• Chapter 28 Oligopoly
• Key Concept: We have the first taste of
how to solve a game!
• Stackelberg, Cournot, Bertrand.
• Chapter 28 Oligopoly
• It is the case that lies between two
extremes, i.e., pure competition and pure
monopoly.
• Often there are a number of competitors
but not so many as to regard each of them
as having a negligible effect on price.
• This is the situation known as oligopoly.
• We will look at how strategic interactions
arise.
• There are several relevant models since
firms may behave in different ways in an
oligopolistic environment.
• It is unreasonable to expect one grand
model since many different behavior
patterns are observed in real life.
• We will introduce some models to
understand some possible patterns.
• We will restrict to the case of two firms or
duopoly.
• The duopoly case captures important
essence of strategic interactions.
• We will also focus on cases in which each
firm is producing an identical product so
that strategic interactions instead of
product differentiation is the key.
• The first model is the Stackelberg model,
for instance, Apple and its follower.
• It is used to describe industries with a
dominant firm or a natural leader.
• A typical pattern is smaller firms wait for
the leader’s announcement of new
products and then adjust their decisions
accordingly.
• In this case we could model the dominant
firm as the Stackelberg leader and the
other as followers.
• It is a sequential quantity setting game.
• 1 is the leader who chooses to produce y1
while after seeing that, the follower, 2
decides to produce y2.
• The total output of the market is therefore
y1+y2 and the price is p(y1+y2).
• What should the leader do?
• Depends on how the leader thinks that
follower will react to its choice.
• The leader should expect that the
follower will attempt to maximize profits
given the choice of the leader.
• Hence for the leader to have a sensible
decision, it has to consider the follower’s
maximization.
• This suggests we could solve backwards.
• Follower’s problem is
• maxy p(y1+y2)y2-c2(y2)
2
• FOC: p(y1+y2)+p’(y1+y2)y2 = c2’(y2)
• FOC: p(y1+y2)+p’(y1+y2)y2=c2’(y2)
• The marginal unit gives p, but it pushes
all units sold before
• So basically, from FOC, we can derive
the reaction function of 2.
• That is, given y1, there is an optimal level
of y2 or y2 =f(y1).
• FOC: p(y1+y2)+p’(y1+y2)y2=c2’(y2).
• In the following we will work with the
case where p(Y)=a-bY and c2(y2)=0 for
all y2.
2=(a-b(y1+y2))y2=ay2-by1y2-by22
• FOC: a-by1-2by2=0 or y2=(a-by1)/2b, 2’s
reaction function.
2=(a-b(y1+y2))y2=ay2-by1y2-by22
• FOC: a-by1-2by2=0 or y2=(a-by1)/2b, 2’s
reaction function.
• The isoprofit line 2=ay2-by1y2-by22=k
• This looks similar in shape to -y1-y22=0 or
y2=-x (verify by yourself).
2=(a-b(y1+y2))y2=ay2-by1y2-by22
• Given y2, if y1 is smaller, the profit of
firm 2 is bigger.
• Hence firm 2’s profits increase as we
move further to the left.
• Two points worth mentioning.
• When y1=a/b (1 has flooded the market),
then y2=0.
• On the other hand, when y1=0, it is as if 2
is the monopolist, so y2=a/2b.
• For each y1, 2 will choose its output to
make its profit as large as possible or pick
y2 so that its isoprofit line is furthest to
the left.
• The optimum will satisfy the tangency.
The slope of isoprofit line will be vertical
at the optimal choice.
• Let us suppose c1(y1)=0 for all y1 and
work out the leader’s problem.
• maxy p(y1+y2)y1-c1(y1)
1
• s.t. y2 =f(y1)
• Now, we can plug in 2’s reaction curve.
So 1=p[y1+ f(y1)]y1
• For the linear demand case, we can plug
in 2’s reaction curve.
• So 1=(a-b[y1+(a-by1)/2b])y1=(a-by1)y1/2.
• FOC: a/2-by1=0 So y1=a/2b.
• Plugging this into 2’s reaction curve, we
get y2=[a-b(a/2b)]/2b=a/4b.
• Show this graphically.
• We draw firm 1’s reaction curve in order to
help draw firm 1’s isoprofit.
• They have the same shape as those of firm
2’s, except they are rotated 90 degrees.
• Since firm 2 will choose an output so that 1
and 2’s choices will be on 2’s reaction
curve, 1 wants to choose an output
combination on the reaction curve that
gives it the highest profit.
• This means picking an isoprofit that is
tangent to the reaction curve.
• Two firms simultaneously decide output
levels.
• In this case each firm has to forecast the
other firm’s output choice.
• We seek an equilibrium in forecasts, a
situation where each firm finds its belief
about the other firm to be confirmed.
• This is the Cournot model.
• We look for the case where 1’s output is a
best response to 2’s and vice versa 2’s is a
best response to 1’s.
• Graphically, it is the intersection of the two
reaction curves.
• Since y2=(a-by1)/2b and symmetrically
y1=(a-by2)/2b, solving these two together,
we get y1=y2=a/3b.
• The Cournot equilibrium can be justified
by the following dynamics.
• Suppose the two quantity setting firms get
together and attempt to set outputs so that
their joint profit is maximized, i.e., they
collude or form a cartel, what will the
output levels be?
• Their problem becomes
• maxy ,y (a-b(y1+y2))(y1+y2)
1 2
• FOC: y1+y2=a/2b
• Together they should produce the
monopoly output.
• But there is the temptation to cheat.
• Look at FOC again.
• Profit is PY=(a-bY)Y where Y=y1+y2
• FOC: P+(dP/dY)Y=0 or P+(dP/dY)
(y1+y2)=0
• P+(dP/dY)(y1+y2)=0
• From 1’s perspective, increasing output
(if 2’s output is fixed), its profit increases
by P+(dP/dY)y1=-(dP/dY)y2>0.
• Hence it is a big issue for the cartel to
deter members from cheating.
• P+(dP/dY)(y1+y2)=0
• Intuitively, when 1 consider increasing
output, the extra profit from selling the
marginal unit is P, but the joint monopoly
will have to lower the price for all units
sold before.
• However, those include sold by firm 2.
• Often some repeated interactions would
help.
• Consider the punishment strategies. Each
firm produces half of the monopoly
output and gets profit m.
• If there is any cheating in the past, switch
to Cournot competition forever and each
gets c.
• So if a firm deviates, it can at best get d
for one shot and then forever it gets c.
• Note that d> m > c.
• [1/(1+r)]+[1/(1+r)]2+[1/(1+r)]3+…
• =1/r
• So a firm is comparing
m +m /r> d+c/r.
• As long as r<(m-c)/(d-m) or future
important enough, it will not deviate.
• The intuition is that a deviation in one
shot is not worthwhile as long as future is
important enough, making punishment
severe enough.
• Sometimes, price matching is one
possibility to detect deviation and hence
is used to maintain collusion.
• Lastly, we talk a bit about Bertrand
competition.
• Two firms simultaneously announce
prices pi and pj.
• There is a market demand x(p).
• If pi<pj, then xi(pi,pj)=x(pi).
• If pi = pj, then xi(pi,pj)=x(pi)/2.
• If pi>pj, then xi(pi,pj)=0.
• Suppose both firms have marginal cost c.
• There is a unique pure Nash equilibrium.
• Suppose pi≤pj and pi<c, then i can deviate
to pic.
• Suppose pi=c and pj>c, i can increase the
price a bit to earn positive profit.
• Suppose c<pi≤pj, j earns at most (pi-
c)x(pi)/2, undercutting i a bit, it could
earn close to (pi-c)x(pi).
• We are left with the case where pi=pj=c.
• Chapter 28 Oligopoly
• Key Concept: We have the first taste of
how to solve a game!
• Stackelberg, Cournot, Bertrand.
