example, one could solve for a competitive equilibrium even in the presence
of one firm (see an exercise at the end of this chapter). Note that this confusion often
arises since certain market structures yield market allocations similar to the competitive
allocation when the number of firms increases (see for example subsection 6.1.2 which
shows that the Cournot allocation may converge to the competitive allocation when the
number of firms increases to infinity).
2. A major reason for studying and using alternative (noncompetitive) market structures
stems from the fact that the competitive market structure very often “fails” to explain
why concentrated industries are observed.
3. Another major reason for studying alternative market structures stems from the nonexis-
tence of a competitive equilibrium when firms’ technologies exhibit IRS.
Finally, note that this chapter does not solve for a competitive equilibrium under decreasing
returns to scale technologies for two reasons: (i) other market structures analyzed in this book
are also developed mainly for CRS (unit cost) technologies, and (ii) most students are familiar
with the DRS from their intermediate microeconomics class. However, the exercise at the end
of this chapter deals with a DRS technology.
Answers to Exercises
1. Firm 1 takes p as given and chooses q
1
to
max
q
1
π
1
= pq
1
−wL
1
= p

L
1
−wL
1
.
10 Perfect Competition
The first and second order conditions are given by
0 =

1
dL
1
=
p
2

L
1
−w,
d
2
π
1
d(L
1
)
2
= −
p
4L
3/2
1
< 0.
Hence, q
1
=

L
1
= p/(2w).
2. Given that there is only one firm, the supply equals demand equilibrium condition yields
that 120 −p
e
= p
e
/2. Therefore, p
e
= 80; hence, q
e
1
= Q
e
= p
e
/2 = 40.
3. From the production function, we can find the equilibrium employment level to be L
e
1
=
(q
e
1
)
2
= 1600. Hence,
π
1
= p
e
q
e
1
−wL
e
1
= 80 ×40 −1600 = 1600.
4. Given that the firms have the same technologies, they have the same supply functions.
Therefore, the supply equals demand condition becomes 120 − p
e
= p
e
/2 + p
e
/2, which
yields p
e
= 60, Q
e
= 120 −60 = 60, hence, q
e
1
= q
e
2
= 30.
5. Clearly, the competitive price is lower and the aggregate production is higher when the
industry consists of two firms.
6.


p p = 2q
1
q
1


p p = 2q
2
q
2


p p = q
1
+q
2
Q = q
1
+q
2
30 30 60
60
• • •
120
120
p = 120 −Q
Solution-Figure 4.1: Competitive equilibrium with two firms
Chapter 5
The Monopoly
Most students encounter the monopoly problem in their intermediate microeconomics class.
However, the instructor would probably want to make sure that all students fully understand the
monopoly’s choice problem and the effect of price elasticity on the monopoly’s price, as well as
the arguments against monopoly (section 5.2). I also urge the instructor not to skip discussing
discriminating monopoly (section 5.3).
The remaining sections, the cartel (section 5.4), and the durable goods monopolies (sec-
tion 5.5) are more optional depending on the instructor’s tastes and students’ ability.
Answers to Exercises
1. (a)
η
p

dQ
dp
p
Q
= ap
−−1
p
ap

= −.
Hence, the exponential demand function has a constant price elasticity. By Propo-
sition 3.3,
MR(Q) = p

1 +
1

= p

−1

.
(b) Equating marginal revenue to marginal cost yields
1
MR = p
M

−1

= c = MC; hence, p
M
=
c
−1
.
(c) As increases, the demand becomes more elastic, hence, the monopoly price must
fall. Formally,
dp
M
d
=
c( −1) −c
( −1)
2
< 0.
(d) The first edition (first printing) contains a typo. The question asks what happens to
the monopoly’s price when →+1. Clearly, p
M
→+∞. The reason is, that when
= 1, the (entire) demand has a unit elasticity, implying that revenue does not vary
with price (or quantity produced). Hence, given that the revenue is constant (in
fact TR = a when = 1), then the profit maximization problem is reduced to cost
minimization which yields that the monopoly would “attempt” to produce as little
as possible (but still a strictly positive amount).
1
The instructor may want to emphasize to the students that in the case of constant-elasticity demand functions
(exponential demand functions) it is easier to solve for the monopoly’s price first (using Proposition 3.3) and then
solve for the quantity produced by substituting the price into the demand function. This procedure becomes very
handy when solving monopolistic competition equilibria analyzed in section 7.2.
12 The Monopoly
(e) Inverting the demand function yields p(Q) = a
1/
Q
−1/
. Thus,
TR(Q) ≡ p(Q)Q = a
1

Q
1−
1

.
Hence,
MR(Q) ≡
dTR(Q)
dQ
= a
1

Q

1

1 −
1

.
(f) Equating marginal revenue to marginal cost yields
a
1

Q

1

1 −
1

= c,
yielding that the monopoly’s profit maximizing output is
Q
M
= a

−1
c

.
Note that the same result is achieved by substituting p
M
(calculated before) into the
demand function.
2. (a) Solution-Figure 5.1 illustrates an aggregate demand composed of the two groups of
consumers, where each group shares a common valuation for the product.

Q
p
V
H
V
L
n
H
n
H
+n
L

Solution-Figure 5.1: Aggregate demand composed of two consumer groups
(b) The monopoly has two options
2
: setting a high price, p = V
H
, or a low price,
p = V
L
. Solution-Figure 5.1 reveals that the profit levels (revenue since production
is costless
3
) are given by
π|
p=V
H = n
H
V
H
, and π|
p=V
L = (n
H
+n
L
)V
L
.
Comparing the two profit levels yields the monopoly’s profit maximizing price. Hence,
p
M
=

V
H
if V
H
> (n
H
+n
L
)V
L
/n
H
V
L
otherwise.
2
Instructors are urged to assign or discuss this exercise, since it provides a good opportunity to introduce the
student to a discrete (logic based) analysis which is used later in a wide variety of topics (see for example the
section on tying [section 14.1]).
3
The first printing of the first edition neglects to assume that production of G-Jeans is costless.
The Monopoly 13
Thus, the monopoly sets a high price if either there are many high valuation con-
sumers (n
H
is large) and/or these consumers are willing to pay a very high price
(V
H
is high).
3. (a) In market 1, p
1
= 2 −q
1
. Hence, by Proposition 3.2, MR
1
(q
1
) = 2 −2q
1
. Equating
MR
1
(q
1
) = c = 1 yields q
1
= 0.5. Hence, p
1
= 1.5.
In market 2, p
2
= 4 −q
2
. Hence, by Proposition 3.2, MR
2
(q
2
) = 4 −2q
2
. Equating
MR
2
(q
2
) = c = 1 yields q
2
= 1.5. Hence, p
2
= 2.5.
(b) π
1
= (p
1
− c)q
1
= (0.5)
2
= 0.25, and π
2
= (p
2
− c)q
2
= (1.5)
2
= 2.25. Summing
up, the monopoly’s profit under price discrimination is π = 2.5.
(c) There are two cases to be considered: (i) The monopoly sets a uniform price p ≥ 2
thereby selling only in market 2, or (ii) setting p < 2, thereby selling a strictly positive
amount in both markets. Let us consider these two cases:
i. If p ≥ 2, then q
1
= 0. Therefore, in this case the monopoly will set q
2
maximize
its profit in market 2 only. By subquestion 3a above, π = π
2
= 2.25.
ii. Here, if p < 2, q
1
> 0 and q
2
> 0. Therefore, aggregate demand is given by
Q(p) = q
1
+q
2
= 2−p+4−p = 6−2p, or p(Q) = 3−0.5Q. By Proposition 3.2,
MR(Q) = 3 − Q. Equating MR(Q) = c = 1 yields Q = 2, hence, p = 2.
Hence, in this case π = (p −c)2 = 2 < 2.25.
Altogether, the monopoly will set a uniform price of p = 2.5 and will sell Q = 1.5
units in market 2 only.
4
4. (a)
π(q
1
, q
2
) = (100 −q
1
/2)q
1
+ (100 −q
2
)q
2
−(q
1
+q
2
)
2
.
(b) The two first order conditions are given by
0 =
∂π
∂q
1
= 100 −q
1
−2(q
1
+q
2
)
0 =
∂π
∂q
2
= 100 −2q
2
−2(q
1
+q
2
).
Solving for q
M
1
and q
M
2
yields that q
M
1
= 25 and q
M
2
= 12.5.
(c) Substituting the profit maximizing sales into the market demand functions yield
p
M
1
= p
M
2
= 87.5. Hence,
π(q
M
1
, q
M
2
) = 87.5 ×12.5 + 87.5 ×25 −(12.5 + 25)
2
= 1875.
(d) Now, that each plant sells only in one market, the two first order conditions become
0 =
∂π
∂q
1
= 100 −q
1
−2q
1
0 =
∂π
∂q
2
= 100 −2q
2
−2q
2
.
4
Note that consumers in market 1 are better off under price discrimination than without it, since under no
discrimination no output is purchased in market 1. Given that the price in market 2 is the same under price
discrimination and without it, we can conclude that in this example, price discrimination is Pareto superior to
nonprice discrimination, since both consumer surplus and the monopoly profit are higher under price discrimination.
14 The Monopoly
yielding q
M
1
= 100/3 and q
M
2
= 25.
(e) p
M
1
= 100 −100/6 = 250/3 and p
M
2
= 100 −25 = 75. Hence,
π = π
1
(q
1
) +π
2
(q
2
) =
250
3
×
100
3

100
3

2
+ 75 ×25 −25
2
= 2917.
(f) This decomposition increases the monopoly’s profit since the technology exhibits
DRS.
5. By section 5.3, the discriminating monopoly will set quantities to satisfy MR
1
(q
M
1
) =
MR
2
(q
M
2
). Hence, by Proposition 3.3
MR
1
(q
1
) = p
M
1

1 +
1
−2

= p
M
2

1 +
1
−4

= MR
2
(q
2
).
Thus, p
M
1
= 1.5p
M
2
.
6. (a) We solve for the monopoly’s profit maximizing prices starting from the second period.
The second period outcome may depend on two cases:
First-period consumer does not buy in period 1: Clearly, in this case, the sec-
ond period profit maximizing price is p
2
= 50, yielding a profit level of π
2
=
3 ×50 = 150.
First-period consumer buys in period 1: In this case the second period profit max-
imizing price is again p
2
= 50, yielding a profit level of π
2
= 2 ×50 = 100.
Altogether, the second-period price is independent of the action of the first-period
buyer in the first period. Therefore, the maximum price the monopoly can charge
the first-period buyer in the first period is p
1
= 150.
(b) The second period outcome may depend on two cases:
First-period consumer does not buy in period 1: In this case, the monopoly has
two choices: (i) charging p
2
= 20, and sell to all three consumers, thereby
earning a second period profit of π
2
= 3 × 20 = 60; or, (ii) charging p
2
= 50,
and selling only to the second period consumers, thereby earning a second-period
profit of π
2
= 2 ×50 = 100.
First-period consumer buys in period 1: In this case the second period profit max-
imizing price is again p
2
= 50, yielding a profit level of π
2
= 2 ×50 = 100.
Altogether, the second period price is independent of the actions of the first-period
buyer. Now, in order to attract the first-period buyer to purchase in period 1, the
monopoly should set p
1
= 40, thereby extracting all surplus from all consumers.
7. The two cases are illustrated in Solution-Figure 5.2.
(a) Equating marginal revenue to the tax inclusive unit cost yields a − 2Q = c + t, or
Q
M
= (a − c − t)/2. Hence, p
M
= (a + c + t)/2. Now, checking the effect of a
tax rate change on the monopoly’s price yields that dp
M
/dt = 1/2 < 1. Hence, as
illustrated in Solution-Figure 5.2 (left), an increase in t raises the monopoly price by
less than t.
The Monopoly 15


Q
p
c
c +t
MR(Q)
a


Q
p
c
c +t
p
M
(c)
p
M
(c +t)
MR
D
D
p
M
(c +t)
p
M
(c)




Solution-Figure 5.2: How the monopoly price varies with a specific tax
(b) Using Proposition 3.3,
MR = p
M

1 +
1
−2

= c +t, yielding p
M
= 2(c +t).
Now, checking the effect of a tax rate change on the monopoly’s price yields that
dp
M
/dt = 2 > 1. Hence, as illustrated in Solution-Figure 5.2 (right), an increase in
t raises the monopoly price by more than t.
Chapter 6
Markets for Homogeneous Products
The Cournot, Bertrand, and sequential moves market structures are essential for a further study
of industrial organization.
The important issues to discuss and emphasize are:
1. Market structures are assumed by the researcher rather than solved for. Specifying a
market structure is the same as specifying the rules of the game before predictions (e.g.,
equilibrium outcomes) are sought.
2. Cournot and Bertrand market structures select Nash equilibrium outcomes, where under
the Cournot game firms’ actions are the quantity produced, and in the Bertrand game
firms’ actions are prices.
3. Free entry versus a fixed number of firms.
Self-enforcing collusion (section 6.5) is suited for more advanced students. Instructors may
want to go over repeated games (section 2.3) before covering self-enforcing collusion, however,
this topic is written in a way that enables the instructor to teach self-enforcing collusion even
without formally teaching repeated games. Instructors who choose to teach this topic may
want to emphasize that a trigger strategy is a function of the entire history of what each player
has played in earlier periods, including the player’s own earlier actions, and therefore provides
self-discipline which restricts the gains from deviation from the cooperative actions.
Finally, subsection 6.6.2 (preferential trade agreements) provides a good example to the law
of (no) second best, by showing that a partial removal of trade barriers need not be welfare
improving.
Answers to Exercises
1. (a) By Lemma 4.1, both firms produce a finite amount of output only if p ≤ c
1
and
p ≤ c
2
. Hence, since c
2
> c
1
, if a competitive equilibrium exists then it must be
that p
e
≤ c
1
. However, if p
e
< c
1
then both firms produce zero output, but at these
prices demand exceeds zero (since Q(c
1
) = α − c
1
> 0). Hence, if a competitive
equilibrium exists, then p
e
= c
1
. Therefore, q
e
1
= Q
e
= α −c
1
and q
e
2
= 0.
(b) Each firm i takes the output of its opponent, q
j
, as given and solves
max
q
i
π
i
= (α −q
i
−q
j
)q
i
−c
i
q
i
, i, j = 1, 2, i = j,
yielding first order conditions given by
0 =
∂π
1
∂q
1
= α −2q
1
−q
2
−c
1
, and 0 =
∂π
2
∂q
2
= α −q
1
−2q
2
−c
2
.
18 Markets for Homogeneous Products
Solving the two equations for the two variables, q
1
and q
2
yields
q
c
1
=
α −2c
1
+c
2
3
, q
2
=
α −2c
2
+c
1
3
, and Q =
2α −c
1
−c
2
3
.
Then, p
c
= α −q
c
1
−q
c
2
= (α +c
1
+c
2
)/3.
(c) From the first order conditions of the previous subquestion we can immediately solve
for firm 2’s best-best response function. Hence,
R
2
(q
1
) =
α −c
2
2

q
1
2
.
In a sequential-moves equilibrium, firm 1 (the leader) takes firm 2’s best-response
function as given and chooses q
1
to
max
q
1
π
1
=
¸
α −q
1

α −c
2
2

q
1
2

q
1
−c
1
q
1
,
yielding the first order condition
0 =
∂π
1
∂q
1
= α −2q
1

α −c
2
2
+q
1
−c
1
.
Therefore,
q
s
1
=
α +c
2
−2c
1
2
, q
s
2
=
α −3c
2
+ 2c
1
4
, and Q
s
=
3α −c
2
−2c
1
4
.
Hence, p
s
= (α +c
2
+ 2c
1
)/4.
(d) When firm 2 is the leader, by symmetry (replacing c
1
by c
2
and vice versa in the
previous subquestion) we can conclude that
q
s
2
=
α +c
1
−2c
2
2
, q
s
1
=
α −3c
1
+ 2c
2
4
, and Q
s
=
3α −c
1
−2c
2
4
.
Hence, p
s
= (α +c
1
+ 2c
2
)/4.
Comparing the two sequential-moves equilibria we see that (i) aggregate output
decreases when the less efficient firm (firm 2) moves before firm 1; (ii) price is higher
when firm 2 moves first; (iii) the market share of the less efficient firm increases
when it moves before firm 1; (iv) the production of the more-efficient firm increases
when it moves first; and (v) the market share of the more-efficient firm increases
when it moves first.
(e) In a Bertrand equilibrium, the more efficient firm (firm 1) completely undercuts the
price set by firm 2. Assuming that money is continuous, firm 1 sets its price to
equal the unit cost of firm 2, thereby ensuring that firm 2 will not find it profitable
to produce any amount. Hence, we can approximate the Bertrand equilibrium by
p
b
1
= c
2
− and p
b
2
= c
2
. Thus, q
b
1
= Q
b
= α − c
2
+ and q
b
2
= 0. Notice that
when the unit costs are not equal (the present case), the Bertrand equilibrium price
exceeds the competitive equilibrium price. Also, instructors should point out to the
students that this is not really a Nash-Bertrand equilibrium since the profit of firm 1
increases as decreases.
Markets for Homogeneous Products 19
2. (a) We first focus on the problem faced by firm 1. In a Cournot market structure, firm 1
takes q
2
, q
3
, . . . , q
N
as given and chooses q
1
that solves
max
q
1
π
1
= (100 −q
1
−q
2
−· · · −q
N
)q
1
−F −(q
1
)
2
.
The first order condition is given by
0 =
∂π
1
∂q
1
= 100 −2q
1
−q
2
−· · · −q
N
−2q
1
.
Therefore, the best-response function of firm 1 is given by
q
1
= R
1
(q
2
, . . . , q
N
) =
100 −
¸
N
j=2
q
j
4
.
Given that all firms have identical technologies (identical cost functions), we can
attempt to search for a Cournot equilibrium where all firms produce the same output
levels. That is, q
c
≡ q
c
1
= q
c
2
= · · · = q
c
N
. In this case,
q
c
= 25 −
(N −1)q
c
4
, or q =
100
N + 3
, hence, Q =
100N
N + 3
.
Therefore,
p
c
= 100 −
100N
N + 3
=
300
N + 3
, and π
i
= p
c
q
c
−F −(q
c
)
2
=
20, 000
(N + 3)
2
−F.
(b) Under free entry, firms enter as long as they make above normal profits. Entry stops
when a further entry generates a loss to firms. Hence, if we approximate the number
of firms, N, by a real number, under free entry all existing firms make zero profit.
Thus,
0 = π
i
=
20, 000
(N + 3)
2
−F, or N =

20, 000
F
−3.
Thus, the (endogenously determined) number of firms in this industry declines with
the fixed cost parameter F.
3. In the third period, firm 3 takes q
1
and q
2
as given and chooses q
3
to maximize its profit.
The solution to this problem yields firm 3’s best-response function R
3
(q
1
, q
2
) which is
given in (6.8). Hence, using (6.8) we have it that
R
3
(q
1
, q
2
) =
120 −c
2

q
1
+q
2
2
.
In the second period, firm 2 takes ¯ q
1
and R
3
(¯ q
1
, q
2
) as given and chooses q
2
that solves
max
q
2
π
2
= [120 − ¯ q
1
−q
2
−R
3
(¯ q
1
, q
2
)]q
2
−cq
2
.
The first-order condition is given by 0 = dπ
2
/dq
2
= 60 − ¯ q
1
/2 −q
2
−c/2. Therefore, the
best-response function of firm 2 is given by
R
2
(q
1
) =
120 −q
1
−c
2
.
20 Markets for Homogeneous Products
In the first period, firm 1 takes the best-response functions of firms 2 and 3 as given and
chooses q
1
that solves
max
q
1
π
1
= {120 −q
1
−R
2
(q
1
) −R
3
[q
1
, R
2
(q
1
)]}q
1
−cq
1
= {30 +
3c
4

q
1
4
}q
1
−cq
1
.
The first-order condition is given by 0 = dπ
1
/dq
1
= 30 −c/4 −q
1
/2. Therefore, solving
for q
s
1
, and then substituting into R
2
(q
s
1
), and then into R
3
[q
s
1
, R
2
(q
s
1
)] yields,
q
s
1
=
120 −c
2
, q
s
2
=
120 −c
4
, and q
s
3
=
120 −c
8
.
Hence, Q
s
= q
s
1
+q
s
2
+q
s
3
= 7(120 −c)/8 and p
s
= (120 + 7c)/8.
4. (a) Under zero production cost, all Bertrand equilibrium outcomes yield zero profits to
both firms. Thus the Bertrand outcome is p
b
1
= 0 and p
b
2
= 0.
(b) We’ll construct an equilibrium in trigger strategies where firms’ threat point is p
b
1
= 0
and p
b
2
= 0, respectively. Formally, let firm i’s price strategy, i = 1, 2, be given by
p
i
(τ) =

10 as long as p
1
(t) = p
2
(t) = 10 for all t = 1, . . . τ −1
0 otherwise.
Now, if no firm deviates in any period, the present value of the sum of discounted
profit to each firm i is π
i
= 5N/(1 − ρ). If firm 1 deviates in period τ, then
π
1
= (10 −)N + 0 ≈ 10N (since effective period τ + 1 all firms revert to p
i
= 0).
Hence, deviation is not profitable to either firm if 5N/(1 −ρ) > 10N, or ρ > 1/2.
(c) All Bertrand equilibria are of the form p
b
1
= 4 − ≈ 4 and p
b
2
= 4.
(d) We can reconstruct firms’ trigger strategies where we assume that the punishment
point is now p
b
1
= 4 and p
b
2
= 4. Since now firms have different profit functions, we
need to check profitability from deviation by each firm separately.
Firm 1: In each period τ, if no firm deviates, the sum of discounted profits is π
1
=
5N/(1 −ρ). If firm 1 deviates in period τ, then its sum of discounted profit is
π
1
= 10N +ρ4N/(1 −ρ). Comparing the two profit levels shows that deviation
is not profitable for firm 1 if ρ > 5/6.
Firm 2: If no firm deviates, the sum of discounted profit is π
2
= (N/2)(10−4)/(1−
ρ). If it deviates, then the sum of discounted profit is π
2
= (10 − 4)N + 0.
Comparing the two profit levels reveals that deviation is not profitable if ρ > 1/2.
(e) Collusion is less likely to be sustained when one firm has a cost advantage over the
other because deviation by firm 1 (low cost) will leave it with a positive stream of
profit compared with a zero profit generated by deviation when the two firms have
an identical cost structure.
5. (a) Let p
A
k
denote the (tariff inclusive) price consumers in country A pay for a unit of
import from country k, k = B, C. Therefore, under a uniform specific tariff of $10
per unit of import, p
A
B
= 70 and p
A
C
= 50. Clearly consumers buy from country C,
and the government earns a revenue of $10 per unit. A FTA with country B reduces
p
A
B
to p
A
B
= 60 > 50 = p
A
C
. Hence, A’s consumers still purchase from country C
implying that this FTA is ineffective and therefore does not alter A’s welfare.
Markets for Homogeneous Products 21
(b) Now, a FTA with country B reduces p
A
B
to p
A
B
= 60 < 60.01 = p
A
C
. Hence, A’s
consumers switch to buying from country B and no tariff revenues are collected. Note
that in this case the consumer price falls from p
A
C
= 60.01 to p
A
B
= 60 (a reduction
of 1 cent per unit of import). Hence, the FTA hardly changes consumer surplus.
However, in this case the government faces a large reduction in tariff revenues due
to the elimination of a $10 tariff per unit of import. Altogether, with the exception
of highly elastic demand, country A loses from the FTA.
6. (a) Clearly p
L
is a dominant action for each firm in the static one-shot game. Hence,
by Proposition 2.2, (p
L
, p
L
) is a unique NE for the one-shot game. Therefore, if a
SPE exists, it must yield that both firms charge p
L
.
We can derive the SPE directly by formulating the extensive form game which is
illustrated in Solution-Figure 6.1 (consider the first two stages only).



p
L
p
H
p
H
p
L
p
L
p
H
Firm 1
Firm 2
π
1
= 70
π
2
= 70
π
1
= 120
π
2
= 0
π
1
= 0
π
2
= 120
π
1
= 100
π
2
= 100

p
L
π
1
= 70
π
2
= 70
Firm 1
(II.:)
(I.:)
(III.:)
Solution-Figure 6.1: Sequential price game: Meet the competition clause
From the figure, it is straightforward to conclude that the following prices constitute
a SPE:
p
2
=

p
L
if p
1
= p
H
p
L
if p
1
= p
L
and p
1
= p
L
.
(b) We now add a third stage to the game, where firm 1 commits to reduce its price to
firm 2’s price whenever firm 2 charges p
L
. This third stage is (partly) illustrated at
the bottom of Figure 6.1. In this case, the SPE if given by
p
2
=

p
H
if p
1
= p
H
p
L
if p
1
= p
L
and p
1
= p
H
,
implying that the firms charge the industry’s profit maximizing price and earn a profit
of 100 each.
7. Note that the answer for exercise 3 provides a solution for N = 3 firms, which intuitively
can be generalized to the N > 3 case. In addition, in what follows, I provide a formal
solution for this problem.
22 Markets for Homogeneous Products
(a) Let q
s
i
denote the sequential-moves equilibrium output level produced by firm i,
i = 1, . . . , N. Define Q
s

¸
N
i=1
q
s
i
and Q
s
−N
=
¸
N−1
i=1
q
s
i
, for i = 1, . . . , N where
Q
s
−1
≡ 0 for i = 1.
Since there is no fixed cost, there is no limit on the number of firms that will enter if
free entry is allowed. We therefore look at how the Nth firm decides on how much
to produce. Firm N solves
max
q
N
π
N
= (a −Q
s
−N
−q
N
−c)q
N
,
yielding q
s
N
= (a −Q
s
−N
−c)/2.
Before we proceed to analyzing firm N −1, let us define a
N
≡ a −Q
s
−N
. Hence, it
can be verified that
a
N−1
= a
N
+q
s
N−1
for allN.
Firm N −1 knows what firm N will be producing and solves
max
q
N−1
π
N−1
=
¸
(a −Q
s
−(N−1)
−q
s
N−1
) −
a
N
−c
2
−c

q
N−1
=
[a
N
−c]q
N−1
2
=
[a
N−1
−q
N−1
−c]q
N−1
2
.
Hence, q
s
N−1
= (a
N−1
−c)/2.
Also, since a
N
= a
N−1
−q
s
N−1
and a
N
= a
N−1
−q
s
N−1
= (a
N−1
+c)/2,
q
s
N
= (a
N
−c)/2 = (a
N−1
−c)/4.
We now analyze firm N − 2. Recall that a
N−2
− q
s
N−2
= a
N−1
. Also, recall that
q
s
N
+q
s
N−1
= (1/2 + 1/4)(a
N−1
−c). Therefore, firm 2 solves
max
q
N−2
π
N−2
=
¸
(a
N−2
−q
N−2
) −
3(a
N−2
−q
N−2
−c)
4
−c

q
N−2
=
[a
N−2
−q
N−2
−c]q
N−2
4
.
The solution is q
s
N−2
= (a
N−2
− c)/2. Hence, q
s
N−1
= (a
N−2
− c)/4 and q
s
N
=
(a
N−2
−c)/8. Therefore,
q
s
N
=
a
N
−c
2
, for all N.
In particular, for N = 1, a
1
= a −Q
s
−1
= a −0 = a. Thus, q
s
1
= (a −c)/2.
Finally, since
q
s
N
=

1
2

k
q
s
N−k
, for all 0 < k < N,
we have it that
q
s
N
=

1
2

N−1
q
s
1
=
1
2
N−1
a −c
2
=
a −c
2
N
.
Altogether, the N-firm sequential moves equilibrium output levels are
q
s
i
=
a −c
2
i
, for all i = 1, . . . , N.
Markets for Homogeneous Products 23
(b) Aggregate output is given by
Q
s
=
¸
i=1
Nq
s
i
= (a −c)

1
2
+
1
4
+· · · +
1
2
N

.
Using the same arguments as in the proof of Lemma 9.2 (see mathematical appendix,
section 9.9), we can show that
1
Q
s
= (a −c)

1 −
1
2
N

.
(c) As the number of firms increases, Q
s
→ a − c (competitive output level). Hence,
p
s
→c (the competitive price level).
1
The first printing of the first edition contains a typo in the specification of Q
s
in the question.
Chapter 7
Markets for Differentiated Products
The first section extends the market structures defined in Chapter 6 to markets with differentiated
products. Using the two-brand environment enables us to analyze the difference in market
outcomes generated by the Cournot and the Bertrand market structures, and to define and
contrast the concepts of strategically substitutes and strategically complements best-response
functions.
Section 7.2 (somewhat more advanced) introduces the student to the endogenous determi-
nation of the variety of brands, and its application to the theory of international trade.
Location models (section 7.3), in particular the Hotelling model (subsection 7.3.1), are im-
portant for a further study of industrial organization. The linear city model is used several times
in the manuscript, for example, to distinguish between vertically and horizontally differentiated
products (see section 12.2). Instructors should cover at least the Hotelling linear city model
which students always find very intuitive and therefore very appealing. The existence proof
(Appendix, section 7.5) can be skipped. The instructor can demonstrate the parameter restric-
tions needed to ensure existence by imposing symmetric locations (a = b) into the conditions
of Proposition 7.6.
Answers to Exercises
1. (a) The best-response functions are drawn in Figure 7.1 and are derived as follows:
max
α
1
π
1
⇒ R
1

2
) = 2 +
3
2
α
2
, or, α
2
= −
4
3
+
2
3
R
1

2
);
max
α
2
π
2
⇒ R
2

1
) = 1 +
1
2
α
1
.


α
2
α
1
R
1

2
)
R
2

1
)
8
1
0

4
3
14

Solution-Figure 7.1: Advertising best-response functions
26 Markets for Differentiated Products
(b) The two best-response functions are upward sloping, hence strategically comple-
ments.
(c) Solving for a NE yields α
N
1
= 14 and α
N
2
= 8. The firms’ profit levels in a NE are:
π
1
= 14
2
and π
2
= 8
2
.
2. (a) Given a low reservation price (a low B), we’ll attempt to find a monopoly equilibrium
where not all the market is served. Figure 7.2 illustrates the location of the two
indifferent consumers, one on each side of the firm. The indifferent consumer (on
0 1
1
2
1
2
+a
1
2
−a
2a
✲ ✛
Solution-Figure 7.2: Single-firm location model
each side) is determined by the reservation utility, that is, B − a − p = 0, or,
a = B −p. Hence, the monopoly chooses p that solves
max
p
π = p2a = 2p(B −p).
The solution is given by
p
M
=
B
2
, a =
B
2
, and π
M
= 2pa =
B
2
2
.
Now, it is easy to verify that 0 < B < 1 implies that 0 < a < 1/2, hence, not all
the market is served.
(b) When substituting B > 1 into the solution for ‘a’ found in the previous subquestion,
we get that the indifferent consumers lie “outside” the city. This implies that for
B > 1, the monopoly can increase the price and still having the entire street purchase
the product. Thus, the monopoly will pick the highest possible price subject to having
the consumers living at the edges of town purchase the product. Formally, set p to
satisfy B −1/2 −p = 0. Hence,
p
M
= B −
1
2
, a =
1
2
, and π
M
= B −
1
2
.
3. (a) The indifferent consumer, denoted by ˆ x, must satisfy
ˆ x ×1 +p
1
= (1 − ˆ x) ×R +p
2
.
Hence,
ˆ x =
R +p
2
−p
1
1 +R
.
(b) Substituting p
1
= p
2
yields ˆ x = R/(1 +R). Hence,
lim
R→∞
ˆ x = lim
R→∞
R
1 +R
= 1.
Markets for Differentiated Products 27
Therefore, all the consumers will eat in restaurant 1 (i.e., ˆ x = 1) only if the trans-
portation cost for traveling to the east is infinite. Otherwise, some consumers will
always eat at restaurant 2.
4. From (7.24), we can define the location of the indifferent consumer, ˆ x, by the implicit
function:
F(p
A
, p
B
) ≡ p
A
−p
B
+τ(ˆ x −a)
2
−τ(ˆ x −L +b)
2
= 0.
Firm A chooses p
A
that solves max
p
A
π
A
= p
A
ˆ x subject to ˆ x satisfying F(p
A
, p
B
) = 0.
The first-order condition is given by
0 =

A
dp
A
= ˆ x +p
A
∂ˆ x
∂p
A
= ˆ x +p
A


∂F
∂p
A
∂F
∂ˆ x

= ˆ x −
p
A
2τ(L −a −b)
.
Similarly, firm B chooses p
B
that solves max
p
B
π
B
= p
B
(L − ˆ x), yielding a first-order
condition given by
0 =

B
dp
B
= L − ˆ x −p
B
∂ˆ x
∂p
B
= ˆ x +p
B

∂F
∂p
B
∂F
∂ˆ x

= L − ˆ x −
p
B
2τ(L −a −b)
.
(a) Under a = b, we search for a symmetric solution. Substituting a = b and ˆ x = L/2
into firm A’s first-order condition (or B’s first-order condition) yields that
p
A
= p
B
= τL(L −2a).
Note that p
A
= p
B
→0 when a →1/2, reflecting the fact that prices drop to zero
when the two brands become homogeneous.
(b) We now investigate the total effect of varying the location of firm A (varying a) on
the profit of firm A.
1
Using the two first-order conditions, we have it that p
A
= 2τ(L − a − b)ˆ x and
p
B
= 2τ(L −a −b)(L − ˆ x). Substituting into the function F(, ) = 0 yields
F(, ) = 2τ(L −a −b)(2ˆ x −L) +τ(ˆ x −a)
2
−τ(ˆ x −L +b)
2
= 0.
Using the implicit function theorem, we have it that
∂ˆ x
∂a
= −
−(2ˆ x −L) −2(ˆ x −a)
2(L −a −b) + 2(ˆ x −a) −2(ˆ x −L +b)
=

x=1/2
a=b
1
4
.
Now, using A’s first-order condition, we can express A’s profit by π
A
= p
A
ˆ x =
2τ(L −a −b)(ˆ x)
2
. Therefore,

A
da
< 0 whenever −(ˆ x)
2
+ (L −a −b)2ˆ x
1
4
< 0,
which always hold since L − 2a < L. Hence, firm A increases its profit by moving
toward point 0. Similarly, firm B increases its profit by moving toward point L.
1
This problem may be too difficult for most undergraduate students since it involves deviating from the
symmetric solution.
Chapter 8
Concentration, Mergers, and Entry Barriers
This chapter contains four major topics: How to measure concentration; merger and merger
guidelines; entry barriers; and entry deterrence. I urge the instructor not to skip the discussion
of concentration measures (section 8.1) and the appendix on merger regulations and guidelines
(section 8.6).
Regarding all other topics, the instructor will find a wide variety of models to choose from.
Most of the topics are logical and require minimum calculations and very little calculus (if
any).
1
Finally, note that you can combine the topic of double monopoly markup, analyzed
under the topic of dealership (subsection 14.3.1), with the analysis of vertical integration given
in subsection 8.2.2.
Answers to Exercises
1. (a) I
4
= 20 + 20 + 20 + 10 = 70.
(b) I
HH
= 4 ×10
2
+ 3 ×20
2
= 1, 600.
(c) i. After the merger of firms 1 and 2, I
HH
= 20
2
+ 10
2
+ 10
2
+ 3 ×20
2
= 1, 800.
ii. ∆I
HH
= 1, 800 −1, 600 = 200.
iii. The merger may be challenged by the FTC or the Justice Department since the
postmerger I
HH
exceeds 1, 000 and ∆I
HH
= 200 > 100.
2. (a) The solution for a Cournot equilibrium with N firms is given in (6.9) and (6.10).
Substituting c = 0 (zero production cost) and N = 3 yields q
c
i
= 25 and π
c
i
= 625,
i = 1, 2, 3.
(b) Substituting N = 2 into (6.10) yields π
1
= π
4
= 1111.111 . . ..
(c) Since firms 2 and 3 split the profit from the merged firm, π
2
= π
3
= π
4
/2 = 555 <
625. Hence, firms 2 and 3 lose from this merger whereas firm 1 gains.
(d) A merger into a monopoly cannot reduce aggregate profit, simply because a monopoly
can always mimic the same production pattern of any oligopoly. To see this, sub-
stituting N = 1 into (6.10) yields a monopoly profit of π
M
= 2500. Hence,
π
i
= 2500/3 = 833
1
3
> 625 > 555. Thus, the profit per firm is the highest
under the monopoly market structure (cartel).
(e) The difference between the two mergers is as follows: When two out three firms
merge, competition is reduced to a duopoly; however, the extra aggregate industry
profit generated from the change in the number of firms is equally divided between
1
Subsection 8.4.2 can be presented using diagrams only. However, the first printing of the first edition lacks a
justification why firm 1’s (high and low) best-response functions do not shift with a change in
ˆ
k. In the second
printing I intend to add an appendix at the end of this chapter which derives both best-response functions. Also,
note that the equilibrium first-period capital level in Figure 8.7 is
ˆ
k
3
and not
ˆ
k
2
as wrongly stated in the first
printing.
30 Concentration, Mergers, and Entry Barriers
the two unmerged groups of firms. Thus, this share of the change in aggregate
industry profit is lower than the profit each firm makes under the triopoly.
In contrast, merging into a monopoly can only increase profit since a monopoly can
always produce like a duopoly, and under monopoly, all industry profit is equally
divided between the three firms.
3. (a) We look for a Nash equilibrium in prices.
2
The X-seller takes p
Y
as given and solves
max
p
X
π
X
= p
X
Q = p
X
(α −p
X
−2p
Y
), yielding p
X
=
α
2
−p
Y
.
The Y -seller takes p
X
as given and solves
max
p
Y
π
Y
= p
Y
2Q = 2p
Y
(α −p
X
−2p
Y
), yielding p
Y
=
α −p
X
4
.
Hence,
p
N
X
=
α
3
, p
N
Y
=
α
6
, p
N
S
=

3
, Q
N
=
α
3
, π
N
X
= π
N
Y
=
α
2
9
.
(b) We define a system as one unit of X bundled with two units of Y . Then, the merged
firm chooses a system price p
S
that solves
max
p
S
π
S
= p
S
(α −p
S
) yielding p
S
= Q
S
=
α
2
, and π
S
=
α
2
4
.
(c) To make a welfare judgment on the gains from this merger it is sufficient to compare
prices and profit levels. Before the merger, the price of one system is p
0
S
= p
X
+
2p
Y
= 2α/3 > α/2 = p
1
S
which is the price after the merger. Also, before the
merger, aggregate profit is π
X

Y
= 2α
2
/9 < α
2
/4 = π
1
S
which is the profit after
the merger takes place. Since the system price falls and industry profit increases, the
merger is welfare improving.
4. In view of Definition 8.1, Figure 8.1 shows that the pair (p
I
, q
I
) = (20, 10) constitutes a
contestable-market equilibrium.
Formally, the (incumbent’s) industry configuration (p
I
, q
I
) = (20, 10) is feasible since
demand equals supply and the incumbent makes a nonnegative profit. It is also sustainable
since no potential entrant can undercut the incumbent’s price while making a positive
profit.
2
A typo in the first printing of the first edition: the question should state that Q = x = y/2.
Concentration, Mergers, and Entry Barriers 31

q
I

p
I

20
10
p = 60 −4Q
AC(q
I
) =
100
q
+q

Solution-Figure 8.1: Contestable-markets equilibrium
Chapter 9
Research and Development
Most of the topics in this chapter are not technically demanding (and less tedious in terms
of technical derivations). The classifications of process innovation (section 9.1) uses only de-
mand and supply diagrams. Innovation race (section 9.2) relies on simple discrete probability
calculations. Cooperation in R&D (section 9.3) relies on a Cournot market structure in the
second stage and a Nash equilibrium in R&D in the first stage. The calculation of the optimal
patent life (section 9.4) does not reach a closed-form solution. However Figure 9.3 enables the
instructor to explain the cost and benefit of the patent system without performing any calcu-
lations. Then, the instructor can simply reproduce the government-innovation two stage game
and demonstrate how to build the social welfare function for calculating the optimal patent
life. Licensing of innovation (section 9.5) uses supply and demand analysis. The international
analysis (section 9.6) is divided into a simple matrix game of product innovation and a two stage
subsidy Cournot game of process innovation. Finally, the appendix sections discuss patent law
and R&D joint ventures (sections 9.7 and 9.8) and can be assigned as homework reading.
Answers to Exercises
1. Using Definition 9.1, innovation is major if the monopoly price is below the initial unit
cost; that is if p
M
< c
0
. Innovation is minor if p
M
> c
0
. To find p
M
, the monopoly solves
MR(Q
M
) = a − 2Q
M
= c
1
= 2c
0
− a. Therefore, the monopoly’s profit maximizing
output is Q
M
= a − c
0
. Hence, p
M
= c
0
. Therefore, according to Definition 9.1 the
present innovation is neither major nor minor.
2. (a) The expected profit of a firm engaged in R&D, given that the other two firms are
also engaged in R&D, is the prize times the sum of the probability that the firm
discovers while the other two do not, plus twice the probability that it discovers
while one competing firm discovers and one does not, plus the probability that all
the three discover simultaneously, minus innovation cost. Formally, for each firm i,
i = 1, 2, 3,

i
=
1
2
1
2
1
2
V +
1
2
1
2
1
2
V
2
(×2) +
1
2
1
2
1
2
V
3
−I =
7V
24
−I.
Thus, all the three firms find it profitable to engage in R&D if V > 24I/7 = 24/7
(since I = 1).
(b) The now single firm can operate zero, one, or two labs.
Operating a single lab: In this case, Eπ|
1 lab
= V/2 −I.
Operating two labs: In this case, the probability of discovery is one minus the
probability that none of the labs discovers. Thus,
Eπ|
2 labs
=

1 −
1
2
1
2

V −2I =
3V
4
−2I.
34 Research and Development
Now, Eπ|
2 labs
> Eπ|
1 lab
if V > 4I, which constitutes a sufficient condition for
having the firm choosing to operate two labs.
3. (a) The probability that a firm does not discover is (1 −α). Hence, the probability that
none of the n firms discovers at a particular date is (1 −α)
n
.
(b) Clearly, one minus the probability that none discovers. Hence, 1 −(1 −α)
n
.
(c) Similar to the derivation given in (9.5),
ET(n) = [1 −(1 −α)
n
] ×1 + (1 −α)
n
[1 −(1 −α)
n
] ×2
+ (1 −α)
2
n[1 −(1 −α)
n
] ×3 + (1 −α)
3
n[1 −(1 −α)
n
] ×4 +· · ·
Therefore,
ET(n) = [1 −(1 −α)
n
]

¸
t=1
[(1 −α)
n
]
t−1
×t
(Lem 9.1)
=
1 −(1 −α)
n
[1 −(1 −α)
n
]
2
=
1
α(2 −α)
.
4. (a) If the EC provides Airbus with a subsidy of 10 (contingent on developing the megacar-
rier, then Airbus’ dominant action is to develop.
(b) If the U.S. government provides Boeing with a subsidy of 10 (contingent on devel-
oping the megacarrier, then Boeing’s dominant action is to develop. Note that this
U.S. subsidy is independent whether the EC provides Airbus with a subsidy of 10 or
15.
(c) None! After the EC subsidy to Airbus is implemented, Airbus’ dominant action is to
develop (hence, independent of the U.S. subsidy to Boeing).
(d) Like any other welfare argument in economics, there are pros and cons to having
both governments subsidizing the development of the new megacarrier. Arguments
against the subsidies include (i) duplication of R&D which require twice the amount
of resources needed to develop and test the plane, (ii) the government may not
possess the means to determine whether the benefits from the development to their
own economy dominate the cost, (i.e., economists generally believe that the private
sector can collect more accurate information), and (iii) income distribution effects;
that is, some economists claim that the development of the Concord by some EC
countries constitute a net transfer from poor to rich people, since most poor and
middle-class people do not get to fly in the Concord.
Arguments for the subsidy include the introduction of more competition between
aircraft manufacturers that would reduce aircraft prices, boost the airline industry,
and may therefore increase the amount of travel.
Chapter 10
The Economics of Compatibility and Standards
The introduction to this chapter provides some basic terminology and a simple “Battle of the
Sexes” game of product standardization, which can be taught to the less technically proficient
students. The behavior of a monopoly telephone company under network externalities (Sub-
section 10.1.1) uses very simple calculus (basically one first-order condition). The remainder of
section 10.1 is devoted to a simple discrete (noncalculus) standardization-variety tradeoff model.
The remaining two sections, the supporting services approach (section 10.2) and the compo-
nents approach (section 10.3) are somewhat more difficult despite the fact that no calculus is
used. In particular, whereas the components approach model is very simple, some students may
get confused because of the existence of multiple equilibria under incompatibility, in which case,
the instructor may want to construct only one equilibrium and compare it to the equilibrium
under compatible systems.
The important points to emphasize are:
• A network externality is a particular type of a consumption externality, where a consumer’s
utility is affected by the number of consumers purchasing the same brand. Thus, network
externalities constitute an assumption about consumer preferences.
• The supporting services and the components approaches do not assume any externality.
The “network effects,” where the utility of a consumer is affected by the brand choices of
other consumers is an equilibrium result (rather than an assumption).
1
Thus, the network
effects are generated by having hardware and software (or other types of hardware) treated
as perfect complements.
Answers to Exercises
Unfortunately, the two exercises cover only the supporting services and the components ap-
proaches. In future printings/editions, I intend to add two exercises covering the two topics
analyzed in the network externalities section.
1. (a) Firm A gains control over the entire market when
ˆ
δ = 1. In this case, by (10.10),
p
B
has to be sufficiently high to satisfy
1 =
ˆ
δ =
Y −p
A
2Y −p
A
−p
B
, or p
B
= Y.
(b) i. Let
¯
δ denote the market share for firm A after income has doubled from Y to
2Y . Then, by (10.10) we have that
¯
δ =
2Y −p
A
−p
B
4Y −p
A
−p
B
>
ˆ
δ =
Y −p
A
−p
B
2Y −p
A
−p
B
1
Some authors use the term indirect network externalities to describe the behavior generated by the supporting
services and the components approaches. I feel that this term was poorly chosen since externalities are either
assumed or not assumed at all, and therefore the term indirect cannot be used to describe models that do not
explicitly assume any externality.
36 The Economics of Compatibility and Standards
if p
A
> p
B
. This result can also be demonstrated by differentiating
ˆ
δ with
respect to Y , yielding
d
ˆ
δ
dY
=
p
A
−p
B
(2Y −p
A
−p
B
)
2
> 0 if p
A
> p
B
.
Thus, the firm with the lower market share increases its market share when
consumer income rises.
ii. Since
ˆ
δ increases, by (10.9) N
A
/N
B
must increase.
2. (a) Potential equilibria can be classified into two types: (i) Firm A sells to consumers AA,
and AB; and firm B sells to BB and BA (alternatively, Firm A sells to consumers
AA, and BA; and firm B sells to BB and AB). (ii) Firm A sells to AA, AB, and
BA; whereas firm B sells to BB only (alternatively, firm A sells to AA only, whereas
firm B sells to BB, AB, and BA).
In what follows we will demonstrate, by a way of contradiction, that both types of
equilibria do not exists.
i. Under this type, p
I
A
= p
I
B
since consumers AB and BA are indifferent between
systems AA and BB as long as the prices are equal. However, this prices
cannot support a Bertrand-Nash equilibrium since firm A can increase its profit
by undercutting firm B by charging p
A
= p
I
B
− which will cause consumer BA
to switch to system AA (this is the usual Bertrand undercutting argument).
ii. Under this type, we must have that p
I
B
= p
I
A
+ 2λ (otherwise firm B is not
maximizing its profit). However, since consumers’ reservation utility is zero, see
equation (10.12), equilibrium prices must satisfy p
I
B
≤ 2λ, hence, p
I
A
= 0 and
π
I
A
= 0. However, firm A can always increase its profit by raising its price by
thereby making strictly positive profit.
(b) Following the proof of Proposition 10.14, it can be established that there exists an
equilibrium in which each consumer buys his or her ideal system. In this equilibrium,
p
x
A
= p
y
A
= p
x
B
= p
y
B
= λ, and π
c
A
= π
c
B
= 4λ.
Chapter 11
Advertising
From a technical point of view, the analyses given in this chapter are pretty much straight
forward, perhaps, with the exception of the calculation of the socially optimal level of persuasive
advertising given in equation 11.13, where an integral is computed. The appendix describing
advertising regulations can be assigned as a home reading.
Answers to Exercises
1. (a) Since

A
=
%∆Q
%∆A
= 0.05, and
p
=
%∆Q
%∆p
= −0.2,
then by the Dorfman-Steiner condition (Proposition 11.1), we have that the profit
maximizing ratio of advertising to revenue is
A
M
pQ
M
=

A

p
=
1
4
.
Hence, A
M
= $2.5 mil..
(b) Now,
p
= −0.5. Therefore A
M
= 10 ×0.05/0.5 = $1 mil..
(c) When the demand becomes more price elastic, the monopoly reduces the ratio of
advertising expenditure to revenue.
2. (a) A
α
= A
β
= 1 implies that n
α
= n
β
= 3, hence, π
α
= π
b
= 30 − 1 = 29.
When A
α
= A
β
= 2, we have it that n
α
= n
β
= 3, hence, π
α
= π
b
= 30 −2 = 28.
Conclusion: if both firms spend the same amount on advertising, then industry profit
is maximized at A
α
= A
β
→0. That is, advertising in this city serves as a mean by
which each firm attracts consumers from the competing firm, but cannot increase
industry sales beyond the level determined by zero advertising.
(b) Fortune teller β takes A
α
as given and chooses A
β
to maximize his or her profit.
The first-order condition is given by
0 =
∂π
β
∂A
β
=
30A
α
(A
β
)
2
−1.
The second-order condition can be easily verified. Hence, A
β
= R
β
(A
α
) =

30A
α
.
By symmetry, A
α
= R
α
(A
β
) =

30A
β
. The two best-response functions are drawn
in Solution-Figure 11.1.
(c) Solution-Figure 11.1 illustrates that the two best response functions intersect twice,
reflecting the fact that there exist two Nash equilibria for this advertising game. For
the purpose of this question, we define 0/0 (yes, zero divided by zero) as equal to 1.
38 Advertising

............................... .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A
α
A
β
0 30
30
R
β
(A
α
)
R
α
(A
β
)




Solution-Figure 11.1: Fortune tellers’ advertising best-response functions
Then, clearly A
α
= A
β
= 0 constitutes one NE. To obtain the other NE, substituting
R
α
(A
β
) into R
β
yields A
β
=

30

30A
β
. Hence, A
N
α
= A
N
β
= 30 which is also
illustrated in Solution-Figure 11.1.
(d) In the interior NE, π
N
i
(30, 30) = 30 − 30 = 0, i = α, β, which is lower than
π
N
i
(1, 1) = 29. Thus, the industry’s profit is not maximized at the interior NE. How-
ever, at the other NE, industry’s profit is maximized and each firm earns π
N
i
(0, 0) =
30.
(e) The interior NE (30, 30) is stable. The interior NE (0, 0) is unstable since any
unilateral deviation by one of the firm would trigger a sequential chain of responses
that would move the advertising equilibrium to (30, 30) (see Solution-Figure 11.1
for an illustration). Thus, in this example, the NE that maximizes industry profit is
unstable.
3. In order for (I, P) to constitute a NE outcome, we need to show that unilateral deviation
is not profitable for each firm. That is, using Table 11.1,
π
2
(I, P) = N ≥ π
1
(I, I) = (1 −θ)E when θ ≥ 1 −N/E;
π
1
(I, P) = θE ≥ π
1
(P, P) = N/2 when θ ≥ N/(2E).
Chapter 12
Quality, Durability, and Warranties
This is a long chapter with a wide variety of topics dealing with supply and demand for quality.
The analyses are pretty much straight forward and do not require any preparation with the
exception of section 12.2 that relies on the knowledge of location models developed in Chapter 7.
Other sections use only simple calculations of expected values. The appendix on the legal
approach to liability (section 12.8) may be assigned as a home reading.
An Alternative Model for Section 12.6
The model suggested below provides some improvement as it assumes that consumers are ex-
pected utility maximizers (similar to what is assumed in Section 12.5).
1
Suppose that consumers
believe that the monopoly is a high-quality producer with probability α, and a low-quality pro-
ducer with probability 1 − α, where 0 < α < 1. Then, without any additional information,
EU = αH + (1 −α)L −p. Therefore, the maximum price that consumers are willing to pay is
¯ p = αH +(1 −α)L. Then a low-quality producer could set the price at ¯ p, and make a profit of
(¯ p −c
L
) · 1 = αH + (1 −α)L −c
L
= αH + (1 −α)L −c
L
> 0. (∗)
The price and quantity pair in Proposition 12.10 should be modified to
p
m
= H and q
m
=
αH + (1 −α)L −c
L
H −c
L
.
Had the monopolist been a low-quality producer, he or she could have made the same profit by
setting ¯ p and selling to all consumers [as is indicated in (*)] rather than setting p = p
m
and
q = q
m
. Therefore, in choosing this price-quantity pair the monopolist could signal his or her
quality to the consumers. Finally, since H > L > c
H
> c
L
,
(H −c
H
)
αH + (1 −α)L −c
L
H −c
L
> αH + (1 −α)L −c
H
= (¯ p −c
H
) · 1.
Hence, the high-quality monopoly would indeed find it profitable to signal its high quality to the
consumers.
1
I thank Keiichi Koda for proposing this improvement.
40 Quality, Durability, and Warranties
Answers to Exercises
1. (a) Using the utility function (12.2) and the zero reservation utility assumption, the
consumer who is indifferent between purchasing from A and not purchasing at all
must satisfy U
z
(A) = az −p
A
= 0, yielding z = p
A
/a.
(b) Follows directly from (12.4).
(c) For a given p
B
, firm A chooses p
A
to
max
p
A
π
A
= p
A
¸
p
B
−p
A
b −a

p
A
a

, yielding 0 =
p
B
−2p
A
b −a

2p
A
a
.
For a given p
A
, firm B chooses p
B
to
max
p
B
π
B
= p
B
¸
1 −
p
B
−p
A
b −a

, yielding 0 = 1 −
2p
B
−p
A
b −a
.
Second-order conditions are easily obtained. Solving the two first-order conditions
yield the equilibrium prices. Substituting into the profit functions yield the equilibrium
profit levels.
(d) We first verify that given b
e
= 1, the profit maximizing location of firm A is a
e
= 4/7.
Thus, given b
e
= 1, firm A chooses a
e
to
max
a
π
a
=
a(1 −a)
(4 −a)
2
, yielding a
e
=
4
7
.
We now verify that given a
e
= 4/7, firm B would choose to locate on the east edge
of the street.
2
For a given a
e
firm B chooses b
e
to maximize π
B
which is given
above. We would like to show that for a
e
= 4/7,
0 <
∂π
B
(4/7, b)
∂b
≈ 49b
2
−21b + 8 ≡ φ(b).
To demonstrate this inequality, we need to prove that φ(b) > 0 for all b ∈ [0, 1].
Clearly φ

(b) = 98b − 21, which is negative for 0 ≤ b < 21/98 and positive for
b > 21/98. Hence, the function φ hits a minimum at b = 21/98. Now, φ(21/98) =
5.75 > 0. Hence φ(b) = ∂π
B
/∂b > 0 for all b ∈ [0, 1] implying that firm B would
locate at b
e
= 1.
2. Let U
i
(k) denote the utility of consumer i when he buys the brand with quality k, k = H, L
and i = 1, 2. We want to show that
U
2
(L) = L(I
2
−p
L
) > H(I
2
−p
H
) = U
2
(H).
From (12.1), we have it that since consumer 1 buys the low-quality brand, then
U
1
(L) = L(I
1
−p
L
) > H(I
1
−p
H
) = U
1
(H).
Therefore,
(H −L)I
1
< Hp
H
−Lp
L
.
Now, since I
2
< I
1
by assumption, (H−L)I
2
< Hp
H
−Lp
L
, implying that H(I
2
−p
H
) <
L(I
2
−p
L
).
2
Many students will find this part of the proof to be difficult.
Quality, Durability, and Warranties 41
3. Under these prices, we seek to characterize the demand and supply patterns of our agents.
3
New buyers (who do not yet own a car): Recalling our assumption that N
L
= U
L
=
0, if a new buyer buys a new car, then his or her expected utility is V
b
= 0.5N
G
+0.5N
L

¯ p
N
= 0. In contrast, if a new buyer buys a used car then V
b
= 0.5U
G
+0.5U
L
− ¯ p
U
> 0.
Hence, under the assumed prices, new buyers buy used cars.
Good-used-car sellers: The question simply assumes that good-used-car owners must
leave the country and therefore sell their good-used cars.
Lemon-used-car sellers: If a lemon-used-car seller sells his or her lemon and buys a new
car, then V
s,L
= 0.5N
G
− ¯ p
N
+ ¯ p
U
= ¯ p
U
. In contrast, if she or he buys a used car then
V
s,L
= 0.5U
G
− ¯ p
U
+ ¯ p
U
= 0.5U
G
. By assumption ¯ p
U
< 0.5U
G
, hence, lemon-used-car
sellers buy (and sell) used cars.
Altogether, under the assumed prices, nobody buys a new car.
4. (a) The monopoly’s production cost is the initial production cost plus the expected (as
most once) replacement cost, which is equal to c + (1 −ρ)c = (2 −ρ)c.
(b) The maximum monopoly price is the expected consumer surplus from purchasing
this product with this type of warranty. This surplus equals V if the product does
not fail at all and if the product fails only once (in which case it is replaced). The
probability of a double failure is (1 −ρ)
2
. Hence, the probability of no double failure
is 1 −(1 −ρ)
2
= ρ(2 −ρ). Therefore, p
M
= ρ(2 −ρ)V .
(c) If the monopoly provides this warranty, then π
W
= ρ(2 − ρ)V − (2 − ρ)c = (2 −
ρ)(ρV − c). If the monopoly sells with no warranty, then π
NW
= ρV − c. Hence,
π
W
> π
NW
since it is assumed that ρV > c. Thus, the monopoly will bundle the
product with this warranty.
3
The purpose of this question is demonstrate that used cars are traded when some good-used-car owners are
“forced” to sell their car, say, because they leave the country. Note: If you are using the first printing of the first
edition, change the last sentence in this question to “. . . the four types of agents. . . .”
Chapter 13
Pricing Tactics: Two-Part Tariff and Peak-Load Pricing
The analyses given in this relatively short chapter rely mainly on logical arguments and therefore
do not require any special technical preparation. Section 13.4 presents an attempt to endogenize
the seasons in determining profit maximizing peak-load pricing, and therefore may be given a
lower priority in case the instructor is short of time.
1
Answers to Exercises
1. (a) The two cost functions are drawn in Solution-Figure 13.1.
✲ ˜
k (# copies)

$
TC(k) = 0.05k
TC(k) = 300 + 0.02k

˜
k
300
Solution-Figure 13.1: The cost of photocopying with or without two-part tariff
(b) The number of copies yielding identical cost under the two payment systems is found
by equating 0.05
ˆ
k = 300 + 0.02
ˆ
k. Therefore,
ˆ
k = 10, 000 copies. Thus, for k >
ˆ
k,
the two-part tariff system is less costly, and for k <
ˆ
k, the fixed-per-unit price system
is less costly.
2. (a) By Proposition 13.4, to find the profit maximizing pricing structure we need (i) to
equate the “high-season” marginal revenue to the cost of an aircraft seat plus the
per-passenger cost, and (ii) to equate the “low-season” marginal revenue to the
per-passenger cost only.
Given that the Summer demand curve lies completely below the Winter demand
curve, we can conclude that Winter is the high season. However, since the two
demand curves are “relatively close,” we need to be careful about the capacity con-
straint determined by the Winter condition.
In the Winter, capacity is determined by having the monopoly equating
MR
W
(q
W
) = 10 −2q
W
= r +c = 2, and MR
S
(q
S
) = 5 −q
S
= c = 1.
1
If you are using the first printing of the first edition note that the calculations of consumer surplus in
section 13.2 are wrong. Ask for the errata file by e-mail.
44 Pricing Tactics: Two-Part Tariff and Peak-Load Pricing
The first equality implies that q
W
= 4, hence, p
W
= 6. The second equality implies
that q
S
= 4, hence, p
S
= 3. Altogether, π
I
= 6 ×4 + 3 ×4 −2 ×4 −1 ×4 = 24.
(b) Similar to the previous subquestion, the monopoly equates
MR
W
(q
W
) = 10 −2q
W
= r +c = 4, and MR
S
(q
S
) = 5 −q
S
= c = 1.
The first equality implies that, q
W
= 3, hence, p
W
= 7. The second equality implies
q
S
= 4 > 3 = q
W
. Hence, due to the capacity constraint, the airline will fly only
3 passengers during the Summer season. Thus, p
S
= 5 − 3/2 = 3.5. Altogether,
π
I
= 7 ×3 + 3.5 ×3 −4 ×3 −1 ×3 = 16.5.
Chapter 14
Marketing Tactics: Bundling, Upgrading, and Dealerships
This chapter introduces several topics related to profit enhancing marketing tools that were not
discussed (at least not in full) in earlier chapters.
Answers to Exercises
1. (a) Under pure tying, we need to check two options: (i) Under p
T
= 4, all three con-
sumers buy the packages, hence, π
T
(4) = 3 × 4 = 12. (ii) Under p
T
= 10, only
consumer 2 buys a package, hence, π
T
(10) = 10. Clearly, the first option maximizes
the profit under pure tying.
(b) The profit maximizing product and package prices are p
MT
X
= p
MT
Y
= 4 and p
MT
=
8. Under this price structure, consumer 1 buys one unit of X, consumer 3 buys one
unit of Y and consumer 2 buys the package (or one unit of each good). Altogether,
π
MT
= 4 + 4 + 8 = 16.
(c) Notice that mixed tying yields the same profit maximizing pricing structure as no
tying, since in the present example, the monopoly cannot charge consumer 2 more
than the sum of the prices of the two products sold separately. In fact, in the
present example, pure tying is profit reducing. Therefore, mixed tying (which yields
an identical allocation as no tying) generates a higher profit than pure tying.
2. (a) We need to compare the profit levels under the commitment of not introducing a
new edition, in which case the publisher can charge a higher price in the first period
by adding the resale value to the price, to the profit level when no commitment is
made and the monopoly introduces a new edition.
By (14.17), the profit level under commitment is nV , and under no commitment is
2n(V − c) − F. Comparing the two profit levels reveals that commitment not to
introduce a new edition is more profitable if F > nV −2c.
(b) Indeed, this condition in weaker (i.e., implied by) the condition given in (14.17) since
it compares in the case of commitment, in which first-period students believe that
no new edition will be published, hence the monopoly can raise the book’s price
by its resale value, to the case where there is no commitment and a new edition is
introduced.
In contrast, the condition given in (14.16) and hence in (14.17) is derived by com-
paring the monopoly’s second period profit only and therefore yields the condition
under which a new edition will not be introduced in the second period. In other
words, the condition given in (14.16) and (14.17) “does not take into account” that
first period price can be raised by the resale value when a commitment is made not
to introduce a new edition.
3. (a) When a new edition is not introduced in the second period, the monopoly competes
with period 1 students on selling the first edition. First period students will undercut
46 Marketing Tactics: Bundling, Upgrading, and Dealerships
the monopoly (since first-period students do not have any production cost) and for
a given p
N
2
will set p
U
2
so that
V −p
N
2
≤ αV −p
U
2
.
Now, in a Bertrand equilibrium, if used-book sellers can undercut the monopolist,
then the monopoly price is reduced to cost( p
N
2
= c). Hence, the maximum price
user-book sellers can charge is
p
U
2
=

c −(1 −α)V if c ≥ (1 −α)V
0 otherwise
Notice that if c < (1 − α)V , or α < 1 − c/V , then the monopoly undercuts the
used-book sellers (since consumers do not value used books very much). Under this
condition, the monopoly would set p
N
2
= (1 − α)V , and therefore earn a profit of
π
N
2
= n[(1 −α)V −c].
(b) There are two cases:
High value for used books (α > 1 −c/V ): A new edition would yield a second-
period profit of π
N
2
= n(V − c) − F. When a new edition is not introduced, then
π
2
= 0. Hence, a new edition is introduced if F ≤ n(V − c), which is the same as
condition (14.16) when α = 1.
Low value for used book (α < 1 − c/V ): When α is low, used book sellers
are undercut by the monopoly. In this case, if a new edition is not introduced in
the second period, then p
N
2
= (1 − α)V − c (the monopoly undercuts used-book
sellers). Hence, π
N
2
= n[(1 − α)V − c]. Altogether, a new edition is introduced
if n(V − c) − F ≥ n[(1 − α)V − c], or if F ≤ nαN, which is different from
condition (14.16).
4. (a) Under this contract, the dealer chooses Q (sales) to
max
Q
π
d
= (1 −φ)[p(Q)Q−cQ] = (1 −φ)[(a −Q)Q−cQ].
The first-order condition yields Q
d
= (a − c)/2, p
d
= (a + c)/2, and π
d
= (1 −
φ)(a − c)
2
/4. Thus, the dealer’s sales and price are independent of (1 − φ) (share
of the profit of the dealer). Under this contract, the industry profit is
π
M

d
= φ
(a −c)
2
4
+ (1 −φ)
(a −c)
2
4
=
(a −c)
2
4
,
which equals the industry’s profit when the firms vertically integrate into a monopoly
given in (14.24).
(b) Under this contract, the dealer chooses Q that solves
max
Q
π
d
= (1 −φ)p(Q)Q−cQ = (1 −φ)(a −Q)Q−cQ.
Marketing Tactics: Bundling, Upgrading, and Dealerships 47
The first-order condition is (1 −φ)(a −2Q) −c = 0, hence,
Q
d
=
a
2

c
2(1 −φ)
, and p
d
=
a
2
+
c
2(1 −φ)
.
Now, industry profit is given by the sum
π
M

d
=

φQ
d
p
d

+

(1 −φ)Q
d
p
d
−cQ
d

=

p
d
−c

Q
d
=

a
2
+
c −2(1 −φ)c
2(1 −φ)

a
2

c
2(1 −φ)

=

a −c
2

a
2

c
2(1 −φ)

<
(a −c)
2
4
.
Thus, this contract yields a lower profit than the industry’s profit when the firms
vertically integrate into a monopoly given in (14.24).
(c) Whereas the “profit sharing” contract analyzed in part (a) is optimal (i.e., maxi-
mizes industry’s profit) it is rarely observed since it requires having the manufacturer
monitoring the profit made by the dealer. This monitoring may be too costly for the
manufacturer. In contrast, the per-unit contract analyzed in the text, despite being
not optimal, does not require any monitoring since the dealer pays up front for each
unit it acquires from the manufacturer.
5. We will demonstrate how a monopoly can benefit from establishing a trade-in procedure
by constructing a simple model. Suppose that there are two consumers: one owns an old
refrigerator and one that is a new buyer. The utility from using an old refrigerator is V
U
whereas the utility from buying a new one is max{V
N
−p
N
, 0}, where V
N
> V
U
> 0. We
further assume that V
N
< 2V
U
.
Single price policy: I will argue that the profit maximizing price is p

N
= V
N
, thereby
selling only to the first-time buyer and earning a profit of π

= V
N
. To see this, note that
the monopoly can set a lower price, p
L
= V
N
−V
U
, thereby selling two units and earn a
profit of π
L
= 2(V
N
−V
U
). However, by assumption, π
L
< π

.
Trade-in price policy: We now show that by employing a trade-in price policy, the
monopoly can price discriminate between new buyers and old-model owners. Suppose
that the monopoly announces that current users can trade-in their old refrigerator for a
new one for a price of p
TI
= V
N
−V
U
, whereas the price with no trade-in is p
N
= V
N
. Note
that under these prices, current users trade-in their old refrigerator for a new one, and also
a new buyer buys a new one. Altogether, π
TI
= V
N
−V
U
+V
N
= 2V
N
−V
U
> V
N
= π

.
Chapter 15
Management, Compensation, and Regulation
Instructors who choose to teach from this chapter are urged to go over the principal-agent
problem (section 15.1), which is developed three times in order to allow for a a gradual in-
crease in the degree of difficulty. More precisely, the first subsection illustrates the problem
with no uncertainty. A subsequent subsection introduces uncertainty with no risk aversion.
The third subsection of introduces risk aversion implicitly by assuming asymmetric information.
Altogether, these three subsections analyze the principal-agent problem in different degrees of
complexity, thereby enabling the instructor to stop at the level that exceeds the students’ ability.
Section 15.2 introduces a simple free-rider team effort problem. Section 15.3 introduces
Fershtman-Judd model, which is recommended only if the students are capable of solving
Cournot-type problems very easily. The remaining sections: executives’ compensation (sec-
tion 15.4) and the regulation of the firm (section 15.4) do fit the more advanced students.
Answers to Exercises
1. (a) Under collusion, the “social planner” chooses a common effort level for all scientists
e that solves
max
e

w −e
2

=
V
N
−e
2
=
Ne
N
= e −e
2
,
yielding e

= 1/2.
(b) Each scientist i takes all other effort levels (e
j
, j = i) as given and solves
max
e
i
U
i
=
V
N
−(e
i
)
2
=
¸
j=i
e
j
+e
i
N
−(e
i
)
2
,
yielding a NE where e
i
= 1/(2N) for all i = 1, 2 . . . N.
(c) Yes, since the optimal effort level is e

= 1/2 which is independent of the number of
scientists. In constrast, the NE effort levels decrease further below the optimal level
when the number of scientists increases, reflecting a stronger free-rider effect.
2. (a) The manager of firm 1 chooses q
1
that solves
max
q
1
M
1
= 0.5[(a −q
1
−q
2
)q
1
+ (1 −c)q
1
],
yielding a best-response function given by
q
1
= BR
1
(q
2
) =
a −c + 1 −q
2
2
.
The best response function of firm 2 (the firm that maximizes profit only) is given
by q
2
= BR
2
(q
1
) = 0.5(a − c − q
1
). Therefore, the equilibrium output levels are
given by
q
e
1
=
a −c + 2
3
, and q
e
2
=
a −c −1
3
.
50 Management, Compensation, and Regulation
Clearly, q
e
1
> q
e
2
confirming that the managerial compensating scheme of firm 1 is
output increasing at the expense of the firm that maximizes profit only (firm 2).
(b) The owner of firm 1 earns π
O
1
= π
1
−M
1
. From the previous calculation we conclude
that p
e
= (a − 2c − 1)/3. Hence, π
e
1
= (a − c − 1)(a − c + 2)/9. We now set µ
1
sufficiently small (as in [15.29]), so that we approximate the owner’s profit by the
firm’s profit. Formally, π
O
1
= π
1
.
Finally, comparing the profit of the owner of firm 1 under the present managerial
compensation scheme to her profit when this manager maximizes profit only (the
simple Cournot profit given in (6.7)) yields that π
O
1
≥ π
c
1
if and only if a ≥ 2 + c.
That is, the present compensation scheme dominates a simply Cournot behavior
(given that the other firm behaves in a Cournot fashion) if the demand intercept is
sufficiently higher than the unit production cost.
Chapter 16
Price Dispersion and Search Theory
Price dispersion (section 16.1) demonstrates that a discount store and an ‘expensive’ store can
coexist in a market where consumers have different search costs (resulting, say, from different
values of time). The analysis relies on a simple of location model. Thus, the students should
find this section easy to follow. Search theory (section 16.2) is introduced with no calculus, so
the analysis relies on simple probability arguments.
Answers to Exercises
1. (a) To see what happens, we look at the equilibrium prices when L → 0. From (16.9)
we see that
lim
L→∞
p
e
D
= lim
L→∞
p
e
ND
= 2αH and ˆ s
e
→0.
Hence, when L becomes small (i.e., some consumers have a negligible search cost),
both stores charge the same price, implying that there is no discount store.
(b) When α = 3/2, the consumer indifferent between searching or buying at random is
located at L. Hence, when α = 3/2 all consumers buy at random. When α = 1,
ˆ s
e
< L implying that the discount store a negative market share (which is ruled out
by out assumption that α > 3/2).
2. (a) Using (16.13), we have it that
s 0 1 2 3 4 5
¯ p 1 4.77 6.52 7.86 9.00 10.0
Solution-Table 16.1: Reservation price as a function of the search cost parameter
(b) Solution-table 16.1 reveals that for s ≥ 4 the consumer reservation price is ¯ p = 9,
implying that the consumer will accept any price in the first store visit.
(c) Solution-table 16.1 reveals that ¯ p = 1 when s = 0 (i.e., when search cost is zero).
(d) When s = 0,
1
the probability that the consumer finds the price higher than 1 (i.e.,
the reservation price) in a single store visit is σ = 8/9. The probability that prices
exceed 1 in two store visits is σ
2
= (8/9)
2
. Hence, the probability that this consumer
finds prices exceed 1 in all store visits is lim
T→∞
(8/9)
T
= 0.
Note also that this result can be inferred from (16.16) where the calculated expected
number of stores visited is µ = 1/(1 −σ) = 9 < +∞.
1
The first printing of the first edition does not specify which search cost is assumed. This question should
read, under the search cost you found in the previous subquestion, calculate....
Chapter 17
Miscellaneous Industries
The purpose of this chapter is to demonstrate to the student that there is no general model
that can describe all industries; however, I also wish to demonstrate in this last chapter that
the tools developed in this book are indispensable to the analysis of most industries, and are,
therefore, worth learning.
This chapter discusses four types of industries that require special attention since (perhaps
like any other industry) they do not fall in any category and therefore cannot be analyzed by
simply picking one of the market structure analyzed earlier in the book. In future editions I
will attempt to expand this part of the book by introducing at least three more industries: the
health industry, the defense industry, and (you guessed, of course), the lawyers.
Answers to Exercises
1. From the definition, this technology (cost function) exhibits economies of scope if
TC(n, n, n) = (n
α
+n
α
+n
α
)
β
< (n
α
)
β
+ (n
α
)
β
+ (n
α
)
β
= TC(n, 0, 0) +TC(0, n, 0) +TC(0, 0, n),
or,
(3n
α
)
β
< 3n
αβ
, therefore if 3
β
< 3, hence if β < 1.
2. The main message here is that the introduction of a binding price cap by the regulator
will cause the monopoly airline to reduce flight frequency (or any other quality aspect of
the service).
(a) Under the FC network, each route can be analyzed separately. Now, from the utility
function of route 3 passengers (17.2), we see that the monopoly extracts maximum
surplus when p
3
= δ +

f
3
. Since the airfare is regulated (i.e., p
3
= ¯ p
3
) we have it
that the minimum frequency that the airline must provide on route 3 (at this airfare)
is
¯
f
3
= (¯ p
3
−δ)
2
. Clearly, as ¯ p
3
decreases, the corresponding frequency
¯
f
3
decreases.
(b) Under the HS network, the flight frequency on route 3 is the (minimum) frequency
on routes 1 and 2.
1
Hence, similar to (17.5), but with a given ¯ p
3
, the monopoly
chooses
˜
f ≡
˜
f
1
=
˜
f
2
that solves
max
f
π
h
= n¯ p
3
+ 2nδ + 2n

f −2cf,
1
The first printing of the first edition contains a typo: Replace “...charges an airfare
¯
f
3
” with “airfare ¯ p
3
.”
54 Miscellaneous Industries
yielding a first order condition given by 0 = 2n/(2

˜
f) −2c. Hence,
2
˜
f =

n
2c

2
<

3n
4c

2
= f
h
.
Therefore, under the HS network, a price cap on route 3 will reduce the flight
frequency on all routes.
3. The analysis here follows the same stages as in section 17.4, but it uses a different driving
time function.
3
(a) If some people take the train and some drive a car, it must be that L
T
= L
C
. Thus,
v +φ = v(n
C
)
3
, implying that n
e
C
=

v +φ
v

1/3
.
(b) The social planner chooses n
s
C
that solves
minL
s
= (N −n
c
)(v +φ) +n
C
V (n
C
)
3
,
yielding
n
s
C
=

v +φ
4v

1/3
< n
e
C
.
Thus, the social planner would recommend a policy that would result in a reduction
in the number of people who drive a car.
(c) Let τ denote the highway toll on this route. Then, to set the toll that would generate
the socially optimal number of highway users, the social planner sets τ
s
that solves
v +φ = v(n
s
C
)
3
+τ = v

v +φ
4v

+τ,
yielding that τ
s
= 3(v +φ)/4.
2
One has also to check that, in this equilibrium, the airfare on route 3, ¯ p
3
, is lower than the sum of the airfares
on routes 1 and 2; that is, we need to verify that (or find the conditions under which) ¯ p
3
≤ p
1
+ p
2
. Otherwise,
route 3 passengers can purchase two separate tickets and reach their destination at a lower airfare.
3
The first printing of the first edition contains a typo. Simply set t
C
= (n
C
)
3
(the equation stated in this
question equals to the value of the driving time, L
C
and not t
C
).

Copyright c 1995–2004 Oz Shy. All rights reserved.

A The manual was typeset by the author using the LTEX 2ε document preparation software by Leslie Lamport (a special version of Donald Knuth’s TEX program) during the months from September 1995 to November 1995 while I was visiting the Economics Department of the A University of Michigan. All the figures are also drawn in LTEX using software called TEXcad developed by Georg Horn, which can be downloaded from various mainframes.

Version 1.0 [Draft 20], (1995): Prepared for the first printing of the first edition bkman21.tex first draft to be posted on the Web (2000/07/07) (parallel to the 5th Printing of the book) v.24 (2004/04/14): 6.1.d, 6.1.e This version: bkman24.tex 2004/04/14 17:36

Contents
To the Instructor 2 3 4 5 6 7 8 9 v 1 7 Basic Concepts in Noncooperative Game Theory Technology, Production Cost, and Demand Perfect Competition The Monopoly 11 17 25 29 35 9

Markets for Homogeneous Products Markets for Differentiated Products Research and Development 37 33

Concentration, Mergers, and Entry Barriers

10 The Economics of Compatibility and Standards 11 Advertising 12 Quality, Durability, and Warranties 39

13 Pricing Tactics: Two-Part Tariff and Peak-Load Pricing 14 Marketing Tactics: Bundling, Upgrading, and Dealerships 15 Management, Compensation, and Regulation 16 Price Dispersion and Search Theory 17 Miscellaneous Industries 53 51 49

43 45

.

(e-mail addresses are given below).com OzBackup@yahoo.tcl?isbn=0262691795 bkman24. Before reporting errors and typos. • to convey to the instructor my views of what the important concepts in each topic are. (April 14. Haifa. • to suggest which topics to choose for different types of classes and levels of students. Israel. I urge the instructor to read carefully the Preface of the book that suggests different ways of organizing courses for different levels of students and also provides a list of calculus-free topics. 2004) E-mail Backup Web-page Catalog-page This draft : : : : : ozshy@ozshy.mit. Finally. please view the errata files to see whether the error you found was already identified and corrected.ozshy. please alert me to any errors or incorrect presentations that you detect in the book and in this manual.edu/book-home. The goals of this manual are: • To provide the instructor with my solutions for all the problems listed at the end of each chapter.com www.To the Instructor Before planning the course.tex 2004/04/14 17:36 . Note that the errata files (according to the printing sequence) are posted on the Web in PDF format.com http://mitpress.

.

but RR (ω) = φ.Chapter 2 Basic Concepts in Noncooperative Game Theory An instructor of a short course should limit the discussion of game theory to the four most important concepts in this chapter that are essential for the understanding most of the analyses presented in this book: 1.3) before teaching the NE equilibrium concept.4). and the SPE (Definition 2. 2. 4. but RJ (φ) = φ. Answers to Exercises 1. . strategies (compare with actions). and so on. uniqueness. subgames. (a) It is straightforward to conclude that R1 (a2 ) = WAR if a2 = WAR WAR if a2 = PEACE and R2 (a1 ) = WAR if a1 = WAR WAR if a2 = PEACE. There does not exist a NE for this game since there does not exist an outcome the is on both best-response functions. Extensive form games. WAR is each player’s best response to each action taken by the other player (hence. Welfare comparisons among outcomes (Definition 2. Now. (b) That is. you can introduce the equilibrium in dominant actions (Definition 2.1): It is important that the student will understand that a game is not properly defined unless the list of players. The definition of a game (Definition 2. I advise covering repeated games (section 2. but RR (φ) = ω. WAR) a ˆ is a (unique) NE since this outcome is on the best-response function of each player.3). and multiple equilibria in class. I believe that the above can be covered in 2 lectures. More precisely. and the payoff functions are clearly stated. If you wish to emphasis more game theory. a2 ) = (WAR. (ˆ1 . 3.10). That is. it is important that the student will know that in order to prove existence. WAR is a dominant action for each player). It is important that the student will understand the meaning of the term outcome as a list of the specific actions chosen by each player (and not a list of payoffs as commonly assumed by students). RJ (aR ) = ω if aR = ω φ if aR = φ and RR (aJ ) = φ if aJ = ω ω if aJ = φ. I urge the instructor to discuss the issues of existence.6). however. If you wish to devote more time to game theory. so RJ (ω) = ω. it is sufficient to find only one NE outcome. Nash equilibrium (Definition 2. the student must go over all outcomes and show that at least one player benefits from unilateral deviation. RJ (ω) = ω. to prove nonexistence. or in three hours of instruction. the action set of each player.

n2 ) = (100. 2. R) if (a ≥ g and b > h) or (a > g and b ≥ h). L) and π α (T. that is π β (T.2 (c) Rα (aβ ) = Basic Concepts in Noncooperative Game Theory B if aβ = L T if aβ = R and Rβ (aα ) = L if aα = T R if aα = B. K. L) = a ≥ e = π α (B. we have to show that for every outcome (n1 . L) = f ≥ h = π β (T. R). L) Pareto dominates (T. one of the players will benefit from changing his or her declared value. (T. or π α (T.1 which states that an equilibrium in dominant actions is also a NE. where π 1 (100. L) = b ≥ d = π β (T. (b) There is no NE for this game. n2 ) = (100. 1994. R) but π β (T. and L is a dominant action for player β. but Rβ (B) = R. so Rβ (T ) = L. R) but π β (T. Let us observe that the parameter restrictions given in part (a) are also included in part (b) confirming Proposition 2. iii. Let us look at the following outcomes: 1 A typo in the question (first printing) leads to this undesirable result (see Basu. then T has to be a dominant action for player α. R).1 To prove this. 100). 100) = 100. 100) = 101 and π 2 (99. n2 ) = (99. L) Pareto dominates (B. L). 3. (a) If (T. and π β (T.” American Economic Review 84: 391–395. L) = b > h = π β (B. (n1 . (T. where π 1 (100. ii. A NE does not exist for this game since Rα (L) = B. (T. R) if (a ≥ c and b > d) or (a > c and b ≥ d). ii. L) is a NE. L) = a < g = π α (B. 100). . (d) Loosely speaking. L) = b ≥ d = π β (T. whereas the other player prefers (B. (b) If (T. for a mechanism that generates (2. (n1 . L) Pareto dominates (B. where π 1 (99. 99). R). L) = a ≥ e = π α (B. Formally. (c) i. n2 ). L) = a > g = π α (B. L) over (B. (a) There are three Pareto optimal outcomes: i. then π α (T. 100) = 98. R). and so on. but Rα (R) = T . either π α (T. (n1 . L) = b < h = π β (B. 99) = 98 and π 2 (100. L) is an equilibrium in dominant actions. R) over the outcome (T. R). L) if (a ≥ e and b > f ) or (a > e and b ≥ f ). 100) = π 2 (100. R). the outcomes are Pareto noncomparable if one player prefers (T. “The Traveler’s Dilemma: Paradoxes of Rationality in Game Theory. R) = c ≥ g = π α (B. 99) = 101. iii. that is π α (T. 2) as a unique NE outcome). and π β (B. R). L).

no firm finds it profitable to unilaterally change the type of cars it produces. SM. 100). This subquestion may confuse students who are already thinking in terms of backward induction. n2 ) = (99. and two proper subgames labeled JL (for Jacob left) and JR (for Jacob right). SM. an instructor teaching from the first printing is urged to change this subquestion to finding the NE for the entire game only. iii.1: Three subgames of the dynamic Battle of the Sexes (b) One easy way to find NE outcomes for an extensive form game is to construct a normal-form representation. Also. 2). ii. If (n1 . and so on.1) proves that the NE outcomes involve the two players going together either to football or to the opera. If (n1 . 100). SM. there are several NE. Hence. then player 1 can increase his or her payoff to π 1 = 101 by declaring n1 = 99. SM ) is a NE outcome since π C (LG. where the NE of the subgames are used to find the SPE. the normal-form representation is already given in Table 2. LG. φ πR = 2 πJ = 1 JL • ω πR = 0 πJ = 0 φ πR = 2 πJ = 1 φ • ω Rachel ◦ ω Jacob φ φ JR • ω πR = 1 πJ = 2 πR = 0 J • π =0 ω πR = 1 πJ = 2 πR = 0 πJ = 0 πR = 0 πJ = 0 Solution-Figure 2. players always benefit from declaring a value of one dollar lower than the other player. SM. SM ) = α ≥ γ = π C (SM. The three subgames are illustrated in Solution-Figure 2. SM. SM ) π F (LG. and the proof is identical to the one given in part (a).1. The outcome (LG. SM ) π G (LG. SM ) = β = β = π F (LG. Since our analysis is based on SPE you may want to avoid assigning this subquestion (only). n2 ) = (100. SM ) is also a NE for this game.Basic Concepts in Noncooperative Game Theory 3 i. 4. SM. to postpone finding the NE for the subgames to the next subquestion. it is easy to verify that the following three outcomes constitute NE: sJ = 2 φ if sR = ω φ if sR = φ and sR = φ. (b) Again. n2 ) = (2. then player 2 can increase his or her payoff from π 2 = 98 to π 2 = 99 by declaring n2 = 99. (a) There are three subgames: the game itself.2. SM ) = β = β = π G (LG. LG). (LG. Formally. Hence. 5. hence. then player 1 can increase his or her payoff from π 1 = 2 to π 1 = 98 by declaring n1 = 100. For example. equation (2. SM. (a) There are several NE outcomes for this game. If (n1 . .2 However. Also.

Now. . Jacob’s announcement constitutes an incredible threat. In contrast. the strategies sR = aR = φ constitute a unique SPE. sR = φ. Looking at the proper subgames in Solution-Figure 2.4 sJ = sJ = Basic Concepts in Noncooperative Game Theory ω if sR = ω ω if sR = φ ω if sR = ω φ if sR = φ and and sR = ω.1) (c) Using backward induction. himself. who knows Jacob’s best-response function in the second stage. RJ (φ)) = 2 > 1. 0) and (1.2 demonstrates that (θ. 3 Instructors: Here is a good opportunity to discuss the concepts of credible and incredible threats. Altogether. (a) Jacob’s expected payoff is given by: Eπ J (θ. 1/3) is a NE in mixed actions. we look for Rachel’s strategy (given Jacob’s best response). 1] if ρ = 1/3 RJ (ρ) =   1 if λ > 1/3   0    0  and RJ (aR ) = ω if aR = ω φ if aR = φ and RR (θ) = if θ < 2/3 [0.1. (2. 6. since Jacob. Thus. note that the outcomes (0.3 Clearly. (c) Solution-Figure 2. we first construct Jacob’s strategy (which constitutes the Nash equilibria for the two proper subgames). we conclude that RJ (aR ) = ω if aR = ω (subgame JR) φ if aR = φ (subgame JL). 1] if θ = 2/3   1 if θ > 2/3. (b) The players’ best-response functions are given by if ρ < 1/3 [0. (d) No! Jacob’s best response is to play φ whenever Rachel plays φ. ρ) = θρ × 2 + θ(1 − ρ) × 0 + (1 − θ)ρ × 0 + (1 − θ)(1 − ρ) × 1 = 2 + 3θρ − ρ − θ. 1) are also NE outcomes in mixed actions and are the same as the pure NE outcomes. Rachel.2. If Rachel plays aR = ω. would ignore it if Rachel plays φ in the first stage. ρ) = (2/3. The players’ best-response functions are drawn in Solution-Figure 2. Rachel’s expected payoff is given by: Eπ R (θ. if Rachel plays aR = φ. ρ) = θρ × 1 + θ(1 − ρ) × 0 + (1 − θ)ρ × 0 + (1 − θ)(1 − ρ) × 2 = 2 + 3θρ − 2ρ − 2θ. then her utility is given by π R (φ. then her utility is given by π R (ω. should ignore Jacob’s announcement. RJ (ω)) = 1. Also.

there are two equilibria: the two pure NE outcomes and one NE in mixed actions. ρ) = (2/3.3 has only one intersection since in that game NE in pure strategies does not exist. at the NE in mixed actions. 1/3) = Eπ R (2/3.Basic Concepts in Noncooperative Game Theory θ 1 ✻ 5 θ 1 2 3 RJ (ρ) 1 ρ ✻ RR (θ) ✻ RJ (ρ) • • RR (θ) ✲ ρ ✲ ρ 1 3 1 2 3 1 ✲ θ • 1 3 1 Solution-Figure 2. 1/3) into the players’ payoff functions (defined above) yield 2 Eπ J (2/3. 3 (e) The best-response functions in Solution-Figure 2. and therefore the best-response functions intersect only once. meaning that in the mixed extension game. .2: Best-response functions for the Battle of the Sexes in mixed actions (d) Substituting (θ. Figure 2.2 intersect three times. In contrast. 1/3) = .

.

Using the same procedure and Definition 3. However. k) = αlα−1 . ∂Q = αlα−1 k β . I urge you to make a formal definition and refer the students to Definition 3. β > 0. by Definition 3. Therefore.Chapter 3 Technology. Hence. this technology exhibit CRS if α + β = 1.2. when you first encounter a discussion of returns to scale. when addressing elasticity issues. and DRS if α + β < 1.2. (b) M PL (l. (a) AC(Q) = T C(Q) F ∂T C(Q) = c. and Demand This chapter summarizes the basic microeconomic tools the students need to know prior to taking this class. by Definition 3. and M C(Q) = Q Q ∂Q These functions are drawn in Figure 4. ∂l ∂M PL (l. (a) Let λ > 1. k)/∂k = 0. (λl)α + (λk)α = λα (lα + k α ) > λ(lα + k α ). the technology exhibits IRS if (λl)α (λk)β = λα+β lα k β > λlα k β . students should be referred to Definition 3. which holds when α + β > 1. and CRS if α = 1. ∂k which is greater than zero under the assumption that α.3. the factors are supporting since M PL (l. (b) In this technology. 4. the technology exhibits DRS if α < 1. 3. k) ≡ hence.2 in the book. . Let λ > 1. Then. k) = αβlα−1 k β−1 . the factors are neither substitutes nor complements. 2. Similarly. (a) Let λ > 1. For example.2. Answers to Exercises 1. Production Cost. ∂M PL (l. my advice is to return (or refer) to this chapter whenever definitions are needed. My advice for the instructor is not to spend time on this chapter but simply assign this chapter (with or without the exercises) as reading in the first class.2. This (quasi-linear) technology exhibits DRS since √ √ √ λl + λk < λl + λ k = λ(l + k). Then. Similarly. = + c. the technology exhibits IRS if which holds if α > 1.

p =1 MR which is independent of Q. hence if −1 < ηp ≤ 0. Hence. (a) Using Definition 3. hence if 0 ≤ < 1. 5. ηp (Q) = −1 when Q = 99/2 = 49.5. (c) The demand is elastic if ηp < −1.8 Technology. (d) M R(Q) = 0 when Q = 49. Therefore.3. Hence. Hence. > 1. T R(Q) = p(Q)Q = (99 − Q)Q. the marginal revenue is zero at the output level where the demand elasticity is −1 (unit elasticity). ηp (Q) = −2 when Q = 33.5. 1+ 1 = − −1 . The demand is inelastic if (d) By Proposition 3. (a) p = (b) ηp = ∂Q(p) p = A(− )p− ∂Q Q −1 (99 − 33)66 = 2178. which is a special case of Figure 3. Hence.3. 2 and CS(66) = (99 − 66)33 = 544. ηp (Q) ≡ p 99 − Q ∂Q(p) p = (−1) = − .5. CS(33) = 6.5 2 A .3. and Demand (b) Clearly. the elasticity is constant in the sense that it does not vary with the quantity consumed. Q Ap− Thus. (c) The inverse demand function is given by p(Q) = 99 − Q. (b) From the above. M R = p[1 + 1/(− )]. M R(Q) ≡ dT R(Q)/dQ = 99 − 2Q. ∂Q Q Q Q Hence. AC(Q) is minimized when Q = +∞. That is. AC(Q) declines with Q. Q p = A Q− . (c) Declining average cost function reflects an increasing returns to scale technology. 1 1 A(− )p− p = =− . (e) Using Figure 3. Production Cost.

given the importance of this market structure. A major reason for studying and using alternative (noncompetitive) market structures stems from the fact that the competitive market structure very often “fails” to explain why concentrated industries are observed. the term competitive refers to price taking behavior of agents. Answers to Exercises 1.Chapter 4 Perfect Competition A perfectly competitive market is characterized by nonstrategic firms. I urge the instructor to devote some time in order to make sure that the students understand what price-taking behavior means. note that this chapter does not solve for a competitive equilibrium under decreasing returns to scale technologies for two reasons: (i) other market structures analyzed in this book are also developed mainly for CRS (unit cost) technologies. we generally assume that our consumers are competitive. where firms take the market price as given and decide how much to produce (and. and (ii) most students are familiar with the DRS from their intermediate microeconomics class. You may want to emphasize and discuss the following points: 1. whether to enter. Finally. the exercise at the end of this chapter deals with a DRS technology. one could solve for a competitive equilibrium even in the presence of one firm (see an exercise at the end of this chapter).1. However. if free entry is allowed). It is now the right time to emphasize that. 3. For example. Another major reason for studying alternative market structures stems from the nonexistence of a competitive equilibrium when firms’ technologies exhibit IRS. q1 . The assumption of price taking behavior has nothing to do with the number of firms in the industry. Most students probably had some discussion of perfectly competitive markets in their intermediate microeconomics class. in economics. which means that they do not bargain over prices and take all prices and their income as given. Note that this confusion often arises since certain market structures yield market allocations similar to the competitive allocation when the number of firms increases (see for example subsection 6. 2. In a competitive market structure.2 which shows that the Cournot allocation may converge to the competitive allocation when the number of firms increases to infinity). Firm 1 takes p as given and chooses q1 to max π1 = pq1 − wL1 = p L1 − wL1 . we make a similar assumption about the firms. For example. However.

1 4. 2 d(L1 ) 4L 1 L1 = p/(2w).10 The first and second order conditions are given by 0= Hence. From the production function. Hence. q1 = q2 = 30. 2.1: Competitive equilibrium with two firms . Clearly. Qe = 120 − 60 = 60. q1 = Qe = pe /2 = 40. Given that the firms have the same technologies. which e e yields pe = 60. 3. Given that there is only one firm. e π1 = pe q1 − wLe = 80 × 40 − 1600 = 1600. Therefore. the supply equals demand equilibrium condition yields e that 120 − pe = pe /2. hence. Therefore. we can find the equilibrium employment level to be Le = 1 e (q1 )2 = 1600. q1 = √ dπ1 p = √ − w. 5. the competitive price is lower and the aggregate production is higher when the industry consists of two firms. hence. p ✻ p = 2q1 p ✻ p = 2q2 120 • p ✻ p = q1 + q2 60 • 30 ✲ q1 • ✲ q2 p = 120 − Q 120 ✲ Q = q1 + q2 30 60 6. dL1 2 L1 Perfect Competition d2 π1 p = − 3/2 < 0. the supply equals demand condition becomes 120 − pe = pe /2 + pe /2. pe = 80. Solution-Figure 4. they have the same supply functions.

Chapter 5 The Monopoly
Most students encounter the monopoly problem in their intermediate microeconomics class. However, the instructor would probably want to make sure that all students fully understand the monopoly’s choice problem and the effect of price elasticity on the monopoly’s price, as well as the arguments against monopoly (section 5.2). I also urge the instructor not to skip discussing discriminating monopoly (section 5.3). The remaining sections, the cartel (section 5.4), and the durable goods monopolies (section 5.5) are more optional depending on the instructor’s tastes and students’ ability. Answers to Exercises 1. (a) ηp ≡ dQ p = a p− dp Q
−1

p =− . ap−

Hence, the exponential demand function has a constant price elasticity. By Proposition 3.3, −1 1 =p . M R(Q) = p 1 + − (b) Equating marginal revenue to marginal cost yields1 M R = pM −1 = c = M C; hence, pM = c . −1

(c) As increases, the demand becomes more elastic, hence, the monopoly price must fall. Formally, dpM c( − 1) − c < 0. = d ( − 1)2 (d) The first edition (first printing) contains a typo. The question asks what happens to the monopoly’s price when → +1. Clearly, pM → +∞. The reason is, that when = 1, the (entire) demand has a unit elasticity, implying that revenue does not vary with price (or quantity produced). Hence, given that the revenue is constant (in fact T R = a when = 1), then the profit maximization problem is reduced to cost minimization which yields that the monopoly would “attempt” to produce as little as possible (but still a strictly positive amount).
The instructor may want to emphasize to the students that in the case of constant-elasticity demand functions (exponential demand functions) it is easier to solve for the monopoly’s price first (using Proposition 3.3) and then solve for the quantity produced by substituting the price into the demand function. This procedure becomes very handy when solving monopolistic competition equilibria analyzed in section 7.2.
1

12 (e) Inverting the demand function yields p(Q) = a1/ Q−1/ . Thus, T R(Q) ≡ p(Q)Q = a Q1− . Hence, M R(Q) ≡
1 1 dT R(Q) 1 = a Q− 1 − . dQ 1 1

The Monopoly

(f) Equating marginal revenue to marginal cost yields a Q−
1 1

1−

1

= c,

yielding that the monopoly’s profit maximizing output is QM = a −1 c .

Note that the same result is achieved by substituting pM (calculated before) into the demand function. 2. (a) Solution-Figure 5.1 illustrates an aggregate demand composed of the two groups of consumers, where each group shares a common valuation for the product. p VH

VL nH nH + nL
✲Q

Solution-Figure 5.1: Aggregate demand composed of two consumer groups (b) The monopoly has two options2 : setting a high price, p = V H , or a low price, p = V L . Solution-Figure 5.1 reveals that the profit levels (revenue since production is costless3 ) are given by π|p=V H = nH V H , and π|p=V L = (nH + nL )V L .

Comparing the two profit levels yields the monopoly’s profit maximizing price. Hence, pM =
2

VH VL

if V H > (nH + nL )V L /nH otherwise.

Instructors are urged to assign or discuss this exercise, since it provides a good opportunity to introduce the student to a discrete (logic based) analysis which is used later in a wide variety of topics (see for example the section on tying [section 14.1]). 3 The first printing of the first edition neglects to assume that production of G-Jeans is costless.

The Monopoly

13

Thus, the monopoly sets a high price if either there are many high valuation consumers (nH is large) and/or these consumers are willing to pay a very high price (V H is high). 3. (a) In market 1, p1 = 2 − q1 . Hence, by Proposition 3.2, M R1 (q1 ) = 2 − 2q1 . Equating M R1 (q1 ) = c = 1 yields q1 = 0.5. Hence, p1 = 1.5. In market 2, p2 = 4 − q2 . Hence, by Proposition 3.2, M R2 (q2 ) = 4 − 2q2 . Equating M R2 (q2 ) = c = 1 yields q2 = 1.5. Hence, p2 = 2.5. (b) π1 = (p1 − c)q1 = (0.5)2 = 0.25, and π2 = (p2 − c)q2 = (1.5)2 = 2.25. Summing up, the monopoly’s profit under price discrimination is π = 2.5. (c) There are two cases to be considered: (i) The monopoly sets a uniform price p ≥ 2 thereby selling only in market 2, or (ii) setting p < 2, thereby selling a strictly positive amount in both markets. Let us consider these two cases: i. If p ≥ 2, then q1 = 0. Therefore, in this case the monopoly will set q2 maximize its profit in market 2 only. By subquestion 3a above, π = π2 = 2.25. ii. Here, if p < 2, q1 > 0 and q2 > 0. Therefore, aggregate demand is given by Q(p) = q1 +q2 = 2−p+4−p = 6−2p, or p(Q) = 3−0.5Q. By Proposition 3.2, M R(Q) = 3 − Q. Equating M R(Q) = c = 1 yields Q = 2, hence, p = 2. Hence, in this case π = (p − c)2 = 2 < 2.25. Altogether, the monopoly will set a uniform price of p = 2.5 and will sell Q = 1.5 units in market 2 only.4 4. (a) π(q1 , q2 ) = (100 − q1 /2)q1 + (100 − q2 )q2 − (q1 + q2 )2 . (b) The two first order conditions are given by 0= ∂π ∂q1 ∂π 0= ∂q2 = 100 − q1 − 2(q1 + q2 ) = 100 − 2q2 − 2(q1 + q2 ).

M M M M Solving for q1 and q2 yields that q1 = 25 and q2 = 12.5. (c) Substituting the profit maximizing sales into the market demand functions yield pM = pM = 87.5. Hence, 1 2 M M π(q1 , q2 ) = 87.5 × 12.5 + 87.5 × 25 − (12.5 + 25)2 = 1875.

(d) Now, that each plant sells only in one market, the two first order conditions become 0= ∂π ∂q1 ∂π 0= ∂q2 = 100 − q1 − 2q1 = 100 − 2q2 − 2q2 .

Note that consumers in market 1 are better off under price discrimination than without it, since under no discrimination no output is purchased in market 1. Given that the price in market 2 is the same under price discrimination and without it, we can conclude that in this example, price discrimination is Pareto superior to nonprice discrimination, since both consumer surplus and the monopoly profit are higher under price discrimination.

4

(ii) charging p2 = 50. and selling only to the second period consumers. the second period profit maximizing price is p2 = 50. The two cases are illustrated in Solution-Figure 5. (a) Equating marginal revenue to the tax inclusive unit cost yields a − 2Q = c + t. thereby earning a second-period profit of π2 = 2 × 50 = 100. pM = (a + c + t)/2. (f) This decomposition increases the monopoly’s profit since the technology exhibits DRS. Hence. yielding a profit level of π2 = 2 × 50 = 100. First-period consumer buys in period 1: In this case the second period profit maximizing price is again p2 = 50. or. the second period price is independent of the actions of the first-period buyer. as illustrated in Solution-Figure 5.14 M M yielding q1 = 100/3 and q2 = 25. 1 2 π = π1 (q1 ) + π2 (q2 ) = 250 100 × − 3 3 100 3 2 + 75 × 25 − 252 = 2917. Now.3. (a) We solve for the monopoly’s profit maximizing prices starting from the second period. By section 5. Therefore. the monopoly has two choices: (i) charging p2 = 20. Altogether. yielding a profit level of π2 = 3 × 50 = 150. the maximum price the monopoly can charge the first-period buyer in the first period is p1 = 150. an increase in t raises the monopoly price by less than t. Altogether. thereby earning a second period profit of π2 = 3 × 20 = 60. in this case. thereby extracting all surplus from all consumers. and sell to all three consumers. by Proposition 3. yielding a profit level of π2 = 2 × 50 = 100. The Monopoly (e) pM = 100 − 100/6 = 250/3 and pM = 100 − 25 = 75. (b) The second period outcome may depend on two cases: First-period consumer does not buy in period 1: In this case. the second-period price is independent of the action of the first-period buyer in the first period. 1 2 1 −2 = pM 1 + 2 1 −4 = M R2 (q2 ). or QM = (a − c − t)/2. checking the effect of a tax rate change on the monopoly’s price yields that dpM /dt = 1/2 < 1. Now. . 7.2 (left). The second period outcome may depend on two cases: First-period consumer does not buy in period 1: Clearly. the monopoly should set p1 = 40. in order to attract the first-period buyer to purchase in period 1. Hence. M 5.5pM . Hence. 6. the discriminating monopoly will set quantities to satisfy M R1 (q1 ) = M ). pM = 1. Hence. First-period consumer buys in period 1: In this case the second period profit maximizing price is again p2 = 50.2.3 M R2 (q2 M R1 (q1 ) = pM 1 + 1 Thus.

as illustrated in Solution-Figure 5.The Monopoly p ✻ 15 p pM (c + t) ✻ • • • a pM (c + t) pM (c) c+t pM (c) • c MR ✲Q c+t c D ✲ Q M R(Q) D Solution-Figure 5. Now. Hence. M R = pM 1 + 1 −2 = c + t.2: How the monopoly price varies with a specific tax (b) Using Proposition 3.3. an increase in t raises the monopoly price by more than t.2 (right). . checking the effect of a tax rate change on the monopoly’s price yields that dpM /dt = 2 > 1. yielding pM = 2(c + t).

.

subsection 6. and therefore provides self-discipline which restricts the gains from deviation from the cooperative actions.1. since c2 > c1 . Market structures are assumed by the researcher rather than solved for. this topic is written in a way that enables the instructor to teach self-enforcing collusion even without formally teaching repeated games. if pe < c1 then both firms produce zero output. i = j. where under the Cournot game firms’ actions are the quantity produced. Hence. (a) By Lemma 4. However. j = 1. Cournot and Bertrand market structures select Nash equilibrium outcomes. The important issues to discuss and emphasize are: 1. Answers to Exercises 1. qj . ∂q1 and 0 = ∂π2 = α − q1 − 2q2 − c2 . (b) Each firm i takes the output of its opponent. 2. by showing that a partial removal of trade barriers need not be welfare improving. i. but at these prices demand exceeds zero (since Q(c1 ) = α − c1 > 0). 2. including the player’s own earlier actions.2 (preferential trade agreements) provides a good example to the law of (no) second best.6. and in the Bertrand game firms’ actions are prices. 3. if a competitive equilibrium exists then it must be that pe ≤ c1 . Specifying a market structure is the same as specifying the rules of the game before predictions (e. if a competitive e e equilibrium exists. Hence.g. Self-enforcing collusion (section 6. Finally. q1 = Qe = α − c1 and q2 = 0. qi yielding first order conditions given by 0= ∂π1 = α − 2q1 − q2 − c1 . Instructors who choose to teach this topic may want to emphasize that a trigger strategy is a function of the entire history of what each player has played in earlier periods.5) is suited for more advanced students. as given and solves max πi = (α − qi − qj )qi − ci qi .Chapter 6 Markets for Homogeneous Products The Cournot. ∂q2 . then pe = c1 .3) before covering self-enforcing collusion. Instructors may want to go over repeated games (section 2. Free entry versus a fixed number of firms. both firms produce a finite amount of output only if p ≤ c1 and p ≤ c2 . however. equilibrium outcomes) are sought.. and sequential moves market structures are essential for a further study of industrial organization. Therefore. Bertrand.

(e) In a Bertrand equilibrium. Hence. Thus. . thereby ensuring that firm 2 will not find it profitable to produce any amount. the more efficient firm (firm 1) completely undercuts the price set by firm 2. (ii) price is higher when firm 2 moves first. s q1 = α − c2 ∂π1 + q1 − c1 . the Bertrand equilibrium price exceeds the competitive equilibrium price. Comparing the two sequential-moves equilibria we see that (i) aggregate output decreases when the less efficient firm (firm 2) moves before firm 1. 3 3 3 c c Then. firm 1 sets its price to equal the unit cost of firm 2. q1 = Qb = α − c2 + and q2 = 0. pc = α − q1 − q2 = (α + c1 + c2 )/3. 2 4 4 Hence. q1 and q2 yields c q1 = 2α − c1 − c2 α − 2c1 + c2 α − 2c2 + c1 . Hence.18 Markets for Homogeneous Products Solving the two equations for the two variables. we can approximate the Bertrand equilibrium by b b pb = c2 − and pb = c2 . q2 = . 2 4 4 Hence. (iii) the market share of the less efficient firm increases when it moves before firm 1. Assuming that money is continuous. ps = (α + c2 + 2c1 )/4. = α − 2q1 − 2 ∂q1 α + c2 − 2c1 s α − 3c2 + 2c1 3α − c2 − 2c1 . instructors should point out to the students that this is not really a Nash-Bertrand equilibrium since the profit of firm 1 increases as decreases. (d) When firm 2 is the leader. and Q = . R2 (q1 ) = α − c2 q1 − . yielding the first order condition 0= Therefore. and (v) the market share of the more-efficient firm increases when it moves first. Notice that 1 2 when the unit costs are not equal (the present case). q2 = . and Qs = . firm 1 (the leader) takes firm 2’s best-response function as given and chooses q1 to max π1 = α − q1 − q1 α − c2 q1 − 2 2 q1 − c1 q1 . Also. (iv) the production of the more-efficient firm increases when it moves first. 2 2 In a sequential-moves equilibrium. ps = (α + c1 + 2c2 )/4. by symmetry (replacing c1 by c2 and vice versa in the previous subquestion) we can conclude that s q2 = α + c1 − 2c2 s α − 3c1 + 2c2 3α − c1 − 2c2 . and Qs = . (c) From the first order conditions of the previous subquestion we can immediately solve for firm 2’s best-best response function. q1 = .

0 = πi = − F. the best-response function of firm 1 is given by q1 = R1 (q2 . q2 )]q2 − cq2 . . q2 The first-order condition is given by 0 = dπ2 /dq2 = 60 − q1 /2 − q2 − c/2. Given that all firms have identical technologies (identical cost functions). Hence. Therefore. Thus. 2 2 In the second period. the (endogenously determined) number of firms in this industry declines with the fixed cost parameter F . q2 ) which is given in (6. pc = 100 − 100N 20.8) we have it that R3 (q1 . 4 N +3 N +3 (b) Under free entry. we can attempt to search for a Cournot equilibrium where all firms produce the same output c c c levels. q3 . hence. . the ¯ best-response function of firm 2 is given by R2 (q1 ) = 120 − q1 − c . or N = 2 (N + 3) F Thus. and πi = pc q c − F − (q c )2 = N +3 N +3 (N + 3)2 100 100N (N − 1)q c .Markets for Homogeneous Products 19 2. under free entry all existing firms make zero profit. . 20. firms enter as long as they make above normal profits. Hence. 2 . 000 300 − F. . 000 − 3. or q = . firm 1 takes q2 . (a) We first focus on the problem faced by firm 1. 3. 000 20. . ∂q1 100 − 4 N j=2 qj Therefore. . qN ) = . firm 3 takes q1 and q2 as given and chooses q3 to maximize its profit. That is. Q = . In a Cournot market structure. by a real number. qN as given and chooses q1 that solves max π1 = (100 − q1 − q2 − · · · − qN )q1 − F − (q1 )2 . q2 ) as given and chooses q2 that solves ¯ q ¯ q max π2 = [120 − q1 − q2 − R3 (¯1 . q1 The first order condition is given by 0= ∂π1 = 100 − 2q1 − q2 − · · · − qN − 2q1 . N . using (6. In this case. . Entry stops when a further entry generates a loss to firms. if we approximate the number of firms. = .8). In the third period. firm 2 takes q1 and R3 (¯1 . The solution to this problem yields firm 3’s best-response function R3 (q1 . q2 ) = 120 − c q1 + q2 − . q c ≡ q1 = q2 = · · · = qN . q c = 25 − Therefore. .

or ρ > 1/2. and q3 = 4 8 2 s s s Hence. . 4 4 The first-order condition is given by 0 = dπ1 /dq1 = 30 − c/4 − q1 /2. Now. if no firm deviates. let firm i’s price strategy. firm 1 takes the best-response functions of firms 2 and 3 as given and chooses q1 that solves max π1 = {120 − q1 − R2 (q1 ) − R3 [q1 . deviation is not profitable to either firm if 5N/(1 − ρ) > 10N . . 1 2 (d) We can reconstruct firms’ trigger strategies where we assume that the punishment point is now pb = 4 and pb = 4. the sum of discounted profits is π1 = 5N/(1 − ρ). be given by 2 pi (τ ) = 10 as long as p1 (t) = p2 (t) = 10 for all t = 1.20 Markets for Homogeneous Products In the first period. Qs = q1 + q2 + q3 = 7(120 − c)/8 and ps = (120 + 7c)/8. B C and the government earns a revenue of $10 per unit. solving s s s s for q1 . Hence. (a) Let pA denote the (tariff inclusive) price consumers in country A pay for a unit of k import from country k. (a) Under zero production cost. pA = 70 and pA = 50. 1 2 (b) We’ll construct an equilibrium in trigger strategies where firms’ threat point is pb = 0 1 and pb = 0. respectively. Firm 2: If no firm deviates. then π1 = (10 − )N + 0 ≈ 10N (since effective period τ + 1 all firms revert to pi = 0). k = B. (e) Collusion is less likely to be sustained when one firm has a cost advantage over the other because deviation by firm 1 (low cost) will leave it with a positive stream of profit compared with a zero profit generated by deviation when the two firms have an identical cost structure. if no firm deviates in any period. 4. τ − 1 0 otherwise. all Bertrand equilibrium outcomes yield zero profits to both firms. the sum of discounted profit is π2 = (N/2)(10−4)/(1− ρ). If firm 1 deviates in period τ . If firm 1 deviates in period τ . Comparing the two profit levels reveals that deviation is not profitable if ρ > 1/2. A’s consumers still purchase from country C B B C implying that this FTA is ineffective and therefore does not alter A’s welfare. 5. under a uniform specific tariff of $10 per unit of import. R2 (q1 )]}q1 − cq1 = {30 + q1 3c q1 − }q1 − cq1 . and then into R3 [q1 . Hence. and then substituting into R2 (q1 ). (c) All Bertrand equilibria are of the form pb = 4 − ≈ 4 and pb = 4. Since now firms have different profit functions. Thus the Bertrand outcome is pb = 0 and pb = 0. 2. Firm 1: In each period τ . we 1 2 need to check profitability from deviation by each firm separately. If it deviates. A FTA with country B reduces pA to pA = 60 > 50 = pA . the present value of the sum of discounted profit to each firm i is πi = 5N/(1 − ρ). Formally. s q1 = 120 − c s 120 − c 120 − c s . . Therefore. R2 (q1 )] yields. i = 1. . Clearly consumers buy from country C. Therefore. . C. then its sum of discounted profit is π1 = 10N + ρ4N/(1 − ρ). then the sum of discounted profit is π2 = (10 − 4)N + 0. q2 = . Comparing the two profit levels shows that deviation is not profitable for firm 1 if ρ > 5/6.

it must yield that both firms charge pL . I provide a formal solution for this problem. with the exception of highly elastic demand. (a) Clearly pL is a dominant action for each firm in the static one-shot game. (I. 6. Therefore. Altogether.01 to pA = 60 (a reduction C B of 1 cent per unit of import). by Proposition 2. the FTA hardly changes consumer surplus. p2 = pL if p1 = pL (b) We now add a third stage to the game.1: Sequential price game: Meet the competition clause From the figure. it is straightforward to conclude that the following prices constitute a SPE: pL if p1 = pH and p1 = pL .1. Hence. Note that in this case the consumer price falls from pA = 60. Note that the answer for exercise 3 provides a solution for N = 3 firms. Hence. implying that the firms charge the industry’s profit maximizing price and earn a profit of 100 each.:) pL π1 = 70 π2 = 70 (III. However.Markets for Homogeneous Products 21 (b) Now.:) (II.2. Hence.:) pL • Firm 1 ◦ Firm 2 pH pL pH • pH π1 = 100 π2 = 100 π1 = 120 π2 = 0 pL π1 = 0 π2 = 120 • Firm 1 π1 = 70 π2 = 70 Solution-Figure 6. In this case. in this case the government faces a large reduction in tariff revenues due to the elimination of a $10 tariff per unit of import. country A loses from the FTA. . in what follows.01 = pA . the SPE if given by p2 = pH pL if p1 = pH if p1 = pL and p1 = pH . In addition. where firm 1 commits to reduce its price to firm 2’s price whenever firm 2 charges pL . We can derive the SPE directly by formulating the extensive form game which is illustrated in Solution-Figure 6. if a SPE exists. 7. (pL . which intuitively can be generalized to the N > 3 case. This third stage is (partly) illustrated at the bottom of Figure 6. a FTA with country B reduces pA to pA = 60 < 60. A’s B B C consumers switch to buying from country B and no tariff revenues are collected. pL ) is a unique NE for the one-shot game.1 (consider the first two stages only).

let us define aN ≡ a − Qs . recall that s + qs qN N −1 = (1/2 + 1/4)(aN −1 − c). 4 (aN −2 − qN −2 ) − s s s The solution is qN −2 = (aN −2 − c)/2. . 2 for all N. Thus. s In particular. . . qN −1 = (aN −1 − c)/2. Define Qs ≡ N qi and Qs = N −1 qi . 2 we have it that 1 N −1 s 1 a−c a−c s qN = q1 = N −1 = N . Firm N solves max πN = (a − Qs − qN − c)qN . there is no limit on the number of firms that will enter if free entry is allowed. . since 1 k s s qN = qN −k . s qN = (aN − c)/2 = (aN −1 − c)/4. s s Also. Therefore. s We now analyze firm N − 2. the N -firm sequential moves equilibrium output levels are s qi = a−c . . Hence. Firm N − 1 knows what firm N will be producing and solves max πN −1 = qN −1 s (a − Qs −1) − qN −1 ) − −(N = [aN − c]qN −1 2 aN − c − c qN −1 2 [aN −1 − qN −1 − c]qN −1 = . We therefore look at how the N th firm decides on how much to produce. it −N can be verified that s aN −1 = aN + qN −1 for allN. 2 2 2 2 Altogether. . . −1 Since there is no fixed cost. 2i for all i = 1. N . qN −1 = (aN −2 − c)/4 and qN = (aN −2 − c)/8. firm 2 solves max πN −2 = qN −2 = 3(aN −2 − qN −2 − c) − c qN −2 4 [aN −2 − qN −2 − c]qN −2 . for N = 1. s qN = aN − c . N. . a1 = a − Qs = a − 0 = a. −N Before we proceed to analyzing firm N − 1. . 2 s Hence. . for i = 1. −1 Finally. Recall that aN −2 − qN −2 = aN −1 . Therefore.22 Markets for Homogeneous Products s (a) Let qi denote the sequential-moves equilibrium output level produced by firm i. Hence. since aN = aN −1 − qN −1 and aN = aN −1 − qN −1 = (aN −1 + c)/2. . for all 0 < k < N. −N qN s yielding qN = (a − Qs − c)/2. . q1 = (a − c)/2. Also. . s s i = 1. N where i=1 −N i=1 Qs ≡ 0 for i = 1.

ps → c (the competitive price level). Qs → a − c (competitive output level). . Hence. we can show that1 Qs = (a − c) 1 − 1 2N .9). section 9.Markets for Homogeneous Products (b) Aggregate output is given by Qs = i=1 s N qi = (a − c) 23 1 1 1 + + ··· + N 2 4 2 . (c) As the number of firms increases. 1 The first printing of the first edition contains a typo in the specification of Qs in the question.2 (see mathematical appendix. Using the same arguments as in the proof of Lemma 9.

.

1 and are derived as follows: 3 max π1 ⇒ R1 (α2 ) = 2 + α2 . for example. are important for a further study of industrial organization. Section 7. Answers to Exercises 1. Location models (section 7.3). (a) The best-response functions are drawn in Figure 7. α1 2 1 max π2 ⇒ R2 (α1 ) = 1 + α1 .3. Instructors should cover at least the Hotelling linear city model which students always find very intuitive and therefore very appealing. in particular the Hotelling model (subsection 7. section 7.1).Chapter 7 Markets for Differentiated Products The first section extends the market structures defined in Chapter 6 to markets with differentiated products. The instructor can demonstrate the parameter restrictions needed to ensure existence by imposing symmetric locations (a = b) into the conditions of Proposition 7.1: Advertising best-response functions . to distinguish between vertically and horizontally differentiated products (see section 12. The linear city model is used several times in the manuscript.2). α2 2 α2 ✻ 4 2 or. and its application to the theory of international trade.5) can be skipped. and to define and contrast the concepts of strategically substitutes and strategically complements best-response functions.2 (somewhat more advanced) introduces the student to the endogenous determination of the variety of brands. The existence proof (Appendix. 3 3 R1 (α2 ) 8 1 0 −4 3 14 ✲ α1 • R2 (α1 ) Solution-Figure 7.6. α2 = − + R1 (α2 ). Using the two-brand environment enables us to analyze the difference in market outcomes generated by the Cournot and the Bertrand market structures.

one on each side of the firm.26 Markets for Differentiated Products (b) The two best-response functions are upward sloping. denoted by x. (a) Given a low reservation price (a low B). the monopoly will pick the highest possible price subject to having the consumers living at the edges of town purchase the product. ˆ Hence. that is. The indifferent consumer (on ✛ 2a 1 2 ✲ 1 2 0 1 2 −a +a 1 Solution-Figure 7. Figure 7. 2 a= B . the monopoly chooses p that solves max π = p2a = 2p(B − p).2: Single-firm location model each side) is determined by the reservation utility. 1 pM = B − .2 illustrates the location of the two indifferent consumers. 2 1 a= . set p to satisfy B − 1/2 − p = 0. hence strategically complements. B − a − p = 0. we get that the indifferent consumers lie “outside” the city. (a) The indifferent consumer. hence. This implies that for B > 1. π1 = 14 2 2. a = B − p. we’ll attempt to find a monopoly equilibrium where not all the market is served. or. Hence. 2 3. 2 and π M = 2pa = B2 . x= ˆ R + p2 − p 1 . 2 Now. not all the market is served. 1+R R = 1. Formally. 1+R ˆ (b) Substituting p1 = p2 yields x = R/(1 + R). the monopoly can increase the price and still having the entire street purchase the product. Thus. N N (c) Solving for a NE yields α1 = 14 and α2 = 8. Hence. (b) When substituting B > 1 into the solution for ‘a’ found in the previous subquestion. The firms’ profit levels in a NE are: 2 and π = 82 . R→∞ lim x = lim ˆ R→∞ . 2 1 and π M = B − . it is easy to verify that 0 < B < 1 implies that 0 < a < 1/2. p The solution is given by pM = B . Hence. must satisfy ˆ ˆ x × 1 + p1 = (1 − x) × R + p2 .

x = 1) only if the transˆ portation cost for traveling to the east is infinite. Substituting a = b and x = L/2 ˆ into firm A’s first-order condition (or B’s first-order condition) yields that pA = pB = τ L(L − 2a). we have it that pA = 2τ (L − a − b)ˆ and x pB = 2τ (L − a − b)(L − x). we search for a symmetric solution. reflecting the fact that prices drop to zero when the two brands become homogeneous. From (7. Therefore. all the consumers will eat in restaurant 1 (i. 2τ (L − a − b) Similarly. using A’s first-order condition. firm B increases its profit by moving toward point L. Substituting into the function F (. we have it that −(2ˆ − L) − 2(ˆ − a) x x ∂x ˆ =− = ∂a 2(L − a − b) + 2(ˆ − a) − 2(ˆ − L + b) x x x=1/2 a=b 1 .24). Similarly. yielding a first-order ˆ condition given by 0= ∂x ˆ dπB = L − x − pB ˆ = x + pB ˆ dpB ∂pB ∂F ∂pB ∂F ∂x ˆ =L−x− ˆ pB . x x x Using the implicit function theorem. x dπA <0 da whenever 1 − (ˆ)2 + (L − a − b)2ˆ < 0. we can define the location of the indifferent consumer.1 Using the two first-order conditions. 1 This problem may be too difficult for most undergraduate students since it involves deviating from the symmetric solution.Markets for Differentiated Products 27 Therefore. 4. 4 Now. ) = 2τ (L − a − b)(2ˆ − L) + τ (ˆ − a)2 − τ (ˆ − L + b)2 = 0. we can express A’s profit by πA = pA x = ˆ 2τ (L − a − b)(ˆ)2 . Otherwise. Hence. firm B chooses pB that solves maxpB πB = pB (L − x). The first-order condition is given by 0= dπA ∂x ˆ ∂pA = x + pA − ∂F ˆ = x + pA ˆ dpA ∂pA ∂x ˆ ∂F =x− ˆ pA . 2τ (L − a − b) (a) Under a = b. . x x ˆ ˆ Firm A chooses pA that solves maxpA πA = pA x subject to x satisfying F (pA . firm A increases its profit by moving toward point 0. Note that pA = pB → 0 when a → 1/2. some consumers will always eat at restaurant 2.e. x.. ) = 0 yields ˆ F (. x x 4 which always hold since L − 2a < L. (b) We now investigate the total effect of varying the location of firm A (varying a) on the profit of firm A. pB ) ≡ pA − pB + τ (ˆ − a)2 − τ (ˆ − L + b)2 = 0. pB ) = 0. by the implicit ˆ function: F (pA .

.

1 Finally. simply because a monopoly can always mimic the same production pattern of any oligopoly. ˆ ˆ note that the equilibrium first-period capital level in Figure 8.1). however.10) yields π1 = π4 = 1111. (a) I4 = 20 + 20 + 20 + 10 = 70.Chapter 8 Concentration.10) yields a monopoly profit of π M = 2500. However. IHH = 202 + 102 + 102 + 3 × 202 = 1. (a) The solution for a Cournot equilibrium with N firms is given in (6.. The merger may be challenged by the FTC or the Justice Department since the postmerger IHH exceeds 1. Regarding all other topics. with the analysis of vertical integration given in subsection 8.7 is k3 and not k2 as wrongly stated in the first printing. πi = 2500/3 = 833 1 > 625 > 555. 600. firms 2 and 3 lose from this merger whereas firm 1 gains. i = 1. the instructor will find a wide variety of models to choose from. After the merger of firms 1 and 2.111 . iii. ∆IHH = 1. 000 and ∆IHH = 200 > 100. 2. note that you can combine the topic of double monopoly markup. Thus. . (b) Substituting N = 2 into (6. analyzed under the topic of dealership (subsection 14. 600 = 200. Most of the topics are logical and require minimum calculations and very little calculus (if any). (d) A merger into a monopoly cannot reduce aggregate profit. π2 = π3 = π4 /2 = 555 < 625. Also. 800 − 1.9) and (6. . (c) i. In the second printing I intend to add an appendix at the end of this chapter which derives both best-response functions. Mergers.2. 3. To see this.6). entry barriers. and Entry Barriers This chapter contains four major topics: How to measure concentration. substituting N = 1 into (6.4. the profit per firm is the highest 3 under the monopoly market structure (cartel). (c) Since firms 2 and 3 split the profit from the merged firm. 2. 800.2 can be presented using diagrams only. and entry deterrence. the first printing of the first edition lacks a ˆ justification why firm 1’s (high and low) best-response functions do not shift with a change in k. (e) The difference between the two mergers is as follows: When two out three firms merge.3.1) and the appendix on merger regulations and guidelines (section 8. (b) IHH = 4 × 102 + 3 × 202 = 1.10). the extra aggregate industry profit generated from the change in the number of firms is equally divided between Subsection 8. competition is reduced to a duopoly. Hence. Answers to Exercises 1. I urge the instructor not to skip the discussion of concentration measures (section 8.2. Hence. merger and merger guidelines. 1 . ii. c c Substituting c = 0 (zero production cost) and N = 3 yields qi = 25 and πi = 625.

4 The Y -seller takes pX as given and solves max πY = pY 2Q = 2pY (α − pX − 2pY ). merging into a monopoly can only increase profit since a monopoly can always produce like a duopoly. q I ) = (20. Y S 3 6 3 3 9 (b) We define a system as one unit of X bundled with two units of Y . 2 A typo in the first printing of the first edition: the question should state that Q = x = y/2. πX = πY = . before the S 1 merger. 3. and Entry Barriers the two unmerged groups of firms. In view of Definition 8. (a) We look for a Nash equilibrium in prices. and πS = . the merger is welfare improving. Thus. It is also sustainable since no potential entrant can undercut the incumbent’s price while making a positive profit. Before the merger. pY yielding pY = Hence. Mergers. . Also.1. the merged firm chooses a system price pS that solves max πS = pS (α − pS ) yielding pS = QS = pS α α2 . pN = X α α 2α α α2 N N . this share of the change in aggregate industry profit is lower than the profit each firm makes under the triopoly. Then. and under monopoly. 2 α − pX . Formally. q I ) = (20. In contrast. all industry profit is equally divided between the three firms. 4. pN = . QN = .1 shows that the pair (pI . pN = . Figure 8.30 Concentration.2 The X-seller takes pY as given and solves max πX = pX Q = pX (α − pX − 2pY ). the price of one system is p0 = pX + S 2pY = 2α/3 > α/2 = p1 which is the price after the merger. 10) constitutes a contestable-market equilibrium. 10) is feasible since demand equals supply and the incumbent makes a nonnegative profit. aggregate profit is πX + πY = 2α2 /9 < α2 /4 = πS which is the profit after the merger takes place. the (incumbent’s) industry configuration (pI . Since the system price falls and industry profit increases. pX yielding pX = α − pY . 2 4 (c) To make a welfare judgment on the gains from this merger it is sufficient to compare prices and profit levels.

1: Contestable-markets equilibrium . and Entry Barriers 31 pI ✻ p = 60 − 4Q ✒ AC(q I ) = 20 • 100 q +q 10 ✲ qI Solution-Figure 8.Concentration. Mergers.

.

plus twice the probability that it discovers while one competing firm discovers and one does not. However Figure 9. 11 3V − 2I. Eπ|1 lab = V /2 − I.4) does not reach a closed-form solution. (b) The now single firm can operate zero. Hence. Therefore.1) uses only demand and supply diagrams. the monopoly solves M R(QM ) = a − 2QM = c1 = 2c0 − a. 2.3 enables the instructor to explain the cost and benefit of the patent system without performing any calculations. Finally. innovation is major if the monopoly price is below the initial unit cost. Operating two labs: In this case. the appendix sections discuss patent law and R&D joint ventures (sections 9.8) and can be assigned as homework reading. 3. all the three firms find it profitable to engage in R&D if V > 24I/7 = 24/7 (since I = 1). i = 1.5) uses supply and demand analysis. Answers to Exercises 1. The international analysis (section 9.2) relies on simple discrete probability calculations. To find pM . Cooperation in R&D (section 9. given that the other two firms are also engaged in R&D. Eπi = 111 111V 111V 7V V + (×2) + −I = − I. for each firm i.Chapter 9 Research and Development Most of the topics in this chapter are not technically demanding (and less tedious in terms of technical derivations). Thus. the instructor can simply reproduce the government-innovation two stage game and demonstrate how to build the social welfare function for calculating the optimal patent life. 2. Innovation race (section 9. the monopoly’s profit maximizing output is QM = a − c0 .3) relies on a Cournot market structure in the second stage and a Nash equilibrium in R&D in the first stage. The calculation of the optimal patent life (section 9. pM = c0 .1 the present innovation is neither major nor minor. Licensing of innovation (section 9. minus innovation cost. Then. plus the probability that all the three discover simultaneously. the probability of discovery is one minus the probability that none of the labs discovers. Therefore. that is if pM < c0 .1. or two labs. The classifications of process innovation (section 9. V − 2I = Eπ|2 labs = 1 − 22 4 .7 and 9. is the prize times the sum of the probability that the firm discovers while the other two do not. Operating a single lab: In this case. (a) The expected profit of a firm engaged in R&D. 222 222 2 222 3 24 Thus. Using Definition 9. Formally. Innovation is minor if pM > c0 . one. according to Definition 9.6) is divided into a simple matrix game of product innovation and a two stage subsidy Cournot game of process innovation.

since most poor and middle-class people do not get to fly in the Concord. the probability that none of the n firms discovers at a particular date is (1 − α)n . Arguments against the subsidies include (i) duplication of R&D which require twice the amount of resources needed to develop and test the plane. then Airbus’ dominant action is to develop. Hence. 1 − (1 − α)n . (ii) the government may not possess the means to determine whether the benefits from the development to their own economy dominate the cost.5). then Boeing’s dominant action is to develop. Arguments for the subsidy include the introduction of more competition between aircraft manufacturers that would reduce aircraft prices. (c) None! After the EC subsidy to Airbus is implemented. n ]2 [1 − (1 − α) α(2 − α) 4.34 Research and Development Now. (i. economists generally believe that the private sector can collect more accurate information). ET (n) = [1 − (1 − α)n ] ∞ t=1 [(1 − α)n ]t−1 × t (Lem 9. (c) Similar to the derivation given in (9. which constitutes a sufficient condition for having the firm choosing to operate two labs. subsidy to Boeing).1) = 1 − (1 − α)n 1 = .S. . one minus the probability that none discovers. (b) Clearly. that is. and may therefore increase the amount of travel. 3.S. independent of the U. some economists claim that the development of the Concord by some EC countries constitute a net transfer from poor to rich people.S. ET (n) = [1 − (1 − α)n ] × 1 + (1 − α)n [1 − (1 − α)n ] × 2 + (1 − α)2 n [1 − (1 − α)n ] × 3 + (1 − α)3 n [1 − (1 − α)n ] × 4 + · · · Therefore. Eπ|2 labs > Eπ|1 lab if V > 4I. government provides Boeing with a subsidy of 10 (contingent on developing the megacarrier. Note that this U. and (iii) income distribution effects. there are pros and cons to having both governments subsidizing the development of the new megacarrier. subsidy is independent whether the EC provides Airbus with a subsidy of 10 or 15..e. (d) Like any other welfare argument in economics. Airbus’ dominant action is to develop (hence. Hence. (b) If the U. (a) The probability that a firm does not discover is (1 − α). (a) If the EC provides Airbus with a subsidy of 10 (contingent on developing the megacarrier. boost the airline industry.

I feel that this term was poorly chosen since externalities are either assumed or not assumed at all. pB has to be sufficiently high to satisfy ˆ 1=δ= (b) Y − pA . whereas the components approach model is very simple. some students may get confused because of the existence of multiple equilibria under incompatibility. In particular.1) uses very simple calculus (basically one first-order condition). the instructor may want to construct only one equilibrium and compare it to the equilibrium under compatible systems. I intend to add two exercises covering the two topics analyzed in the network externalities section. In future printings/editions. (a) Firm A gains control over the entire market when δ = 1. where a consumer’s utility is affected by the number of consumers purchasing the same brand. Then. • The supporting services and the components approaches do not assume any externality. Let δ denote the market share for firm A after income has doubled from Y to 2Y . The remainder of section 10. the two exercises cover only the supporting services and the components approaches.” where the utility of a consumer is affected by the brand choices of other consumers is an equilibrium result (rather than an assumption).1 Thus. The behavior of a monopoly telephone company under network externalities (Subsection 10.1 is devoted to a simple discrete (noncalculus) standardization-variety tradeoff model. the network effects are generated by having hardware and software (or other types of hardware) treated as perfect complements.Chapter 10 The Economics of Compatibility and Standards The introduction to this chapter provides some basic terminology and a simple “Battle of the Sexes” game of product standardization.10). The important points to emphasize are: • A network externality is a particular type of a consumption externality. Thus. network externalities constitute an assumption about consumer preferences. 1 . ¯ i. The remaining two sections. by (10. Answers to Exercises Unfortunately. The “network effects. the supporting services approach (section 10. and therefore the term indirect cannot be used to describe models that do not explicitly assume any externality.2) and the components approach (section 10.1. ˆ 1.10) we have that ˆ ¯ 2Y − pA − pB > δ = Y − pA − pB δ= 4Y − pA − pB 2Y − pA − pB Some authors use the term indirect network externalities to describe the behavior generated by the supporting services and the components approaches. in which case. In this case.3) are somewhat more difficult despite the fact that no calculus is used. by (10. which can be taught to the less technically proficient students. 2Y − pA − pB or pB = Y.

and BA. Under this type.9) NA /NB must increase. In what follows we will demonstrate.36 The Economics of Compatibility and Standards ˆ if pA > pB . since consumers’ reservation utility is zero. and BA. by (10. Under this type. pI = 0 and B A I πA = 0. Firm A sells to consumers AA. However. see equation (10. ii. px = py = px = py = λ. and BA). However. Thus. In this equilibrium. yielding ˆ dδ pA − pB >0 = dY (2Y − pA − pB )2 if pA > pB . and AB. AB. and firm B sells to BB and BA (alternatively. (a) Potential equilibria can be classified into two types: (i) Firm A sells to consumers AA. However. i. AB. Since δ increases. firm A can always increase its profit by raising its price by thereby making strictly positive profit. it can be established that there exists an equilibrium in which each consumer buys his or her ideal system. this prices cannot support a Bertrand-Nash equilibrium since firm A can increase its profit by undercutting firm B by charging pA = pI − which will cause consumer BA B to switch to system AA (this is the usual Bertrand undercutting argument). whereas firm B sells to BB only (alternatively. (b) Following the proof of Proposition 10. pI = pI since consumers AB and BA are indifferent between A B systems AA and BB as long as the prices are equal. whereas firm B sells to BB. ˆ ii. hence. that both types of equilibria do not exists. by a way of contradiction. 2. equilibrium prices must satisfy pI ≤ 2λ. This result can also be demonstrated by differentiating δ with respect to Y . the firm with the lower market share increases its market share when consumer income rises. .14. (ii) Firm A sells to AA.12). and firm B sells to BB and AB). firm A sells to AA only. we must have that pI = pI + 2λ (otherwise firm B is not B A maximizing its profit). A B A B c c and πA = πB = 4λ.

hence.05/0. Hence. When Aα = Aβ = 2. advertising in this city serves as a mean by which each firm attracts consumers from the competing firm. we have that the profit maximizing ratio of advertising to revenue is AM 1 A = = . (c) Solution-Figure 11.1). Therefore AM = 10 × 0.5 mil. 2. the monopoly reduces the ratio of advertising expenditure to revenue. (b) Fortune teller β takes Aα as given and chooses Aβ to maximize his or her profit.05. ∂Aβ (Aβ )2 √ The second-order condition can be easily verified. Answers to Exercises 1. we have it that nα = nβ = 3. but cannot increase industry sales beyond the level determined by zero advertising.13. pQM − p 4 Hence.. For the purpose of this question. . %∆A and p = %∆Q = −0. AM = $2. (c) When the demand becomes more price elastic. πα = πb = 30 − 1 = 29. (a) Aα = Aβ = 1 implies that nα = nβ = 3. By symmetry. (b) Now. where an integral is computed. Aα = Rα (Aβ ) = 30Aβ . hence. Conclusion: if both firms spend the same amount on advertising. That is. perhaps. the analyses given in this chapter are pretty much straight forward. The appendix describing advertising regulations can be assigned as a home reading. (a) Since A = %∆Q = 0. reflecting the fact that there exist two Nash equilibria for this advertising game.1 illustrates that the two best response functions intersect twice. The first-order condition is given by 0= ∂πβ 30Aα = − 1. then industry profit is maximized at Aα = Aβ → 0.1. with the exception of the calculation of the socially optimal level of persuasive advertising given in equation 11. we define 0/0 (yes.Chapter 11 Advertising From a technical point of view. πα = πb = 30 − 2 = 28. p = −0. The two best-response functions are drawn in Solution-Figure 11.2. %∆p then by the Dorfman-Steiner condition (Proposition 11. Aβ = Rβ (Aα ) = 30Aα . zero divided by zero) as equal to 1.5 = $1 mil.5..

AN = AN = 30 which is also α β illustrated in Solution-Figure 11. P ) = θE ≥ π 1 (P.. 3. . we need to show that unilateral deviation is not profitable for each firm. . 0) = 30. Thus. . Hence. substituting Rα (Aβ ) into Rβ yields Aβ = 30 30Aβ . ✻ . . .... . industry’s profit is maximized and each firm earns πi (0. .. 30 . when θ ≥ N/(2E).. ✲ ... . in this example. . In order for (I. 30) = 30 − 30 = 0.. . . That is. .. i = α. which is lower than N (1. . .. I) = (1 − θ)E π 1 (I... N (d) In the interior NE.. at the other NE. . . P ) = N ≥ π 1 (I. 30) (see Solution-Figure 11.. π 2 (I. β. The interior NE (0. (e) The interior NE (30.1: Fortune tellers’ advertising best-response functions Then. . the NE that maximizes industry profit is unstable. 0) is unstable since any unilateral deviation by one of the firm would trigger a sequential chain of responses that would move the advertising equilibrium to (30. . πi (30.38 Advertising Aα Rβ (Aα ) Rα (Aβ ) . the industry’s profit is not maximized at the interior NE. .. . To obtain the other NE.. Thus. ✻. .. using Table 11. 0 30 ✲ Aβ Solution-Figure 11. .1.. .. 1) = 29..... . ✲ ...1. Howπi N ever.. P ) to constitute a NE outcome. .. P ) = N/2 when θ ≥ 1 − N/E.. .1 for an illustration). .. clearly Aα = Aβ = 0 constitutes one NE. 30) is stable.

p H − cL Hence.10 should be modified to pm = H and q m = αH + (1 − α)L − cL . Therefore.5). p The price and quantity pair in Proposition 12. Therefore. the high-quality monopoly would indeed find it profitable to signal its high quality to the consumers. Finally. and a low-quality producer with probability 1 − α. where 0 < α < 1. and Warranties This is a long chapter with a wide variety of topics dealing with supply and demand for quality.Chapter 12 Quality. Other sections use only simple calculations of expected values. .6 The model suggested below provides some improvement as it assumes that consumers are expected utility maximizers (similar to what is assumed in Section 12. and make a profit of ¯ ¯ (¯ − cL ) · 1 = αH + (1 − α)L − cL = αH + (1 − α)L − cL > 0. H − cL (∗) Had the monopolist been a low-quality producer. Then. in choosing this price-quantity pair the monopolist could signal his or her q=q quality to the consumers. since H > L > cH > cL .8) may be assigned as a home reading. he or she could have made the same profit by setting p and selling to all consumers [as is indicated in (*)] rather than setting p = pm and ¯ m . 1 I thank Keiichi Koda for proposing this improvement. The analyses are pretty much straight forward and do not require any preparation with the exception of section 12. Durability. An Alternative Model for Section 12. (H − cH ) αH + (1 − α)L − cL > αH + (1 − α)L − cH = (¯ − cH ) · 1. the maximum price that consumers are willing to pay is p = αH + (1 − α)L. The appendix on the legal approach to liability (section 12.2 that relies on the knowledge of location models developed in Chapter 7.1 Suppose that consumers believe that the monopoly is a high-quality producer with probability α. Then a low-quality producer could set the price at p. without any additional information. EU = αH + (1 − α)L − p.

2 For a given ae firm B chooses be to maximize πB which is given above. 1] implying that firm B would locate at be = 1. 2. the consumer who is indifferent between purchasing from A and not purchasing at all must satisfy Uz (A) = az − pA = 0.4). 7 We now verify that given ae = 4/7. φ(21/98) = 5. the profit maximizing location of firm A is ae = 4/7. (H −L)I2 < HpH −LpL . which is negative for 0 ≤ b < 21/98 and positive for b > 21/98.2) and the zero reservation utility assumption. b−a For a given pA . we have it that since consumer 1 buys the low-quality brand. b−a a 2pB − pA . (b) Follows directly from (12. 1]. we need to prove that φ(b) > 0 for all b ∈ [0. b−a yielding 0 = pB − 2pA 2pA − . since I2 < I1 by assumption.1). Solving the two first-order conditions yield the equilibrium prices. From (12. firm B would choose to locate on the east edge of the street. ∂b To demonstrate this inequality. given be = 1. b−a a pB − pA . yielding z = pA /a. (c) For a given pB .40 Answers to Exercises Quality. . Clearly φ (b) = 98b − 21. (H − L)I1 < HpH − LpL . then U1 (L) = L(I1 − pL ) > H(I1 − pH ) = U1 (H). Let Ui (k) denote the utility of consumer i when he buys the brand with quality k. Substituting into the profit functions yield the equilibrium profit levels. Thus. (4 − a)2 4 yielding ae = . Now. firm B chooses pB to max πB = pB 1 − pB yielding 0 = 1 − Second-order conditions are easily obtained. We want to show that U2 (L) = L(I2 − pL ) > H(I2 − pH ) = U2 (H). We would like to show that for ae = 4/7. Therefore. firm A chooses ae to max πa = a a(1 − a) .75 > 0. Hence φ(b) = ∂πB /∂b > 0 for all b ∈ [0. 2 Many students will find this part of the proof to be difficult. the function φ hits a minimum at b = 21/98. 0< 2. (d) We first verify that given be = 1. ∂πB (4/7. implying that H(I2 −pH ) < L(I2 − pL ). firm A chooses pA to max πA = pA pA pB − pA p A − . b) ≈ 49b2 − 21b + 8 ≡ φ(b). (a) Using the utility function (12. Durability. L and i = 1. Hence. k = H. and Warranties 1. Now.

Therefore. then π N W = ρV − c.L = 0. . (a) The monopoly’s production cost is the initial production cost plus the expected (as most once) replacement cost. 3 The purpose of this question is demonstrate that used cars are traded when some good-used-car owners are “forced” to sell their car. change the last sentence in this question to “. Under these prices. because they leave the country. then V s.5N L − pN = 0. This surplus equals V if the product does not fail at all and if the product fails only once (in which case it is replaced). . the monopoly will bundle the product with this warranty. 4. In contrast. under the assumed prices. if she or he buys a used car then ¯ ¯ ¯ s. ¯ ¯ Hence. Thus. Altogether. then π W = ρ(2 − ρ)V − (2 − ρ)c = (2 − ρ)(ρV − c).5U L − pU > 0. nobody buys a new car. under the assumed prices. then his or her expected utility is V b = 0. The probability of a double failure is (1 − ρ)2 . if a new buyer buys a new car. Hence.5N G +0. Note: If you are using the first printing of the first edition. . Lemon-used-car sellers: If a lemon-used-car seller sells his or her lemon and buys a new car. the probability of no double failure is 1 − (1 − ρ)2 = ρ(2 − ρ). (b) The maximum monopoly price is the expected consumer surplus from purchasing this product with this type of warranty. if a new buyer buys a used car then V b = 0.5N G − pN + pU = pU . we seek to characterize the demand and supply patterns of our agents. In contrast. the four types of agents. . which is equal to c + (1 − ρ)c = (2 − ρ)c.Quality. By assumption pU < 0. say. Hence. If the monopoly sells with no warranty.” .5U G + 0.3 New buyers (who do not yet own a car): Recalling our assumption that N L = U L = 0. π W > π N W since it is assumed that ρV > c. and Warranties 41 3. Durability.L = 0. hence. new buyers buy used cars. pM = ρ(2 − ρ)V .5U G − pU + pU = 0. (c) If the monopoly provides this warranty.5U G .5U G . lemon-used-car ¯ V ¯ ¯ sellers buy (and sell) used cars. . Good-used-car sellers: The question simply assumes that good-used-car owners must leave the country and therefore sell their good-used cars.

.

Therefore. and for k < k. In the Winter.1. Ask for the errata file by e-mail. and M RS (qS ) = 5 − qS = c = 1. 000 copies.1: The cost of photocopying with or without two-part tariff (b) The number of copies yielding identical cost under the two payment systems is found ˆ ˆ ˆ ˆ by equating 0. .05k = 300 + 0. Section 13. 2. we can conclude that Winter is the high season. However. (a) The two cost functions are drawn in Solution-Figure 13.4 presents an attempt to endogenize the seasons in determining profit maximizing peak-load pricing.1 Answers to Exercises 1.2 are wrong. k = 10. and (ii) to equate the “low-season” marginal revenue to the per-passenger cost only. capacity is determined by having the monopoly equating M RW (qW ) = 10 − 2qW = r + c = 2. for k > k.02k.” we need to be careful about the capacity constraint determined by the Winter condition.4. (a) By Proposition 13. ˆ the fixed-per-unit price system the two-part tariff system is less costly. to find the profit maximizing pricing structure we need (i) to equate the “high-season” marginal revenue to the cost of an aircraft seat plus the per-passenger cost. $ T C(k) = 0.Chapter 13 Pricing Tactics: Two-Part Tariff and Peak-Load Pricing The analyses given in this relatively short chapter rely mainly on logical arguments and therefore do not require any special technical preparation. 1 If you are using the first printing of the first edition note that the calculations of consumer surplus in section 13. is less costly. Thus. and therefore may be given a lower priority in case the instructor is short of time.05k ✻ T C(k) = 300 + 0. since the two demand curves are “relatively close.02k • 300 ✲ k (# copies) ˜ ˜ k Solution-Figure 13. Given that the Summer demand curve lies completely below the Winter demand curve.

. pW = 6.5. pS = 3. hence. Altogether. π I = 6 × 4 + 3 × 4 − 2 × 4 − 1 × 4 = 24. hence. the airline will fly only 3 passengers during the Summer season. The second equality implies that qS = 4. the monopoly equates M RW (qW ) = 10 − 2qW = r + c = 4. Altogether. qW = 3.5 × 3 − 4 × 3 − 1 × 3 = 16. Thus. and M RS (qS ) = 5 − qS = c = 1. (b) Similar to the previous subquestion. hence. π I = 7 × 3 + 3. Hence. due to the capacity constraint. The second equality implies qS = 4 > 3 = qW .5. The first equality implies that. pS = 5 − 3/2 = 3.44 Pricing Tactics: Two-Part Tariff and Peak-Load Pricing The first equality implies that qW = 4. pW = 7.

Clearly. in which first-period students believe that no new edition will be published. Answers to Exercises 1.17) is derived by comparing the monopoly’s second period profit only and therefore yields the condition under which a new edition will not be introduced in the second period. First period students will undercut . mixed tying (which yields an identical allocation as no tying) generates a higher profit than pure tying. in which case the publisher can charge a higher price in the first period by adding the resale value to the price. By (14. (a) When a new edition is not introduced in the second period. (a) We need to compare the profit levels under the commitment of not introducing a new edition. Therefore. to the case where there is no commitment and a new edition is introduced. the condition given in (14. (b) The profit maximizing product and package prices are pM T = pM T = 4 and pM T = X Y 8. 2. Altogether. only consumer 2 buys a package.16) and hence in (14. the monopoly cannot charge consumer 2 more than the sum of the prices of the two products sold separately. all three consumers buy the packages. (a) Under pure tying. the condition given in (14. (b) Indeed. and Dealerships This chapter introduces several topics related to profit enhancing marketing tools that were not discussed (at least not in full) in earlier chapters. π M T = 4 + 4 + 8 = 16.Chapter 14 Marketing Tactics: Bundling. hence. the monopoly competes with period 1 students on selling the first edition. the profit level under commitment is nV . implied by) the condition given in (14.. π T (10) = 10. π T (4) = 3 × 4 = 12. In contrast. hence the monopoly can raise the book’s price by its resale value. Comparing the two profit levels reveals that commitment not to introduce a new edition is more profitable if F > nV − 2c. this condition in weaker (i. Upgrading. consumer 1 buys one unit of X. to the profit level when no commitment is made and the monopoly introduces a new edition.17).17) “does not take into account” that first period price can be raised by the resale value when a commitment is made not to introduce a new edition. in the present example. (c) Notice that mixed tying yields the same profit maximizing pricing structure as no tying. we need to check two options: (i) Under pT = 4. since in the present example. 3.16) and (14.e. In other words. the first option maximizes the profit under pure tying. and under no commitment is 2n(V − c) − F .17) since it compares in the case of commitment. In fact. (ii) Under pT = 10. hence. pure tying is profit reducing. consumer 3 buys one unit of Y and consumer 2 buys the package (or one unit of each good). Under this price structure.

pd = (a + c)/2. Thus. the monopoly would set pN = (1 − α)V . and therefore earn a profit of 2 N π2 = n[(1 − α)V − c]. used book sellers are undercut by the monopoly. the dealer chooses Q (sales) to max π d = (1 − φ)[p(Q)Q − cQ] = (1 − φ)[(a − Q)Q − cQ]. Hence. which is different from condition (14. then the monopoly undercuts the used-book sellers (since consumers do not value used books very much). (b) Under this contract. then π2 = 0. Hence. the maximum price 2 user-book sellers can charge is pU = 2 c − (1 − α)V 0 if c ≥ (1 − α)V otherwise Notice that if c < (1 − α)V . a new edition is introduced if n(V − c) − F ≥ n[(1 − α)V − c]. Upgrading. then pN = (1 − α)V − c (the monopoly undercuts used-book 2 N sellers).16). Under this condition. (a) Under this contract. (b) There are two cases: High value for used books (α > 1 − c/V ): A new edition would yield a secondN period profit of π2 = n(V − c) − F . Hence. or α < 1 − c/V .16) when α = 1. In this case. 4. a new edition is introduced if F ≤ n(V − c). Altogether. then the monopoly price is reduced to cost( pN = c). Low value for used book (α < 1 − c/V ): When α is low. if used-book sellers can undercut the monopolist. which is the same as condition (14. and Dealerships the monopoly (since first-period students do not have any production cost) and for a given pN will set pU so that 2 2 V − pN ≤ αV − pU . or if F ≤ nαN . π2 = n[(1 − α)V − c]. the dealer chooses Q that solves max π d = (1 − φ)p(Q)Q − cQ = (1 − φ)(a − Q)Q − cQ. Under this contract. the industry profit is πM + πd = φ (a − c)2 (a − c)2 (a − c)2 + (1 − φ) = . in a Bertrand equilibrium. Q . the dealer’s sales and price are independent of (1 − φ) (share of the profit of the dealer). Q The first-order condition yields Qd = (a − c)/2. and π d = (1 − φ)(a − c)2 /4. 2 2 Now.46 Marketing Tactics: Bundling. if a new edition is not introduced in the second period. When a new edition is not introduced. 4 4 4 which equals the industry’s profit when the firms vertically integrate into a monopoly given in (14.24).

industry profit is given by the sum πM + πd = = = < φQd pd + (1 − φ)Qd pd − cQd = pd − c Qd c a c − 2(1 − φ)c a + − 2 2(1 − φ) 2 2(1 − φ) c a−c a − 2 2 2(1 − φ) (a − c)2 .Marketing Tactics: Bundling. To see this. πT I = VN − VU + VN = 2VN − VU > VN = π ∗ . by assumption. where VN > VU > 0. current users trade-in their old refrigerator for a new one. Suppose that there are two consumers: one owns an old refrigerator and one that is a new buyer. The utility from using an old refrigerator is VU whereas the utility from buying a new one is max{VN − pN . Altogether. does not require any monitoring since the dealer pays up front for each unit it acquires from the manufacturer. 2 2(1 − φ) 47 Now. thereby selling two units and earn a profit of π L = 2(VN − VU ). maximizes industry’s profit) it is rarely observed since it requires having the manufacturer monitoring the profit made by the dealer. 0}. the per-unit contract analyzed in the text. π L < π ∗ . hence.. and also a new buyer buys a new one. This monitoring may be too costly for the manufacturer. (c) Whereas the “profit sharing” contract analyzed in part (a) is optimal (i. thereby N selling only to the first-time buyer and earning a profit of π ∗ = VN . . despite being not optimal. 5. Upgrading. Trade-in price policy: We now show that by employing a trade-in price policy.24). 4 Thus. note that the monopoly can set a lower price. Single price policy: I will argue that the profit maximizing price is p∗ = VN . We further assume that VN < 2VU . In contrast. whereas the price with no trade-in is pN = VN .e. We will demonstrate how a monopoly can benefit from establishing a trade-in procedure by constructing a simple model. 2 2(1 − φ) and pd = a c + . and Dealerships The first-order condition is (1 − φ)(a − 2Q) − c = 0. However. Note that under these prices. pL = VN − VU . the monopoly can price discriminate between new buyers and old-model owners. Suppose that the monopoly announces that current users can trade-in their old refrigerator for a new one for a price of pT I = VN −VU . Qd = a c − . this contract yields a lower profit than the industry’s profit when the firms vertically integrate into a monopoly given in (14.

.

which is recommended only if the students are capable of solving Cournot-type problems very easily. Answers to Exercises 1. and Regulation Instructors who choose to teach from this chapter are urged to go over the principal-agent problem (section 15. the NE effort levels decrease further below the optimal level when the number of scientists increases. (c) Yes. these three subsections analyze the principal-agent problem in different degrees of complexity. and q2 = .Chapter 15 Management. (b) Each scientist i takes all other effort levels (ej . The third subsection of introduces risk aversion implicitly by assuming asymmetric information. 2. the “social planner” chooses a common effort level for all scientists e that solves Ne V max w − e2 = = e − e2 . The remaining sections: executives’ compensation (section 15. Altogether. 2 The best response function of firm 2 (the firm that maximizes profit only) is given by q2 = BR2 (q1 ) = 0.4) do fit the more advanced students. In constrast. Therefore. since the optimal effort level is e∗ = 1/2 which is independent of the number of scientists. reflecting a stronger free-rider effect. (a) Under collusion. More precisely. thereby enabling the instructor to stop at the level that exceeds the students’ ability.2 introduces a simple free-rider team effort problem. yielding a NE where ei = 1/(2N ) for all i = 1. Section 15.5(a − c − q1 ). (a) The manager of firm 1 chooses q1 that solves max M1 = 0.1). N . A subsequent subsection introduces uncertainty with no risk aversion. j = i) as given and solves max Ui = ei V − (ei )2 = N j=i ej + ei N − (ei )2 . 2 . q1 yielding a best-response function given by q1 = BR1 (q2 ) = a − c + 1 − q2 . Section 15.5[(a − q1 − q2 )q1 + (1 − c)q1 ]. − e2 = e N N yielding e∗ = 1/2.4) and the regulation of the firm (section 15. q1 = 3 3 . the equilibrium output levels are given by a−c+2 a−c−1 e e .3 introduces Fershtman-Judd model. the first subsection illustrates the problem with no uncertainty. . Compensation. . which is developed three times in order to allow for a a gradual increase in the degree of difficulty.

O (b) The owner of firm 1 earns π1 = π1 −M1 . That is. Hence. π1 = π1 . and Regulation e e Clearly. q1 > q2 confirming that the managerial compensating scheme of firm 1 is output increasing at the expense of the firm that maximizes profit only (firm 2). Finally. so that we approximate the owner’s profit by the O firm’s profit. Formally. From the previous calculation we conclude e = (a − 2c − 1)/3. .29]).7)) yields that π1 ≥ π1 if and only if a ≥ 2 + c. π e = (a − c − 1)(a − c + 2)/9. the present compensation scheme dominates a simply Cournot behavior (given that the other firm behaves in a Cournot fashion) if the demand intercept is sufficiently higher than the unit production cost. Compensation. We now set µ that p 1 1 sufficiently small (as in [15. comparing the profit of the owner of firm 1 under the present managerial compensation scheme to her profit when this manager maximizes profit only (the O c simple Cournot profit given in (6.50 Management.

Search theory (section 16. the students should find this section easy to follow. Hence.52 3 7..e.. From (16.1: Reservation price as a function of the search cost parameter (b) Solution-table 16.Chapter 16 Price Dispersion and Search Theory Price dispersion (section 16. (a) To see what happens. the consumer indifferent between searching or buying at random is located at L.0 Solution-Table 16. from different values of time). ˆ D N L→∞ L→∞ Hence. we look at the equilibrium prices when L → 0.. Hence.2) is introduced with no calculus. when search cost is zero). ¯ implying that the consumer will accept any price in the first store visit. 2.00 5 10. we have it that s p ¯ 0 1 1 4. se < L implying that the discount store a negative market share (which is ruled out ˆ by out assumption that α > 3/2).86 4 9.1 reveals that p = 1 when s = 0 (i. Note also that this result can be inferred from (16.1 reveals that for s ≥ 4 the consumer reservation price is p = 9. calculate. This question should read. the probability that this consumer finds prices exceed 1 in all store visits is limT →∞ (8/9)T = 0. (b) When α = 3/2. Thus.77 2 6. so the analysis relies on simple probability arguments. The probability that prices exceed 1 in two store visits is σ 2 = (8/9)2 . (c) Solution-table 16. when α = 3/2 all consumers buy at random.9) we see that lim pe = lim pe D = 2αH and se → 0. some consumers have a negligible search cost). when L becomes small (i. say. under the search cost you found in the previous subquestion.e. both stores charge the same price. implying that there is no discount store. The analysis relies on a simple of location model.1 the probability that the consumer finds the price higher than 1 (i.16) where the calculated expected number of stores visited is µ = 1/(1 − σ) = 9 < +∞. Answers to Exercises 1. the reservation price) in a single store visit is σ = 8/9. .e. 1 The first printing of the first edition does not specify which search cost is assumed. ¯ (d) When s = 0... When α = 1.1) demonstrates that a discount store and an ‘expensive’ store can coexist in a market where consumers have different search costs (resulting..13). (a) Using (16.

.

n. therefore if 3β < 3. 0. and (you guessed. however. Clearly. and are. of course). Answers to Exercises 1. this technology (cost function) exhibits economies of scope if T C(n. similar to (17. In future editions I will attempt to expand this part of the book by introducing at least three more industries: the health industry. from the utility function of route 3 passengers (17. n. the monopoly ˜ ˜ ˜ chooses f ≡ f1 = f2 that solves max π h = n¯3 + 2nδ + 2n f − 2cf. p3 = p3 ) we have it ¯ that the minimum frequency that the airline must provide on route 3 (at this airfare) ¯ ¯ is f3 = (¯3 − δ)2 . (a) Under the FC network.5). hence if β < 1. The main message here is that the introduction of a binding price cap by the regulator will cause the monopoly airline to reduce flight frequency (or any other quality aspect of the service). (3nα )β < 3nαβ .e.2). 0) + T C(0. From the definition. n). Since the airfare is regulated (i.Chapter 17 Miscellaneous Industries The purpose of this chapter is to demonstrate to the student that there is no general model that can describe all industries. the corresponding frequency f3 decreases. I also wish to demonstrate in this last chapter that the tools developed in this book are indispensable to the analysis of most industries. n) = (nα + nα + nα )β < (nα )β + (nα )β + (nα )β = T C(n. p f 1 ¯ The first printing of the first edition contains a typo: Replace “. Now. the lawyers. or. 2.” ¯ . p ¯ (b) Under the HS network. the defense industry.. each route can be analyzed separately. the flight frequency on route 3 is the (minimum) frequency ¯ on routes 1 and 2. worth learning. 0) + T C(0. but with a given p3 . 0. as p3 decreases. therefore..charges an airfare f3 ” with “airfare p3 .. This chapter discusses four types of industries that require special attention since (perhaps like any other industry) they do not fall in any category and therefore cannot be analyzed by simply picking one of the market structure analyzed earlier in the book.1 Hence. we see that the monopoly extracts maximum √ surplus when p3 = δ + f3 .

LC and not tC ). Then. yielding ns = C v+φ 4v 1/3 < ne .2 ˜ f= n 2c 2 < 3n 4c 2 = f h. is lower than the sum of the airfares ¯ on routes 1 and 2. v+φ 4v + τ. a price cap on route 3 will reduce the flight frequency on all routes. implying that ne = C v+φ v 1/3 . but it uses a different driving time function. the airfare on route 3. The analysis here follows the same stages as in section 17. (c) Let τ denote the highway toll on this route. the social planner sets τ s that solves v + φ = v(ns )3 + τ = v C yielding that τ s = 3(v + φ)/4. One has also to check that. Simply set tC = (nC )3 (the equation stated in this question equals to the value of the driving time. 3. v + φ = v(nC )3 . under the HS network. C Thus. Hence. we need to verify that (or find the conditions under which) p3 ≤ p1 + p2 . 2 . 3 The first printing of the first edition contains a typo. the social planner would recommend a policy that would result in a reduction in the number of people who drive a car. (b) The social planner chooses ns that solves C min Ls = (N − nc )(v + φ) + nC V (nC )3 .4. Thus. that is. Otherwise. in this equilibrium.3 (a) If some people take the train and some drive a car. ¯ route 3 passengers can purchase two separate tickets and reach their destination at a lower airfare. to set the toll that would generate the socially optimal number of highway users. Therefore.54 Miscellaneous Industries ˜ yielding a first order condition given by 0 = 2n/(2 f ) − 2c. p3 . it must be that LT = LC .

Sign up to vote on this title
UsefulNot useful