Lecture Notes Handout

Microeconomics 2 ITMI 1017

Prof. Swagata Bhattacharjee
Topic: Duopoly
Ref: Serrano and Feldman Ch 13; Varian Ch 28.

Oligopoly
So far, we have studied market structures characterized by perfect competition and monopoly. These are,
however, two extreme scenarios and rarely seen in practice. Most often a market is served by two more …rms
who have some amount of market power. An oligopoly is a market structure with few …rms. Here, each
…rm realizes that each of their decisions a¤ects the market equilibrium. So, here …rms act strategically, i.e.,
base their decisions on what the other …rms are most likely to do. We will discuss two ways these …rms in a
duopoly compete with each other: (A) quantity competition, and (B) Price competition.

Cournot Competition
If the …rms in an oligopolistic market compete with each other through their choices of quantity, this market
structure can be modeled as Cournot Oligopoly (…rst discussed by Antoine Augustin Cournot, 1810-77).
Let us assume there are two …rms in the market, each choosing how much to produce. Let y1 and y2 be
the quantity produced by …rm 1 and 2, respectively. The total market supply is then: y1 + y2 = y; and this
determines the market price through the negatively sloped inverse demand curve:

p(y) = p(y1 + y2 )

Assume that …rm i has the cost function given by: Ci (yi ):
Firm i wants to maximize its pro…t:

i (y1 ; y2 ) = yi p(y1 + y2 ) Ci (yi ) (1)

The basic assumption in Cournot model is: when each …rm decides how much to produce, it takes the
other …rm’s output choice as given and …xed. Therefore, …rm i maximizes its own pro…t by treating yj as
constant.
The FOC for each …rm i is:
@ dp(y) dCi (yi )
= p(y) + yi =0
@yi dy dy
| {z i } | {zi }
M Ri M Ci

For …rm 1; solving FOC gives an optimal output choice y1 as a function of the other …rm’s output choice:
y2 :
y1 = r2 (y2 )
and similarly, for …rm 2:
y2 = r1 (y1 )
These are called Reaction functions/ Best Response Functions and for every output level chosen by the
other …rm, determines the optimal pro…t maximizing output level to be chosen for each …rm.
A Cournot equilibrium is a pair of output levels y1 and y2 ; that are consistent: each …rm i is maximizing
its pro…t at yi ; given that the other …rm is choosing yj : I.e., in equilibrium,

y1 = r2 (y2 );
y2 = r1 (y1 ) (2)

In short, Cournot equilibrium is a self-sustaining, consistent, self-reinforcing outcome in this duopoly scenario.
(They are the Nash Equilibrium, as you will …nd out later in your Game Theory course.)

1
Cournot Market With Multiple Firms
Suppose there are n …rms in the market, each setting quantity to produce. Total market output is:

y = y1 + y2 + ::: + yn

The inverse market demand curve is given by:

p(y) = p(y1 + y2 + ::: + yn )

Each …rm wants to maximize pro…t:

i (y1 ; y2 ; :::; yn ) = yi p(y) Ci (yi ) (3)

FOC for each …rm:

@ dp(y) dCi (yi )

= p(y) + yi =0 (4)
@yi dy dy
| {z i } | {zi }
M Ri M Ci

Solving these n equations for the n output levels, we …nd the Cournot equilibrium: (y1 ; :::yn ; p ):
From the FOC 4, we get:
dp(y)
p(y) + yi = M Ci (yi )
dyi
dp y yi
p(y)[1 + ] = M Ci (yi )
dy p y
si
p(y)[1 ] = M Ci (yi ) (5)

where si = yyi = market share of …rm i:

This looks like a monopolist’s optimization, just with the relevant elasticity of demand curve that each
…rm faces being si instead of the entire demand elasticity. Clearly, the smaller the market share for …rm i;
the more elastic the demand curve it faces.
In case of a monopoly, si = 1; so the FOC 4 describes monopolist’s optimal, whereas when the market
has in…nitely many …rms, n ! 1; si ! 0; hence the FOC 4 gives

p = MC

which resembles perfect competition. This is one justi…cation for perfect competition: with may …rms a
Cournot oligopoly resembles perfect competition.

Example:
Assume the market demand curve is given by:

p(y1 + y2 ) = 100 y

Cost curves are:

C1 (y1 ) = 25y1
C2 (y2 ) = 25y2

Firm 10 s and 20 s pro…t, respectively:

1 (y1 ; y2 ) = y1 [100 (y1 + y2 )] 25y1

2 (y1 ; y2 ) = y2 [100 (y1 + y2 )] 25y2

The FOC for Firm 1:

@
= 100 2y1 y2 25 = 0 (6)
@y1

2
 Solving:
y2
y1 = r1 (y2 ) = 37:5 (7)
2
So, this is the reaction function/ best response function for Firm 1.
And, for Firm 2:
@
= 100 2y2 y1 25 = 0 (8)
@y2
So, the reaction function for …rm 2:
y1
y2 = r2 (y1 ) = 37:5 (9)
2
Equilibrium is the point where the reaction functions 7 and 9 intersect (See Figure 1). Solving these two
equations, we get:
y1 = y2 = 25
Total Market quantity is:
y1 + y2 = y = 50
Hence:
p = 100 50 = 50

Figure 1: Cournot Equilibrium

Each …rm earns a pro…t of 625:

Comparison with Perfect Competition:

If this market were served by perfectly competitive …rms, price would be pc = M C = 25: Total market
supply at this price would be: yc = 75: Firms would earn zero economic pro…t.
So, Cournot market produces less than the e¢ cient level of output and charges a higher price.

3
Comparison with Monopoly:
If this market were served by a monopolistic …rm, he would set

MR = MC
100 2y = 25
ym = 37:5
pm = 62:5

Clearly, Cournot market is less ine¢ cient than the monopolist, produces higher quantity and charges
lower price, so …rms obtain lower pro…t than the monopolist. The following Figure illustrates the social
surplus in each of these cases.

Figure 2: Comparison of Social Surplus in Cournot, Monopoly, and Perfect Competition

Collusion
In the previous example, if the …rms choose quantities independently as in a Cournot oligopoly, the total
pro…t is much less than the case when the market is served by a monopolist. If the …rms could collude, the
…rms would do better to act like a monopolist, set a quantity level that maximizes the joint pro…t, and then
share it among themselves. This collusion is termed as Cartel. Here, we will examine the situation when
two Cournot …rms collude.
Firm 1 and 2 choose to maximize the joint pro…t:

max (y1 ; y2 ) = [y1 + y2 ]p(y1 + y2 ) C1 (y1 ) C2 (y2 )

y1 ;y2

4
 FOCs:
@ dp(y)
= p(y) + y M C1 (y1 ) = 0
@y1 dy
@ dp(y)
= p(y) + y M C2 (y2 ) = 0
@y2 dy
So, at optimal point:
M R = M C1 = M C2
So, a …rm that has a cost advantage will always produce more output. From the example above, the
cartel will optimally produce:
y10 + y20 = 37:5
and since the cost conditions are same for both the …rms, each will produce half of it : y10 = y20 = 18:75;and
the market price will be
p0 = 62:5
Thus, each …rm will share a pro…t of
0 0
1 = 2= (100 37:5)18:75 25(18:75)
= 703:125 > 625(= ci )
Collusion is better for both the …rms.

Figure 3: Cartel

However, collusion is not easy to sustain in Cournot Oligopoly. Since the cartel is not an equilibrium,
each …rm has an incentive to cheat on the agreement. For example, if the …rms agree to produce 18:75 units,
for …rm 1, the optimal way to respond to …rm 2 producing 18:75 units is to increase its own output, because
according to the best response function,
y2
y1 = r1 (y2 = 18:75) = 37:5
2
= 28:125 > 18:75
As soon as Firm 1 deviates and produces this amount, it gets a pro…t of

1 = (28:125)(100 28:125 18:75) 25(28:175)

= 791 > 703:125

5
 But, discovering that Firm 1 has cheated, Firm 2 reacts to this new output level:

y2 = r2 (y1 = 28:125) = 23:44

And so on, until the original Cournot equilibrium (y1 = y2 = 25) is restored.

Q: Can collusions be ever sustained?

Yes! In a dynamic scenario, when the …rms in the market compete with quantities a-la-Cournot every period
and market conditions do not change, it is possible to sustain collusion by using punishment strategies.

Price Competition
Next, we consider the case where the …rms in an oligopolistic market decide on what price to charge rather

than the quantity to produce.

The price competition model where …rms produce homogeneous goods is formulated by Joseph Bertrand,
we will study that in Game Theory. Here, we consider the case where the products are slightly di¤erentiated.

Di¤erentiated Goods
If …rms are producing similar, but not identical goods, even with price competition, they can set separate
prices and still attract consumers. The …rms’demand curves will now be a function of the prices set by both
the …rms:

y1 = y1 (p1 ; p2 )
y2 = y2 (p1 ; p2 )

Thus, …rm 1 maximizes own pro…t:

1 (p1 ; p2 ) = p1 y1 (p1 ; p2 ) C1 (y1 (p1 ; p2 ))

And Firm 2 maximizes own pro…t:

2 (p1 ; p2 ) = p2 y2 (p1 ; p2 ) C2 (y2 (p1 ; p2 ))

The FOCs are, respectively:

@ 1 @y1 @y1
= y1 + p1 M C1 =0 (10)
@p1 @p1 @p1
@ 2 @y2 @y2
= y2 + p2 M C2 =0
@p2 @p2 @p2
This looks very similar to the Cournot model, just the …rms choose prices rather than quantities. Thus,
we get best response functions for each …rm i :

pi = ri (pj )

And solving these we …nd the consistent Bertrand equilibrium.

Note: With di¤erentiated goods, Bertrand equilibrium price is greater than MC.
From FOC for …rm 1 (Equation 10), we …nd:

@y1
y1 = (p M C)
@p1 1

@y1
If y1 > 0; the slope of the demand curve: @p1 < 0; hence p1 > M C:

6
Let us consider the same example as above with marginal costs M C1 = M C2 = 25
Assume the demand curve for …rm 1 and 2 are, respectively:
p2
y1 = 50 p1 +
2
p1
y2 = 50 p2 +
2
Then, …rm 1 maximizes:
p2
1 (p1 ; p2 ) = (p1 25)[50 p1 + ]
2
and Firm 2 maximizes:
p1
2 (p1 ; p2 ) = (p2 25)[50 p2 + ]
2
The FOCs give the reaction functions:
75 p2
p1 = r1 (p2 ) = +
2 4
75 p1
p2 = r2 (p1 ) = +
2 4
So, in equilibrium:

p1 = p2 = 25
y1 = y2 = 25
1 = 2 = 625

Figure 4: Bertrand Equilibrium with Di¤erentiated Goods