1

JV; October 2016 – ECON 4925 (2016)

Lecture Note 2: Resource Extraction and Imperfect Competition1

This note presents some results related to resource extraction when we have a market
structure between perfect competition and pure monopoly. Perfect competition has
been the topic of several lectures, whereas monopoly extraction is covered in Stiglitz
(1976) and the lecture on Sept 14. 2

We focus first on a model with a dominant firm (a cartel, like OPEC) and a competitive
fringe (independent resource owners, outside the cartel). We start with a static model so
as to provide a simple representation of the “leader-follower”-notion, which highlights
the importance of the sequence of moves by the players. (The natural equilibrium
concept used here is the so-called Stackelberg equilibrium.) Thereafter we reverse the
sequence of moves, and consider the resulting outcome as a Nash-Cournot equilibrium,
and show the meaning of dynamic inconsistency within this setting. Then we turn to a
dynamic representation of this dominant firm - competitive fringe model, but
concentrate mainly on one solution concept - the so-called open-loop Nash equilibrium (a
binding contract equilibrium or a pre-commitment equilibrium). In this game the
dominant firm announces ex ante an extraction plan, while the members of the fringe,
while pursuing their extraction plans, believe that the dominant firm will honour or
stick to its announced plan. The strategies for all players will therefore be functions only
of time and the resulting market outcome will constitute a Nash equilibrium. However,
such an open-loop equilibrium will normally not be dynamically consistent, or
constitute a subgame perfect Nash equilibrium. One should, by the way, bear in mind
that there is an important distinction between subgame perfection and time or dynamic
consistency, according to Karp and Newbery (1993; p. 891): “Perfect equilibria must be
credible even if some agent deviates; time-consistent equilibria are defined without

1 This note is covering a theme, “Oligopolistic Extraction”, which has not been covered in class, but is
meant as additional reading so as to see the role of mixed market structure. The other theme, “Dominant
Firm – Competitive Fringe” was very briefly presented, but is not covered by the literature on the reading
list. However, this issue, which is rather difficult, is so important that you should have some knowledge
about it.
2 Joseph Stiglitz (1976), Monopoly and the Rate of Extraction of Exhaustible Resources, American Economic

Review, 66 (4), pp. 655 – 661.

 2

regard to out-of-equilibrium behavor.”3 Because transactions are made between

sovereign nations, “… contracts can only be enforced by mutual self-interest and not by
appeal to higher authority…”, to quote Newbery (p. 617) 4. Hence, if the dominant firm
should announce an extraction plan (“that seems to be the best one for the cartel”), then,
if the fringe should believe that this plan will not be revised later, the dominant firm
will normally have an incentive not to stick to the original plan, because it is possible to
achieve an even higher profit from deviating from the original extraction plan. Because
we assume that players are rational, the fringe will correctly predict that the dominant
firm will deviate (at the same time the dominant firm is also able to predict the best
strategy for the fringe and hence the outcome of the game), and the only outcome that
no player will renege on, is the Nash equilibrium, not only for the entire game, but also
for any conceivable subgame (or for any conceivable state that might happen). We
should therefore look for a plan that is time consistent or involving a plan for any
player that is best, and therefore credible, from any point in time, in the sense that it
eliminates all empty threats. However, finding such (closed-loop) equilibria is normally
very difficult because we have to leave the principle of optimality underlying optimal
control theory which relies on the fact that future states or future values of the variables
values do not affect current decisions.5 At last we sketch a rather simple dynamic model
for resource extraction under an oligopolistic (or duopolistic) market structure. This is a
model where there is some “strategic interaction” through the impact each resource
owner has on the market price. We describe the dynamic (open-loop) Nash Equilibrium
under some rather restrictive assumptions.

Let us start with the introduction made by David Newbery, in a working paper from
1980, on “Credible Oil Supply Contracts”. He says: “The world oil market differs from
the conventional text-book commodity market in three important respects. As oil is an
exhaustible resource, current supply decisions depend on future prices, but there are no
futures markets, and so suppliers must predict the future.6 Second, the market is

3 Larry Karp and David M. Newbery (1993), Intertemporal Consistency Issues in Depletable Resource, in
Handbook of Natural Resources and Energy Economics, vol. III, edited by A.V. Kneese and J.L. Sweeney,
Elsevier Science Publiushers B.V.
4 David Newbery (1981), Oil Prices, Cartels, and the Problem of Dynamic Inconsistency”, Economic Journal,

91 (363), pp. 617 – 646.

5 A closed-loop equilibrium is a Nash equilibrium with strategies that depend both on time and the state.

6 Comment by JV: The time period between deciding to start operating an oil field (and building up

production capacity) and when to start extracting may be very long, up to 10 years.
 3

dominated by a small number of large suppliers who face competition from a fringe of
small producers. The market structure is neither one of pure monopoly, nor perfect
competition, but closer to the more complex form of a dominant producer (in this case a
cartel) facing a competitive fringe. Finally, oil is a highly politicised commodity,
conferring substantial scarcity rents to those in a position to exploit them, and traded
between sovereign nation states. In such a situation, the only credible long term supply
contracts are those which continue to be mutually advantageous, for there is no
supranational legislature to ensure that such contracts are honoured.”

The issue of dynamic inconsistency can the illustrated by having a cartel, facing
competition by a fringe, when the cartel wants to produce after the fringe has exhausted
its reserves, and from then on operate as a monopolist. In that case we can imagine that
the cartel will announce that it will sell oil at a low price in the future, which might
induce the fringe to sell oil early and therefore speed up extraction so that the fringe
reserves are depleted sooner. After the fringe is left with no reserves and hence left the
market, the cartel remains the sole producer acting as a monopolist, having an obvious
incentive to renege on the former incredible promise of selling at a low price, and rather
raise the price. In a rational equilibrium, the fringe will see through this line of
reasoning and will not find the original cartel announcement credible, and hence will
not speed up extraction early. The outcome sketched above cannot be part of
equilibrium. (If a price path should make a jump at some point in time in the future – a
jump foreseen by the players having rational expectations – such a price path cannot be
part of equilibrium, because some producers will have an incentive to change their
initial plans. As long as there are incentives to revise the plans announced earlier, we
cannot have an equilibrium. This is similar to the problem we discussed in Lecture #3
(Sept 7), with sequential extraction from several deposits having different extraction
costs, where we explained why the equilibrium price path is continuous.)

a) A dominant firm – competitive fringe model: The static case

Consider the following market structure: We have a profit-maximizing cartel and a
competitive fringe (consisting of a number of small, independent producers, each
maximizing profit as price taker). The cartel is a leader, in the sense of setting the price
that is taken as exogenous by the fringe, so as to maximize the profit given the fringe’s
supply response. The fringe members take their output decisions after the dominant
firm has announced a price.
 4

Suppose market demand is known by all parties to be D(p) , with D (p) 0 , and a
supply response taken by the fringe as given by S (p) , being increasing in the price set by
the cartel. This supply function is common knowledge. The cartel’s objective is to
maximize the profit, as a monopolist on the residual demand curve, given by the
difference between market demand and the supply of the fringe; D(p) S (p) , when
assuming no cost of production for the cartel. (This residual demand facing the cartel
can be zero if the price is set so high that the fringe takes over the entire market, or can
be equal to the market demand if the price is below the lowest price compatible with
any production by the fringe; i.e. below a lower bound p for which S (p) 0 .) The
monopolist therefore will choose a price so that p[D(p) S (p)] is maximized. If we
suppose that p[D(p) S (p)] : v(p) is concave in p , the profit-maximizing price, p̂ , will
obey v (pˆ) D(pˆ) S (pˆ) p D (pˆ) S (pˆ) 0 ; i.e., marginal revenue derived from the
residual demand curve set equal to zero. If this price is compatible with positive
production by the fringe, the market is divided between the two groups of producers.
This is the Stackelberg-outcome. (This outcome is easy to illustrate.)

Within this setting, it was assumed that the dominant firm will announce the price p̂
and the fringe then responds by producing S (pˆ) , known, of course, with perfect
certainty by the cartel. When the model is to be applied for the world market for oil, it
may not be very useful. This point has been elaborated in a paper by Michael Hoel; see
Hoel (1983) in a variant of the model above, but with a different decision sequence: 7
Rather than assuming the sequence above, with the subsequent response behavior by
the fringe, he assumes that the fringe (for some reason, say the time it takes to build up
production capacity) has to decide on the level of output before the dominant firm sets
the price.8 Instead of taking the stipulated or announced price as given and then decide
on output, the fringe now has to guess on a price, call this price conjecture for p * , and
makes a supply decision, according to S (p * ) . Given this predetermined output level, the
cartel will choose a price, so as to maximize w(p; p * ) : p D(p) S (p * ) , with S (p * ) now

being given. (The predetermined output of the fringe works like a uniform negative

7 Michael Hoel (1983), Monopoly Resource Extractions under the Presence of Predetermined Substitute
Production, Journal of Economic Theory, 30 (1), pp. 201 – 212.
8 This might be realistic when it comes to whether we should start drilling outside Lofoten, or as

remarked in an unpublished note by Michael Hoel, when the producers of shale oil or oil sand decide to
invest.
 5

shift in demand facing the dominant firm.) Let the solution to this problem be given by
the price p ** obeying wp (p **; p * ) D(p ** ) S (p * ) p **D (p ** ) 0 . This price choice will

depend on the fringe’s guess, and with p ** pˆ . We also have v(pˆ) w(p**; p * ) ; see
below. In the leader-follower case the dominant firm will have a wider opportunity set
than if the fringe precommits to a given output.

Now, we assume that all players are fully rational in the sense that any player can “see
through the entire game”, and each player can put herself in the other players’ shoes
and calculate what one’s best strategy against all the others’ strategies will be. Then
what each player expects the other players will choose, will in fact be realized or
confirmed in equilibrium. (If we have reached equilibrium, no player will regret.) The
equilibrium can therefore be regarded as a Nash-Cournot equilibrium.

Let the fringe have rational expectations (here in the sense of perfect foresight because
all features of the game are known by all players), with p * p ** . In that case the
optimal price set by the cartel, given the predetermined output level by the fringe, or
given the fringe’s strategy, will obey the first-order condition w p (p ** ; p ** ) 0 , with

p ** pˆ . To see this, note that we can consider

v (p** ) D(p** ) S (p ** ) p **(D (p ** ) S (p ** )) w p (p **; p ** ) p **S (p ** ) p **S (p ** ) 0
because w p (p ** ; p ** ) 0 . Hence, as v(p) is concave, by assumption, we have p ** pˆ .
The dominant firm will set a higher price when the fringe commits to a given output
than what is the case when the fringe responds to an announced price. (Such
commitment will in this model relax the competitive pressure.)

To see what dynamic inconsistency means we can now use a very elegant line of
reasoning provided by Hoel (1983). Suppose that the dominant firm makes a
preliminary price announcement before the fringe makes her output decision. Let this
announcement be p̂ . If the fringe (naïvely) believes that this will be the final price, her
output will be S (pˆ) . Also, if the dominant firm should stick to this price, his profit will
be v(pˆ) . Hence it seems that by announcing the price p̂ , the dominant firm will be able
to avoid the less profitable outcome w(p **; p ** ) . But now we come to the clue: Despite
this option, the dominant firm can make it even better! The argument above was based on
naïve beliefs by the fringe and also irrational behavior by the dominant firm. Given the
output set by the fringe, S (pˆ) , we have seen above that the price being the solution to
 6

the problem, arg max p p D(p) S (pˆ) , will be different from p̂ !9 The preliminary
announcement p̂ is therefore not dynamically consistent. The reason is that if the fringe
should believe in this price announcement, the dominant firm will, according to what
we have demonstrated above, benefit from deviate from that announcement. And
because we have no higher legal authority enforcing binding contracts, a fringe
consisting of rational players will correctly anticipate or predict that that the dominant
firm will deviate from any price announcement different from p ** . Whatever price
announcement should be made, the fringe will correctly predict the price to be p ** and
decide its output equal to S (p ** ) . The dominant firm cannot then do better than by
setting p p ** .

b) A dominant firm – competitive fringe model: The dynamic case10

Consider now extraction over time, and let us look for a dynamic equilibrium, when the
cartel and the fringe operate simultaneously, for some time period. After the fringe has
exhausted its reserve, the cartel operates as a monopoly from then on until exhaustion
when the terminal price hits the price ceiling, given by the cost of supplying,
competitively, a perfect substitute – a “backstop” technology. In order to see what
might happen in the mixed market structure we use what we have learnt about
equilibrium extraction both under perfect competition and pure monopoly, when we
make use of Hotelling’s “No-Arbitrage Principle” to derive conditions for a dynamic
equilibrium as balancing the cost and benefit of delaying extraction so that no producer
has any incentive to alter its intertemporal behaviour. Extraction costs are ignored, and
it is assumed that the cartel’s reserve is larger than the reserves of the fringe producers.
Also, there is a perfect substitute supplied competitively at a price b (a backstop
technology which puts an upper limit on the price of the resource). Demand is time-
invariant, and the demand elasticity is higher in absolute value the higher is the price
(or lower is output sold in the market). All market participants have the same discount
rate r; say the rate of interest in a perfect international capital market.

Both the dominant firm and each fringe producer maximize their present discounted
value of future profits. The dominant firm, taking the fringe’s supply response as given,

9The “argmax” means the value of the argument p that maximizes the function.
An important reference is Stephen W. Salant, 1976, Exhaustible Resources and Industrial Structure: A
10

Nash-Cournot Approach to the World Oil Market, Journal of Political Economy, 84 (5), pp. 1079-1094.
 7

is maximizing with respect to the residual demand. The price path, p(t ) , set by the
dominant firm initially when taking into account the behavior of the fringe producers
who take the price path as given, while the price clears the market at any point in time.
What are the characteristic features of the (open-loop) Nash equilibrium price path in
this mixed market context?

First, as long as there is some extraction the price cannot exceed b (a higher price will
give zero demand). Secondly, because the market participants are rational, the
equilibrium price path must be continuous, showing no jumps, as we have explained
above. Third, with potential speculators operating, the rate of increase in the price
(equal to the rent) cannot exceed r . (If not, speculators would buy up as much as
possible to capture a capital gain.)
As long as the fringe has some reserves left, the equilibrium price path must obey the
p(t )
Hotelling Rule r . If it increases at a rate below r at some point in time t, the
p(t )
fringe would extract and sell the entire stock so as to make a higher profit by
purchasing assets, with an infinite resource supply – with excess supply, forcing the
price down to zero.
During some initial period, with price below b, both the fringe and the cartel will
supply the market, with the price increasing at a rate equal to r, but also the cartel’s
marginal revenue, derived from the residual demand curve, increasing at the same rate.
As long as the fringe is adjusting its supply as the price is increasing with a smaller and
smaller market demand, it is possible to have both marginal revenue and the price
increasing at the same rate as long as the fringe has some reserves left.11 At some point
in time the fringe will have exhausted its stock. From then on, the cartel will be the sole
supplier and will act as a monopolist for the period remaining until price being equal to
b , with equilibrium (no-arbitrage) characterized by the intertemporal monopoly
m p p
equilibrium condition r r , where marginal revenue is m p,
m p p
1
with : 1 , and where is the absolute value of the demand elasticity, with 1

by assumption. It is assumed that (t ) is increasing as price is increasing over time.

11 See Salant (op.cit.) for details.

 8

Hence, the dynamic equilibrium is made up of two phases: The first one, with both the
cartel and the fringe supplying the market, where both the equilibrium price and the
cartel’s marginal revenue (derived from the residual demand) will increase at a rate
equal to the rate of interest. From the point in time when the fringe’s reserves are
depleted, we have the intertemporal monopoly equilibrium, as analyzed in Stiglitz
(op.cit.).

Compared to the optimal solution (competitive equilibrium under complete set of

forward markets), the resources are exhausted once price equals b, with the initial price
TC
C rt C
set so that p(t ) p (0)e , p(T ) b and x C (t )dt S , where market demand is
0

derived from U (x ) p and S the total initial reserves.

Under pure monopoly, we know that price is increasing at a lower rate; hence for the
same amount S of resources to be extracted, the initial price pM (0) pC (0) , and with
pM (T M ) b for T M TC .
In the dynamic dominant firm – competitive fringe equilibrium, there is competition in
between the two extremes above, with an initial price pCF (0) pC (0), p M (0) , as the

fringe exhausts its stock earlier than the dominant firm, and we have the monopoly
equilibrium path in the last phase. The corresponding monopoly price will hit the
backstop price at T CF T C ,T M .

An interesting feature is that if a group of price takers should form a cartel and then act
like a dominant firm, with some outsiders –acting like the fringe above – then according
to the findings above, the resulting equilibrium will not only benefit the cartel members,
but also benefit the (free-riding) fringe producers.

c) Oligopolistic extraction over time

We will just sketch the main principles for a dynamic duopolistic equilibrium for a
resource market. Each resource owner knows that how much I want to sell will affect
the payoff or profit accruing to the other player, through the market price as given by
p(X ) p(x1 x 2 ) , declining in X , where x i is extraction of player no. i , with i 1,2 .
The game is a non-cooperative game, with rational players and a rational expectations
Nash equilibrium must rely on being self-enforcing.
 9

Suppose that resource owner #i has a known resource deposit initially as given by S 0i ,
and ignore extraction costs. Both producers have the same discount rate r , and an
infinite planning horizon. All relevant features of the model are known to both
producers. We then know that the resource constraint can be expressed as
t

Si (t ) Si 0 x i (s )ds , as the remaining reserve of resource stock i at some point in

0

time t .

As seen from ex ante the resource owner i’s objective function; PDV of future profits, is

rt
given by e p(x i (t ) x j (t )) x i (t )dt which is to be maximized subject to
0

Si (t ) x i (t ) Si (0) Si 0 and limt Si (t ) 0.

Each player will now be able to influence the price, but no producer has full control of it
as in the monopoly case. Here we assume that each producer is able to solve the entire
game ex ante (no cooperation), as if each resource owner is able to see through the
entire game during the entire horizon. This is equivalent to saying that at the ex ante
stage, each resource owner announces her entire plan on 0, . What I do is the best
response to what other player is doing, in the same way that what the other is doing is
the best response to what I do. Mutual best responses; my expectations are confirmed ex
post.
Technically this means that each player can solve for all the conditions characterizing
the dynamic (open-loop) Nash equilibrium.
The CV Hamiltonian for player i is then:

H i (xi , Si , x j , i , t ) p(x i x j )x i x
i i

Any rational player can now solve all the conditions that must hold in a dynamic
equilibrium, with X 0 x10 x 20 and price p(X 0 ) :

Hi
p(X O ) x iO p (X O ) i
0 ( 0 if x i0 0) or mi (x iO (t ), x j0 ) i
for
xi
i, j 1,2; i j.
 10

where mi is resource owner i’s marginal revenue, and i

the current shadow value of
her resource. Suppose that both producers are active so that mi (x iO (t ), x j0 ) i
0 . Each
shadow price or costate variable must obey

Hi
i
(t ) r i (t ) r i (t ) i
(t ) i
(0)e rt
Si

(Marginal cost of delaying is equal to the marginal benefit from delaying extraction.)

Hence we get: For any point in time, the equilibrium output sold at the market at some
point in time t, must obey:

m1(x10 (t ), x 20 (t )) 1
(0)e rt

m2 (x10 (t ), x 20 (t )) 2
(0)ert

where the initial shadow values are determined so that the resource constraint is
binding as time goes to infinity. For each resource owner marginal revenue is equal to
the current shadow value of the remaining resource deposit, and we have that

mi x i0
X0
r and mi p(X 0 ) 1 0 0
p (X 0 ) : p(X 0 ) 1 i
0
: i
p(X 0 ) .
mi X p(X ) (X )

Here we have, as before, as the absolute value of the market demand elasticity, and
x i0
i
: is producer i ‘s market share, and with the ratio between marginal revenue
X0
mi p
and price, i
, depending on the player herself. Note that we have i
r . If
mi p i

there are cases where i

(t ) 0 , as in the case with identical resource owners S10 S20
and iso-elastic demand, then the dynamic duopoly equilibrium will coincide with the
dynamic competitive equilibrium.

Consider the case where resource owner 1 has a larger initial reserve than resource
owner 2; S10 S20 . We also have a choke price, b, above which all demand will vanish.
Then we know that 1
(0) 2
(0) , and from the two equilibrium conditions, with
inequality if no extraction, we have:
 11

p(X ) x1p (X ) m1(x1(t), x 2 (t)) 1

(0)e rt
p(X ) x 2 p (X ) m2 (x1(t), x 2 (t)) 2
(0)e rt

If both resource owners are active we must have – on taking the difference between the
two above:

(*) p (X ) x 2 x1 ert 1
(0) 2
(0)

Because 1
(0) 2
(0) , due to S10 S20 and no extraction costs, it follows from (*) that
x 2 (t ) x1(t ) , saying that the resource-rich owner extracts most as long as both are
extracting.
The next question: Will they stop extracting at the same point in time? Will
x1(T ) x 2 (T ) 0 and S1(T ) S2 (T ) 0?
Suppose that p (0) is finite, and suppose x1(T ) x 2 (T ) 0 . Then, with inequality in (*)
we have: p (X ) x 2 x1 0 erT 1
(0) 2
(0) , a contradiction, and we cannot have
x1 x2 0 for the same T . The two deposits will be depleted at different points in
time; with T1 T2, where x1(T1 ) 0 with S1(T1 ) 0 and x 2 (T2 ) 0 as well as
S2 (T2 ) 0 . Why is T1 T2 ?
This can be seen from the following line of reasoning: For x i 0 , we have mi p(X ) .
As one deposit is gradually being depleted, we are moving up along the marginal
revenue curve for that producer, and once the stock is exhausted, the marginal revenue
has reached the price. Because 2 1
, we must have m2 m1 , and resource owner 2
will reach the price path first; hence T2 T1 . Also, because of the choke price,
equilibrium will then require that resource owner 1 will have depleted her reserves at
T1 , when the price has reached b ; i.e. p(x1(T1 )) b.