Monopoly power comes from a firm's ability to set prices. This ability is dictated by the shape of the demand curve facing that firm. If the firm faces a downward sloping demand curve, it is no longer a price taker but rather a price setter. In our perfect competition model, we assume there exist multiple participants, and because there are so many participants, the slice of the demand curve each firm sees is but a flat line. These firms are price takers.

There is a medium between monopoly and perfect competition in which only a few firms exist in a market. None of these firms faces the entire demand curve in the way a monopolist would, but each does have some power to set prices. A small collection of firms who dominate a market is called an oligopoly. A duopoly is a special case of an oligopoly, in which only two firms exist.

We will begin our discussion with an investigation of duopolies. For the following duopoly examples, we will assume the following:

- The two firms produce homogeneous and indistinguishable goods.
- There are no other firms in the market who produce the same or substitute goods.
- No other firms can or will enter the market.
- Collusive behavior is prohibited. Firms cannot act together to form a cartel.
- There exists one market for the produced goods.

In 1838, Augustin Cournot introduced a simple model of duopolies that remains the standard model for oligopolistic competition. In addition to the assumptions stated above, the Cournot duopoly model relies on the following:

- Each firm chooses a quantity to produce.
- All firms make this choice simultaneously.
- The model is restricted to a one-stage game. Firms choose their quantities only once.
- The cost structures of the firms are public information.

In the Cournot model, the strategic variable is the output quantity. Each firm decides how much of a good to produce. Both firms know the market demand curve, and each firm knows the cost structures of the other firm. The essence of the model is this: each firm takes the other firm's choice of output level as fixed and then sets its own production quantities.

The best way to explain the Cournot model is by walking through examples. Before we begin, we will define the reaction curve, the key to understanding the Cournot model (and elementary game theory as well).

A reaction curve for Firm 1 is a function
*Q*
_{1}
^{*}()
that takes as input the
quantity produced by Firm 2 and returns the optimal output for Firm 1 given Firm
2's production decisions. In other words,
*Q*
_{1}
^{*}(*Q*
_{2})
is Firm 1's best
response to Firm 2's choice of
*Q*
_{2}
. Likewise,
*Q*
_{2}
^{*}(*Q*
_{1})
is Firm 2's best
response to Firm 1's choice of
*Q*
_{1}
.

Let's assume the two firms face a single market demand curve as follows:

Q = 100 - Pwhere

MC_1 = 10

MC_2 = 12

Given this market demand curve and cost structure, we want to find the reaction
curve for Firm 1. In the Cournot model, we assume
*Q*
_{2}
is fixed and proceed.
Firm 1's reaction curve will satisfy its profit maximizing condition,
*MR* = *MC*
.
In order to find Firm 1's marginal
revenue, we first determine its
total revenue, which can be described as follows

Total Revenue = P * Q1 = (100 - Q) * Q1

= (100 - (Q1 + Q2)) * Q1

= 100Q1 - Q1 ^ 2 - Q2 * Q1

The marginal revenue is simply the first derivative of the total revenue with
respect to
*Q*
_{1}
(recall that we assume
*Q*
_{2}
is fixed). The marginal revenue
for Firm 1 is thus:

MR1 = 100 - 2 * Q1 - Q2\

Imposing the profit maximizing condition of
*MR* = *MC*
, we conclude that Firm 1's
reaction curve is:

100 - 2 * Q1* - Q2 = 10 => Q1* = 45 - Q2/2

That is, for every choice of
*Q*
_{2}
,
*Q*
_{1}
^{*}
is Firm 1's optimal choice of
output. We can perform analogous analysis for Firm 2 (which differs only in
that its marginal costs are 12 rather than 10) to determine its reaction curve,
but we leave the process as a simple exercise for the reader. We find Firm 2's
reaction curve to be:

Q2* = 44 - Q1/2

The solution to the Cournot model lies at the intersection of the two reaction
curves. We solve now for
*Q*
_{1}
^{*}
. Note that we substitute
*Q*
_{2}
^{*}
for
*Q*
_{2}
because we are looking for a point which lies on Firm 2's reaction curve as
well.

Q1* = 45 - Q2*/2 = 45 - (44 - Q1*/2)/2

= 45 - 22 + Q1*/4

= 23 + Q1*/4

=> Q1* = 92/3

By the same logic, we find:

Q2* = 86/3

Again, we leave the actual computation of
*Q*
_{2}
^{*}
as an exercise for the reader.
Note that
*Q*
_{1}
^{*}
and
*Q*
_{2}
^{*}
differ due to the difference in marginal costs.
In a perfectly competitive market, only firms with the lowest marginal cost
would survive. In this case, however, Firm 2 still produces a significant
quantity of goods, even though its marginal cost is 20% higher than Firm 1's.

An equilibrium cannot occur at a point not in the intersection of the two reaction curves. If such an equilibrium existed, at least one firm would not be on its reaction curve and would therefore not be playing its optimal strategy. It has incentive to move elsewhere, thus invalidating the equilibrium.

The Cournot equilibrium is a best response made in reaction to a best response and, by definition, is therefore a Nash equilibrium. Unfortunately, the Cournot model does not describe the dynamics behind reaching equilibrium from a non-equilibrium state. If the two firms began out of equilibrium, at least one would have an incentive to move, thus violating our assumption that the quantities chosen are fixed. Rest assured that for the examples we have seen, the firms would tend towards equilibrium. However, we would require more advanced mathematics to adequately model this movement.

The Stackelberg duopoly model of duopolies is very similar to the Cournot model. Like the Cournot model, the firms choose the quantities they produce. In the Stackelberg model, however, the firms do not move simultaneously. One firm holds the privilege to choose production quantities before the other. The assumptions underlying the Stackelberg model are as follows:

- Each firm chooses a quantity to produce.
- A firm chooses before the other in an observable manner.
- The model is restricted to a one-stage game. Firms choose their quantities only once.

To illustrate the Stackelberg model, let's walk through an example. Assume Firm 1 is the first mover with Firm 2 reacting to Firm 1's decision. We assume a market demand curve of:

Q = 90 - P

Furthermore, we assume all marginal costs are zero, that is:

MC = MC1 = MC2 = 0

We calculate Firm 2's reaction curve in the same way we did for the Cournot Model. Verify that Firm 2's reaction curve is:

Q2* = 45 - Q1/2

To calculate Firm 1's optimal quantity, we look at Firm 1's total revenues.

Firm 1's Total Revenue = P * Q1 = (90 - Q1 - Q2) * Q1

= 90 * Q1 - Q1 ^ 2 - Q2 * Q1

However, Firm 1 is not forced to assume Firm 2's quantity is fixed. In fact,
Firm 1 knows that Firm 2 will act along its reaction curve which varies with
*Q*
_{1}
. Firm 2's quantity very much relies on Firm 1's choice of quantity. Firm
1's Total Revenue can thus be rewritten as a function of
*Q*
_{1}
:

R1 = 90 * Q1 - Q1 ^2 - Q1 * (45 - Q1/2)

Marginal revenue for firm 1 is thus:

MR1 = 90 - 2 * Q1 - 45 + Q1

= 45 - Q1

When we impose the profit maximizing condition
(*MR* = *MC*)
, we find:

Q1 = 45

Solving for
*Q*
_{2}
, we find:

Q2 = 22.5

Although much of the logic behind the Stackelberg model is used in the Cournot model, the two outcomes are radically different: being the first to announce creates a credible threat. In the Cournot model, both firms make their choices simultaneously and have no communication beforehand. In the Stackelberg model, Firm 1 not only announces first, but Firm 2 knows that when Firm 1 announces, Firm 1's actions are credible and fixed. This demonstrates how a slight change in the flow of information can drastically impact the outcome of a market.

The Bertrand duopoly Model, developed in the late nineteenth century by French economist Joseph Bertrand, changes the choice of strategic variables. In the Bertrand model, rather than choosing how much to produce, each firm chooses the price at which to sell its goods.

- Rather than choosing quantities, the firms choose the price at which they sell the good.
- All firms make this choice simultaneously.
- Firms have identical cost structures.
- The model is restricted to a one-stage game. Firms choose their prices only once.

Although the setup of the Bertrand Model differs from the Cournot model only in the strategic variable, the two models yield surprisingly different results. Whereas the Cournot model yields equilibriums that fall somewhere in between the monopolistic outcome and the free market outcome, the Bertrand model simply reduces to the competitive equilibrium, where profits are zero. Rather than take you through a series of convoluted equations to derive this result, we will simply show there could be no other outcome.

The Bertrand equilibrium is simply the no profit equilibrium. First, we will demonstrate that the Bertrand outcome is indeed an equilibrium. Imagine a market in which two identical firms sell at market price P, the competitive price at which neither firm earns profits. Implicit in our argument is our assumption that at equal price, each firm will sell to half the market. If Firm 1 were to raise its price above the market price P, Firm 1 would lose all its sales to Firm 2 and would have to exit the market. If Firm 1 were to lower its price below P, it would be operating below cost and therefore at a loss overall. At the competitive outcome, Firm 1 cannot increase profits by changing its price in either direction. By the same logic, Firm 2 has no incentive to change prices. Therefore, the no profit outcome is an equilibrium, in fact a Nash equilibrium, in the Bertrand model.

We now demonstrate uniqueness of the Bertrand equilibrium. Naturally, there can be no equilibrium where profits are negative. In this case, all firms would operate at a loss and exit the market. It remains to be shown that there is no equilibrium where profits are positive. Imagine a market in which two identical firms sell at market price P, which is greater than cost. If Firm 1 were to raise its price above the market price P, Firm 1 would lose all its sales to Firm 2. However, if Firm 1 were to lower its price ever so slightly below P (while still remaining above MC), it would capture the entire market at a profit. Firm 2 is faced with the same incentives, so Firm 1 and Firm 2 would undercut each other until profits are driven to zero. Therefore no equilibrium exists when profits are positive in the Bertrand model.

You may ask yourself why firms don't agree to work together to maximize profits for all rather than competing amongst themselves. In fact, we will show that firms do benefit when cooperating to maximize profits.

Assume both Firm 1 and Firm 2 face the same total market demand curve:

Q = 90 - Pwhere P is the market price and Q is the total output from both Firm 1 and Firm 2. Furthermore, assume that all marginal costs are zero, that is:

MC = MC1 = MC2 = 0

Verify that the reaction curves according to the Cournot model can be described as:

Q1* = 45 - Q2/2

Q2* = 45 - Q1/2

Solving the system of equations, we find:

Cournot Equilibrium: Q1* = Q2* = 30

Each firm produces 30 units for a total of 60 units in the market place.
*P*
is
then 30 (recall
*P* = 90 - *Q*
). Because
*MC* = 0
for both firms, the profit for
each firm is simply 900 for a total profit of 1,800 in the market.

However, if the two firms were to collude and act as a monopoly, they would act
differently. The demand curve and the marginal costs remain the same. They
would act together to solve for the total profit maximizing quantity
*Q*
.
Revenues in this market can be described as:

Total Revenue = P * Q = (90 - Q) * Q

= 90 * Q - Q^2

Marginal Revenue is therefore:

MR = 90 - 2 * Q

Imposing the profit maximizing condition
(*MR* = *MC*)
, we conclude:

Q = 45

Each firm now produces 22.5 units for a total of 45 in the market. The market price P is therefore 45. Each firm makes a profit of 1,012.5 for a total profit of 2,025.

Notice that the Cournot equilibrium is much better for the firms than perfect competition (under which no one makes any profits) but worse than the collusive outcome. Also, the total quantity supplied is lowest for the collusive outcome and highest for the perfectly competitive case. Because the collusive outcome is more socially inefficient than the competitive oligopoly outcome, the government restricts collusion through anti-trust laws.

We now extend the Cournot Model of duopolies to an oligopoly where n firms exist. Assume the following:

- Each firm chooses a quantity to produce.
- All firms make this choice simultaneously.
- The model is restricted to a one-stage game. Firms choose their quantities only once.
- All information is public.

Recall that in the Cournot model, the strategic variable is the output quantity. Each firm decides how much of a good to produce. All firms know the market demand curve, and each firm knows the cost structures of the other firms. The essence of the model: each firm takes the other firms' choice of output level as fixed and then sets its own production quantities.

Let's walk through an example. Assume all firms face a single market demand curve as follows:

Q = 100 - Pwhere

MC_i = 10 for all firms I

Given this market demand curve and cost structure, we want to find the reaction
curve for Firm 1. In the Cournot model, we assume
*Q*
_{i}
is fixed for all firms
*i*
not equal to 1. Firm 1's reaction curve will satisfy its profit maximizing
condition,
*MR*1 = *MC*1
. In order to find Firm 1's marginal revenue, we first
determine its total revenue, which can be described as follows

Total Revenue = P * Q1 = (100 - Q) * Q1

= (100 - (Q1 + Q2 +...+ Qn)) * Q1

= 100 * Q1 - Q1 ^ 2 - (Q2 +...+ Qn)* Q1

The marginal revenue is simply the first derivative of the total revenue with
respect to
*Q*
_{1}
(recall that we assume
*Q*
_{i}
for
*i*
not equal to 1 is fixed).
The marginal revenue for firm 1 is thus:

MR1 = 100 - 2 * Q1 - (Q2 +...+ Qn)

Imposing the profit maximizing condition of
*MR* = *MC*
, we conclude that Firm 1's
reaction curve is:

100 - 2 * Q1* - (Q2 +...+ Qn) = 10

=> Q1* = 45 - (Q2 +...+ Qn)/2

*Q*
_{1}
^{*}
is Firm 1's optimal choice of output for all choices of
*Q*
_{2}
to
*Q*
_{n}
.
We can perform analogous analysis for Firms 2 through
*n*
(which are identical
to firm 1) to determine their reaction curves. Because the firms are identical
and because no firm has a strategic advantage over the others (as in the
Stackelberg model), we can safely assume all would output the same quantity.
Set
*Q*
_{1}
^{*} = *Q*
_{2}
^{*} = ... = *Q*
_{n}
^{*}
. Substituting, we can solve for
*Q*
_{1}
^{*}
.

Q1* = 45 - (Q1*)*(n-1)/2

=> Q1* ((2 + n - 1)/2) = 45

=> Q1* = 90/(1+n)

By symmetry, we conclude:

Qi* = 90/(1+n) for all firms I

In our model of perfect competition, we know that the total market output
*Q* = 90
, the zero profit quantity. In the
*n*
firm case,
*Q*
is simply the sum of
all
*Q*
_{i}
^{*}
. Because all
*Q*
_{i}
^{*}
are equal due to symmetry:

Q = n * 90/(1+n)

As
*n*
gets larger,
*Q*
gets closer to 90, the perfect competition output. The
limit of
*Q*
as
*n*
approaches infinity is 90 as expected. Extending the
Cournot model to the
*n*
firm case gives us some confidence in our model of
perfect competition. As the number of firms grow, the total market quantity
supplied approaches the socially optimal quantity.