Professional Documents
Culture Documents
Very preliminary
Espen R. Moen and Christian Riis
Norwegian School of Management
April, 2018
Contents
1 Introduction 4
3 E¢ ciency 8
3.1 Producer surplus . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Consumer surplus . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Integral formulation . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.1 Example 1: the linear demand curve . . . . . . . . . . 17
3.3.2 Example 2: the iso elastic demand curve . . . . . . . . 18
3.4 Generalizations to more than one market . . . . . . . . . . . . 19
3.5 First best allocation . . . . . . . . . . . . . . . . . . . . . . . 20
3.6 Bargaining and the Coase theorem . . . . . . . . . . . . . . . 21
3.6.1 Exercises: . . . . . . . . . . . . . . . . . . . . . . . . . 23
1
5 Investments in quality 36
5.1 Monopoly and investments in quality . . . . . . . . . . . . . . 37
5.2 Socially optimal quality . . . . . . . . . . . . . . . . . . . . . 39
5.3 Comparing the market solution and the e¢ cient solution . . . 40
5.4 "Constrained e¢ cient" quality level . . . . . . . . . . . . . . . 42
5.5 Quality choice in a vertical chain . . . . . . . . . . . . . . . . 43
5.5.1 The case of competitive retailing: No RPM . . . . . . . 45
5.5.2 The case of competitive retailing: RPM . . . . . . . . . 46
7 Ramsey pricing 68
7.1 Formulation of the Ramsey problem . . . . . . . . . . . . . . . 68
7.2 Ramsey vs monopoly prices . . . . . . . . . . . . . . . . . . . 71
7.3 Implications for regulation . . . . . . . . . . . . . . . . . . . . 72
7.4 Optimal regulation - "price cap" . . . . . . . . . . . . . . . . . 72
7.4.1 Exercises Ramsey pricing . . . . . . . . . . . . . . . . . 73
8 Access pricing 79
2
8.3.3 Entry barriers, raising rivals costs . . . . . . . . . . . . 91
8.3.4 Contracts and externalities . . . . . . . . . . . . . . . . 92
8.4 Access price regulation . . . . . . . . . . . . . . . . . . . . . . 94
8.4.1 "Price-minus" principle . . . . . . . . . . . . . . . . . . 95
8.4.2 Partly crowding out . . . . . . . . . . . . . . . . . . . . 96
8.5 Access price regulation based on average margin . . . . . . . . 98
8.6 Competition between integrated networks . . . . . . . . . . . 99
12 Appendix A 122
12.1 Dependence in demand . . . . . . . . . . . . . . . . . . . . . . 122
3
The compendium is based on the slides used in the course Industrial Orga-
nization and Economics Consulting. It is organized according to sequence of
topics as presented in the course, and is de…nitely not meant to be a "text
book". It focuses primarily on the theory part, and the aim is to explain
more deeply the underlying principles of the models used in the course.
The course’s text book is Pepall et.al. "Industrial Organization", 5. edi-
tion. The text book is rich in examples and cases, however there are weak-
nesses when it comes the formal analysis of the models (examples will be
given).
1 Introduction
Industrial organization analyses the boundary of …rms and the relationship
between …rms and markets. industrial organization constitutes the theoret-
ical underpinning of economics consulting. The purpose of this course is
twofold: First to give the students a sound theoretical understanding of the
…eld of industrial economics. Second to demonstrate how the principals of
industrial organization are applied within economics consulting, through real
world cases.
The cases discussed include:
Port pricing - the appeal court’s verdict in the Hurtigruten - Bodø havn
case
The cases are often real cases from economics consulting. However, the
way we are presenting the cases will typically deviate from the presentations
in consulting reports or for a court of law. The cases are usually complex;
many players are involved, the environment is unstructured, and there are
many mechanisms at play simultaneously. In this text we do not present the
4
cases in detail, that will be too time consuming and the context may be very
speci…c to the case at hand and hence give little general knowledge. Instead
we will go deeper into some of the aspects (typically the main aspect) of the
case. We will do this by deploying economic theory, at a deeper and a more
technical level than typically done in the consulting reports. Our aim is to
show how economic forces are at play in the cases, and how we as economists
can capture these forces in mathematical structures and thereby understand
each of them at a deep level. Our aim in this text is not to be "lawyers", i.e.,
not to give an as comprehensive and exhaustive list of relevant factors in the
case and optimally weigh them together to get the de…nite answer. That is
something better learned at the work place.
We want to point out that the models we construct are not ment to be
general models that capture all relevant aspects of the case. On the contrary,
a model is often tailormade to capture the logic of one particular mechanism.
The underlying assumptions of the model may be violated in other parts of
the case, in which other mechanisms are at play. Sometimes it is possible
and even warranted to integrate several mechanisms in one model, but this
is not the general rule, and may often be too time consuming. However,
we do believe that models are vital for economics consulting, particularly in
complicated matters like antitrust and vertical restraints. History here shows
as we see it (slightly sexist) that "a man with a model against a man without
a model is like a man with a gun against a man without a gun". The man
with the gun usually wins.
In the analysis, the acting agents are typically …rms. The consumers play
a more passive role, as price takers who chose whichever option is the best for
them. We assume that the …rms have as objective to maximize pro…ts. For
instance, when we analyse Telenor’s (or other …rms’) incentives to outsource
part of the production, we assume that they will do so if that maximize their
pro…t, without regard to the employment prospects of their employees. Fur-
thermore, we assume that the …rms respond rationally to the situation at
hand. This is in accordance with the standards of our legal system, where
it is typically assumed that professional, comersial agents are able to pur-
sue and protect their own interests. Note howerver, that the assumption of
rationality may not be as adequate in other settings, for instance in issues
related to the marketing control act, in which private persons are granted
special protection.
The cases represents quite di¤erent market structures, but have two com-
mon features: 1) one or a few …rms have dominant positions, and 2) a core
5
element of the political debate regards the vertical structure in the market.
The dominance problem is well known from introductory courses in mi-
croeconomics, and is discussed under the heading "monopoly pricing". We
know that a …rm with market power has incentives to raise its prices above
marginal cost, which creates an e¢ ciency loss (deadweight loss of monopoly
pricing).
In the course we want to take a broader perspective: in particular, are
there other potential distortions associated with dominance? E.g. has a
dominant …rm incentives to choose the optimal product quality, or is it a risk
at it uses its market control to save costs by reducing quality (after all the
customers have few alternatives)? The monopolists price level is above the
optimal level, but what about the price structure? Is there is a risk that a
monopoly conducts harmful price discrimination, by treating some customers
di¤erently from others?
In the long run, are monopolists more or less innovative than …rms com-
peting …ercely? As John Hicks, the British Nobel-prize winner once said:
"The best of all monopoly pro…ts is a quiet life".
Finally, a dominant …rm may face competition from smaller competitors,
and from future potential entrants. Does a dominant …rm have incentives to
foreclose competitors from the market, given that it has the means?
To provide answers to these questions, we have to understand the incen-
tive structure in the market. And this will be the key part of any assessment
of the various policy proposals that come up in the political debate, in the dis-
cussions about how the government and the competition authorities should
deal with the lack of competition. A recent proposal from Arbeiderpartiet, a
proposal which also got support from some members of Høyre, regards price
setting in the grocery market. The concern is that large food suppliers as
Orkla, Tine, Ringnes etc., provide the dominant grocery chains with more
favorable contract terms than smaller stores. According to the proposal, the
grocery market will function more e¢ ciently if the dominant suppliers were
not allowed to price discriminate between downstream grocery stores. Maybe
the proposal will work, maybe it will not. But to say something about the ef-
fect of the proposal, and whether it is the right medicine (given that that the
competition problem is real) we have to analyze the competitive structure,
and understand the underlying mechanisms driving price setting.
The second important feature of the markets analyzed in the course is the
vertical structure. It is common to think about vertical structures as value
chains - as a sequence of processing activities each activity adding value to the
6
product. How does market dominance impact value creation in the vertical
chain? Say there is a bottleneck somewhere in the chain, a dominant …rm
large enough to dictate the terms of trade: does it matter if the bottleneck
is upstream or downstream? Is it harmful to consumers if a dominant …rm
upstream merges with downstream …rms, and leverage its market dominance
to a new level? Take the Norwegian book market as example, large publishing
houses have over time bought up independent book stores and established
large dominating book store chains owned by the publishing houses. Is this
good or bad news for the readers?
Similar patterns are found in the telecommunication market. Both Te-
lenor and Telia are now integrated with some of the previously independent
service providers, like Chess and OneCall. How will the integration process
impact competition in the mobile market? Would it be better if these service
providers were not owned by the integrated networks?
The vertical structure of markets are mostly ignored in introductory mi-
croeconomics courses. However, the following cannot be stressed to much:
the e¤ects of collaboration and integration vertically (between upstream and
downstream …rms) have substantially di¤erent e¤ects from collaboration hor-
izontally (between competitors). It is a basic insights that …erce competition
horizontally is bene…cial to consumers, and that any collaboration on prices
is harmful – and illegal, according to section 10 of the Norwegian Compe-
tition Act. In contrast, coordination on prices vertically, between upstream
and downstream …rms, are often bene…cial to society. This is key in under-
standing the e¤ect of regulation and competition policy associated with the
complex vertical structure we …nd in the sectors discussed in the course.
This text is organized as follows. I …rst provide a theoretical introduction
to main concepts and models. Then I discuss some om the cases more in
detail.
7
gards unlawful cooperation between …rms, covered by section 10 in the act.
The classic example of and unlawful cooperation is collusion, secret aggre-
ment between …rms to limit competition and thus keep prices at a high level.
The second regards abuse of a dominant position, which is section 11 of the
act. A …rm is considered dominant if its market share exceeds 40-50 percent,
and is expected to have the power to act independently of its competitors.
The …nal area is control of mergers and acquisitions, section 16. According
to the law any merger or aquisition involving …rms with a turnover above a
threshold valued must be approved by the competition authorities.
Enforcement of the competition act is often referred to as ex post com-
petition policy. The competition authorities intervenes if …rms have acted
unlawfully. Note that it is not illegal to have market power or being a domi-
nant …rm. It is the abuse of market power which is illegal. If the competition
authorities concludes that one or more …rms have acted unlawfully, the au-
thority has the power to impose a …ne, or it may take the case to the court.
Market regulations, as sector regulation in telecommunication and energy
markets, are founded on a di¤erent principle. It is referred as ex ante policy.
Market regulation has been particularly important within the telecommu-
nications sector where the Norwegian Communications Athority (nasjonal
kommunikasjonsmyndighet, NKOM) is the regulator. NKOM may intervene
and restrict business behavior if they …nd a …rm has the power to abuse
market power, independent of whether an abusement actually has occured.
This contrast to the competition authority, which intervenesl if and only if
if …nds that an abusement has taken place.
Regulatory authorities can also act ex post, and as the competition au-
thority, has the authority to impose …nes if a …rm abuses its market power.
The telecommunication sector is regulated under the The Electronic Com-
munications Act (Ekomloven) from 2003.
3 E¢ ciency
In any policy analyses, you need a benchmark, a measure of e¢ ciency -
otherwise there is no way you can provide a meaningful assessment of policy
proposals. The standard measure of consumer welfare in competition analysis
is total consumer surplus net of total production production costs. Note that
this measure is neutral with respect to income distribution, it just adds up
consumers’willingnesses to pay for the various goods supplied in the market.
8
Total social surplus is de…ned as the sum of consumer and producer sur-
plus. As mentioned in the previous section, the purpose of the competiton
act is to promote e¢ ciency with special regard to the bene…t of consumers.
As e¢ ciency refers to the sum of consumer and producer surplus, wheras
consumer bene…t refers only the former, the formulation of the aim of the
act has given rise to controversies about its interpretation, and how it should
be implemented in practice.
9
product price, and w as the input price, the …rm’s surplus is:
= pq wv
= pf (v) wv
pf 0 (v) = w
which de…nes the …rm’s input demand as function of input price w (for a
given p),
v (w)
v (w) inserted in the pro…t function yields
d (w) dv (w)
= (pf 0 (v (w)) w) v (w)
dw dw
= v (w)
which is the area between the input demand curve and the input price, as
illustrated in the following …gure.
Figure
10
Exercise: Suppose f (q) = q 1=2 . Derive the input demand function, and
show that the …rm’s surplus can be expressed as the area under the input
demand curve.
We can illustrate the producer surplus from the …rm’s cost function. In-
verting the product function f (v) yields v(q), and then the cost function
c(q) = qv(q)
(q) = pq c(q)
Maximising with respect to q yields the well known …rst order condition
p c0 (q) = 0;
price equal to marginal cost. The …gure below gives a graphical representa-
tion,
11
where producer surplus is the shaded area.
Note that the shaded area in this …gure corresponds exactly in size to the
the shaded area in the previous …gure. However, deriving producer surplus
from the input side has the advantage that it does not depend on the …rm’s
degree of market power. To illustrate this, consider a monopoly …rm. With a
monopoly …rm the …rs order condition is represented by marginal cost equal
to marginal revenue, illustrated in the left part of the …gure below.
If we derive producer surplus from the input side, the procedure is not
impacted by the …rm’s strategic position in the market, the area below the
input demand function represents the surplus independent of whether the
…rm is a monopoly …rm or a competitive …rm. The left part of the …gure
is the classic representation of producer surplus with market power. Note
12
that to provide a precise representation of pro…t, information on the demand
elastisity is needed.
We generalize the result in two directions. Consider a multi product …rm,
which is a common characteristics of larger …rms in regulated sectors, e.g.
telecommunication …rms. Second we allow the …rm to have market power in
some or all submarkets.
With vector notation, the …rm’s net revenue is
(q; v) = p(q)q wv
where p(q) is the product price vector and w the input price vector. The
…rm maximizes (q; v) subject to the production possibility set, the feasible
combinations of inputs v and outputs q.1
Denote by q (w) and v (w) the optimal input/output choices given input
prices w. Hence the …rm’s pro…t is
(w) = p(q )q wv
(w1 ; w2 ) = p(q )q w1 v1 w2 v2
which corresponds to the result in the simple one input, one output case with
a competitive.
1
Mathmatically, the production possibility set can be represented by a relation
H(q; v) = 0, which for each set of input levels v describes the feasible combinations of
outputs q. The technical part is not important for the argument in the text.
13
Hence the impact of an increase in the input price on producer surplus
does not depend on whether the …rm is competitive or has market power
in the product market. Furthermore, the production structure can complex.
Note that with a multiproduct …rm (1; w2 ) is not nesecarily zero, repre-
senting the …rm’s surplus from production which does not use input 1, or can
substitute away from this particular input.
This representation of producer surplus simpli…es the analysis of the im-
pact of changes in input prices on …rm’s pro…t.
14
and T is a …xed fee. If T is high, the consumer has less money to spend
on buying the units, and this may in‡uence her demand. However, in most
cases the expences associated with buying a good is small relative to the
person’s total income, and the income e¤ects therefore of little importance.
In such circumstances the willingness to pay as we de…ne it below will be a
good aproximation for the value of the good to the consumer.
The standard (Marshallian) measure of gross consumer surplus is the
area under the demand curve. Suppose q(p) is a downward sloping market
demand curve. The curve shows how much consumers wish to buy at any
given price p. Let p(q) represent the inverse demand curve, the mirror image
of q(p), and drawn in the …gure below.
Gross consumer surplus is the area under the demand curve is, the shaded
area in the …gure. Analytically, gross consumer surplus is the integral
Z q
p(z)dz
0
Net consumer surplus, or simply "the consumer surplus", is obtained by
subtracting consumer expenses, pq, from the gross measure. Let CS be the
consumer surplus2 , then
Z q
CS = p(z)dz pq
0
2
It can be shown that consumer surplus, de…ned as above, is a precise meausure of
consumers’valuation if there are no income e¤ects.
15
The integral formulation of demand curve is convenient as it allows us to
deal with di¤erent speci…cations of the demand function. A brief introduction
is given here:
Suppose we increase q with one unit. What will be the impact on consumer
surplus? As illustrated in the …gure below, the additional unit is valued at
p(q), which is the marginal valuation at the level of supply q. Hence we
should expect that the derivative of the integral with respect to q equals
p(q), which actually is true.
Rq
Mathematically, we can di¤erentiate 0 p(z)dz with respect to q which
yields excactly p(q): Rq
d 0 p(z)dz
= p(q)
dq
In some cases it is possible to solve the integral explicitly. Two such cases
are the linear curve and the iso-elastic demand curve. Let us do that:
16
3.3.1 Example 1: the linear demand curve
Let the demand curve be
p(q) = D q
where D is a …xed number. Gross consumer surplus is:
Z q Z q
p(z)dz = (D z)dz
0 0
We solve the integral:
Z q
1 2 1 2
(D z)dz = jDz z = Dq q
0 2 2
How is gross consumer surplus impacted by q, total consumption? We take
the derivative
d Dq 21 q 2
=D q
dq
which is exactly the same as we obtain by di¤erentiating the integral in the
…rst hand, recall that Rq
d 0 (D z)dz
=D q
dq
Let us split gross consumer surplus in net surplus and consumer expenses:
1 2
Dq q
2
1
= D q+q q q
2
1
= (D q) q + q q q
2
1
= (D q) q + q 2
2
1 2
= pq + q
2
where 21 q 2 is net consumer surplus, and pq are the consumer expenses, see
the …gure below.
17
3.3.2 Example 2: the iso elastic demand curve
The iso elastic demand curve has a constant demand elasticity. The demand
elasticity is de…ned as
dq p
" = Elp q(p) =
dp q
If we specify the demand function as follows
"
q=p
0 "
+ 1 "
+1 " 1
18
and we can decompose gross consumer surplus into the sum of net consumer
surplus and consumer expenses:
" " 1
q "
" 1
" " 1
= q " pq + pq
" 1
" " 1 1
= q " q " q + pq
" 1
" " 1 " 1
= q " q " + pq
" 1
" " 1
= 1 q " + pq
" 1
1 " 1
= q " + pq
" 1
" 1
where " 1 1 q " is net consumer surplus, and pq are the consumer expenses.
Note also that the derivative of gross consumer surplus with respect to q
is p(q). We …nd:
" 1
"
d " 1
q " " 1 1
1
=q " =q " = p(q)
dq
19
3.5 First best allocation
Given our measure of social welfare, we can derive the conditions for an e¢ -
cient allocation. Considering one market segment in isolation, social welfare
is maximized at the level of production which solves the following maximiza-
tion problem: Z q
max p(z)dz c(q)
q 0
Di¤erentiate with respect to q yields the …rst order condition
p(q) c0 (q) = 0
which says that price should be equal to marginal cost, a result known from
introductory micro courses. This result is straight forward to generalize to
several market segments, e¢ ciency requires that price equals marginal cost
in each single segment. This is trivial with independent demand, in which
case …rst best is de…ned as
Z q1 Z q2
max p1 (z)dz + p2 (z)dz c1 (q1 ) c2 (q2 )
q1 ;q2 0 0
20
3.6 Bargaining and the Coase theorem
The Coase theorem states that if agents can approach and negotiate and
there are su¢ ciently low transaction costs, bargaining will lead to an e¢ cient
outcome regardless of the initial allocation of property rights.
We will here present a simple example of the principle, and we will use the
vertical value chain as example. Suppose two …rms, A and B, can approach
each other and negotiate a contract. A is an upstream …rm which makes
an investment, and B is a downstream …rm which realizes the value of the
investment in the market.
If they make a deal, …rm A invests an amount q which creates value V (q)
at cost C(q). A pays the investment cost C(q) whereas …rm B realizes the
value V (q) in the market. A lumpsum transfer T from B to A splits the gains
between the two parties.
If they sign a contract, the relationship creates net value
V (q) C(q):
If the they do not agree, and split up, we assume that A obtains uA as its
best alternative and B obtains uB . Suppose the investment q and the transfer
T are contractable.
We can make some observations:
1. The relationship creates value beyond what the two parties can make
on their own if and only if
2. The value of the relationship (and thus the size of the pie for which
the two parties bargain over) is as large as possible if q is set such that
V (q) C(q) is maximized. Refer to q as the investment level which
maximizes value added. It follows that q is determined by the …rst
order condition V 0 (q ) = C 0 (q ).
3. The two parties have a con‡ict of interest in how to split the pie.
Let T be the agreed transfer from B to A (the compensation for A’s
investment cost). Then it is clear that A wants T to be as large as
possible, and B wants T to be as small as possible.
21
4. The two parties have a common interest in making the pie as large
as possible, that is to agree on q . Note that if an ine¢ cient level of
investment q^ is proposed in the negotiation, for any T^, there exists a
counter o¤er q and a transfer T which is preferred by both parties
to the o¤er q^ ,T^. Given q^ it follows that the investment q creates
additional value of size
[V (q ) C(q )] [V (^
q) C(^
q )]
V (q ) C(q ) uA uB > 0
22
In the further discussion, we can think about a market where one part
sets the prices, as a market where this part has all bargaining power. This
part then provides a "take it or leave it o¤er", and the other part accepts or
rejects. For example, if the seller sets the price, the buyer’s only choice is
between "buy it" or let it go.
3.6.1 Exercises:
Consumer welfare
Qu. 1: Just to refresh the math: let gross consumer surplus be
Z q
p(z)dz
0
Solution:
Z q
1 2
(D az) dz = jq0 Dz az
0 2
1 2
aq = Dq
2
Observe the following - a linear demand curve corresponds to a quadratic
utility function.
Solution:
Z q
1 1 1
z " dz = jq0 1 z1 "
0 1 "
1 1 1
= 1q
"
1 "
23
4 Vertical market structure
From now on, the vertical structure of markets will be the main focus. Let me
add one comment to this. Isn’t it so that there is a vertical structure already
build into the standard models, as we know them from introductory courses?
Think about the concept "…rm" as it is introduced in the production theory
part of elementary microeconomics. The …rm purchases primary production
inputs in the market, and uses its production function to transform the inputs
to …nal products. In most "real" cases the transformation from inputs to …nal
products is sequence of re…nement processes, in which each step adds some
value to the product.
In some cases, the re…nement process is entirely an internal process in one
single …rm. Thus represents a vertically integrated value chain - everything
takes place inhouse. Hence the model of competition between …rms, as it is
presented in elementary courses, can be considered as competition between
vertically integrated production units.
The question is, does it matter if the sequence of re…nement processes
is split between companies? That some …rms are upstream, buying primary
inputs, transforming them to semi…nished products, and then sell them to
downstream …rms …nalizing the production process. Will this vertical disin-
tegration in the value chain impact e¢ ciency?
The question belongs to the general question about the boundaries of
…rm - should the …rm do the whole process inhouse, or is it more e¢ cient
to "outsource" some some of the tasks. Part of it has to do with the issue
of technical specialization. It could be that a …rm becomes more technically
e¢ cient if it specializes in a more narrow set of tasks - like Raufoss indus-
tripark producing bumpers for Volvo. But more importantly, the vertical
structure impacts the incentive structure, which will be the focus here.
In the discussion we explore alternative speci…cations of the decision
processes in the value chain. Take the relationship between an upstream
and a downstream …rm. As independent …rms, you can think about them
simply as price setters, where each …rm independently set their respective
prices. Alternatively, we can think about prices as the outcome of a bargain-
ing process. Furthermore, the contract governing the transfers between the
…rms can be more or less complex, depending on details in the market. In
the discussion we explore di¤erent alternative descriptions of these processes.
An essential part of the discussion is the degree of market power. The
simplest case, and a natural starting point, is the monopoly model. The intro-
24
ductory microeconomics case can then be thought of as a vertically integrated
monopoly. Let us now consider a market where there is one monopoly …rm
upstream supplying a downstream monopolist. We begin with the double
marginalization problem:
which has the standard interpretation: the left hand side, marginal revenue,
should equal the right hand side, marginal cost.
A convenient reformulation of the …rst order condition is as follows:
!
q 1
p(q) 1 + p0 (q) = p(q) 1 + 0 p = c0 (q)
p q (p) q
25
hence3
1
p(q) 1 = c0 (q)
"(p)
where "(p) = q 0 (p) pq is the demand elasticity. Sometimes we assume that
the elasticity is constant (iso elastic demand curve), in which case " is a …xed
number, independent of p.
The left hand side of (3) is the marginal revenue, and let us refer to that
at as m(q), in which case the …rst order condition can be written
m(q) = c0 (q):
The …rst order condition determines q M , the monopoly output.
We next explore independent price setting.
m(q) = r
with second order condition for maximum
m0 (q) < 0
we see that the standard representation with a downward sloping marginal
revenue curve is su¢ cient for the …rst order condition to represent a solution
to the maximization problem4 .
3
Note that the slope of the inverse function is equal to the inverse of the slope of the
original function. That is, since p(q) is inverse to q(p), we have p0 (q) = 1=q 0 (p).
4
A downward sloping marginal revenue curve is not just as trivial as the assumption of a
downward sloping demand curve. Even with a downward sloping demand curve, marginal
revenue can be non-monotone (if the elasitisity of demand changes in demand level) and
discontinuous (if the demand curve is kinked).
26
Note that m(q) = r is one equation which maps the wholesale price r into
a downstream …rm production level q D (r). Hence it has the interpretation
of the downstream …rm’s demand for deliveries as function of the wholesale
price. That is, m(q) is the demand curve the upstream …rm faces.
Hence the upstream maximizes pro…t
U
= m(q)q c(q)
= p(q)q c(q)
Note that the di¤erence between the two expressions: the upstream …rm faces
a demand curve m(q), whereas the integrated …rm faces demand curve p(q).
We can show that demand curve m(q) yields a higher price than demand
curve p(q). Note that the …rst order condition for the upstream …rm is:
d U
= m(q) + m0 (q)q c0 (q) = 0
dq
whereas for the integrated …rm we have
d
= m(q) c0 (q) = 0
dq
27
The double marginalization problem is severe, not only because joint prof-
its is lower than if the …rms coordinate the price setting, but also because
consumers are worse o¤ than with an integrated monopoly. It is often for-
mulated as: "the only thing which is worse than a monopoly is a chain of
monopolies."
The problem has a simple intuition: the downstream …rm charges a con-
sumer price which is a mark-up on the wholesale price, which already includes
a mark–up. Thus, the downstream …rm sets a "margin on the margin" - cre-
ating a "double margin" - which is bad for producers and bad for consumers.
If double marginalization is a problem, is regulation the proper medicine,
or is it a problem which is better dealt with in the market?
28
con‡ict of interest in how to split the gains from trade - however they have a
common interest in making the pie as large as possible. Hence any contract
o¤er which is ine¢ cient in that it does not realize all gains from trade, is
dominated by an e¢ cient contract o¤er, see the discussion above. The two
parties can simply split the additional value created by the e¢ cient contract.
This argument is related to the more general "Coase theorem" which says
that if there are su¢ ciently low transaction costs, and symmetric information,
bargaining will lead to a Pareto e¢ cient outcome regardless of the initial
allocation of property rights.
Typically, B2B contracts are more complex than simple uniform price
contracts. In most cases, the parties do not bargain over "a price per unit",
but they bargain over the total payment charged for a package of deliveries.
A simple representation of this is a two part tari¤. With a two-part tari¤
the transfer from the downstream …rm to the upstream …rm is
rq + T
where r is the variable part of the tari¤, a payment per unit, and T is a …xed
transfer. Uniform prices is the special case with T = 0.
The nice feature of a two part tari¤ is that the parties have two pricing
instruments to play with, and they have two goals. They can use the variable
part r to maximize value creation, and the …xed transfer T to split the
pie. With only one instrument, it is impossible to distinguish between value
creation and value distribution, which leads to ine¢ ciencies.
More precisely, the integrated monopoly maximizes
= p(q)q c(q)
r = c0 (q M )
D = p(q)q rq T
29
with …rst order condition
m(q) = r
Since r = c0 (q M ) it follows that downstream …rm’s production decision
replicates the integrated monopoly’s:
qD = q M
Another example of a market based contract which deals with the problem
is RPM, resale price maintenance (…xed price contract).
q = qM
m(q) = c0 (q)
30
distortions as it may not re‡ect the heterogeneity in the market - …rms di¤er
in many respects, and the optimal contracts re‡ect these di¤erences.
This is an example of regulatory ine¢ ciencies, that must be taken into
account in that regulation can only be justi…ed if the market distortion is
su¢ ciently severe.
So far we have focused on price as the decision parameter. Firms makes
a lot of other decisions, investment decisions, choice of quality, marketing
decisions, investment in research and development etc. We will now focus on
the set of decisions that impact consumers willingness to pay for a product,
for a given price. We can think about that as the choice of quality, marketing,
sales service etc.
4.4.1 Exercises
Double marginalization:
Derive the marginal revenue for the integrated …rm, and the upstream
…rm, in the following two cases:
m(q) = D 2q
m(q) + m0 (q)q = D 4q
In both cases we see that marginal revenue is smaller for the upstream
…rm (which is a general result).
31
Suppose demand is linear and represented by
p(z) = D z (4)
Qu. 1) Show that marginal revenue equals
mr(z) = D 2z
Solution: Suppose the demand curve is p(z) = D 2z. With this demand
curve, the marginal revenue is
mr(z) = p(z) + p0 (z)z
= D 2z + ( 2) z
= D 4z
The monopoly chooses z such that mr(z) equals marginal cost.
Consider next an upstream and downstream producer signing a linear
price contract. Refer to r as the wholesale price. The downstream …rm
maximizes
(z) = p(z)z rz
= (D z) z rz
32
with …rst order condition
mr(z) r
= D 2z r = 0
The upstream …rms sets r taking into account that downstream production is
determined by the …rst order condition D 2z r = 0. Hence the upstream
…rm solves the following problem:
max rz cz
r
subject to
D 2z r=0
Since the constraint represents a one to one mapping from r to z, we can
reformulate the optimization problem as an maximization problem in z. From
the constraint, solve for r, and then substitute out r in the objective function:
rz cz = (D 2z) z cz
with this formulation, we see that the upstream …rm faces a maximization
problem which corresponds exactly to the maximization problem of a hypo-
thetical monopolist facing demand curve D 2z.
rz + T
the upstream …rm. Given a contract [r; T ] the downstream …rm maximizes
D (z) = p(z)z rz T
= (D z) z rz T
mr(z) = r
D 2z = r
33
The upstream …rm’s pro…t is:
(r c)z + T
With a two part tari¤, the upstream …rm faces two constraints. The …rst
is the incentive constraint, represented by the downstream …rm’s …rst order
condition D 2z = r. The second is the participation constraint, in that
the downstream …rm must be better o¤ signing the contract than rejecting it
(without this constraint the planner would set T = 1). Let the participation
constraint be a break even constraint,
D (z) = p(z)z rz T
= (D z) z rz T 0:
max (r c) z + T
r;T
s:t:
D 2z = r
(D z) z rz T = 0
(r c) z + T
= (r c) z + (D z) z rz
= (D z) z cz
D 2z = c
34
which we recognize as the …rst order condition for the solution of an inte-
grated monopolist’s problem.
This is a well known result: with two instruments the principal can use
the wholesale price to induce the down stream …rm to maximize total pro…t,
and then extract the pro…t using the …xed term T . Hence the contract terms
will be:
r = c
2
M D c
T = =
2
D c 2
where we recognize 2
as the maximum pro…t an integrated monopolist
can obtain.
Qu. 4) Suppose the two …rms negotiate the terms of the contract [r; T ].
Derive the equilibrium in the following three situations:
The upstream …rm has all bargaining power (can make a "take it or
leave it o¤er")
They have equal bargaining power (split the gain …fty …fty)
Solution:
The …rst bullet point refers to a situation which corresponds to the pre-
vious exercise. Take it or leave it o¤er means that upstream …rm sets r and
T , conditional on the incentive and the participation constraints.
Consider the second and third bullet point. Note that the two parties
have a common interest in maximizing total wealth, to split a large pie is
better than to split a small one. Thus, as long as negotiations are e¢ cient
(they typically are if the parties have access to the same information), the
two parties will both strictly prefer to set r = c, which is exactly what is
needed to make the pie as large as possible. Hence, they negotiate over the
distribution of the pie, that is the …xed term T .
35
If the downstream …rm has all bargaining power it will extract all pro…t,
hence the equilibrium contract will be (assuming that the upstream …rm’s
participation constraint is a break even constraint)
r = c
T = 0
With equal bargaining power, they split the pie …fty …fty, hence:
r = c
2
1 D c
T =
2 2
5 Investments in quality
In this section we focus on choice of quality as a decision problem. We can
think about quality in a broad sense, capturing the more dynamic aspects of
competition.
So far we have analyzed price setting. The choice of price is a static
decision as the …rm takes demand and cost structures as given, and maximizes
pro…t. Over time both demand- and cost-structures are subject to changes.
Demand changes as product quality over time increases, and new technology
generations are introduced. On the cost side, process innovations reduce
production costs over time.
Both product- and process innovations are impacted by …rm investments.
One aspect of this is quality improvements, …rms investing in new and better
product types. In the growth literature these processes are typically explored
within an explicit dynamic framework, however deep insights can be gained
from a more static approach; e.g. a two stage approach, where …rms invest
in the …rst stage, and equilibrium prices and quantities are determined in the
second stage. So although the models we will explore are static, they can be
thought of as simple representations of a dynamic market.
How does the market structure impact the incentives to invest in qual-
ity? Suppose the demand function (consumers’marginal willingness to pay)
depend on quality s, described as follows
p(q; s);
36
where p(q; s) is increasing in s for any given q. As illustrated in the …gure,
improved quality s shifts the demand curve upwards.
Suppose the …rm’s cost function is c(q; s), which is increasing and convex
in both arguments. The …rm has now two choice parameters, production
level q and quality level s, and the …rm maximizes
The …rst order conditions for the monopoly’s choice of q and s are
37
where m(q) is the marginal revenue,
The equations (6) are two conditions that simultaneously determine the
monopoly production and quality, q M ; sM . The …rst equation has a known
interpretation: marginal revenue equal to marginal cost. Consider then the
…rst order condition for choice of quality:
dp(q M ; s) M
q cs (q M ; s) = 0
ds
M
The term dp(qds ;s) has the interpretation of the marginal consumer’s mar-
ginal valuation of quality (evaluated at q M ). In a …gure it can be illustrated
as follows:
A higher quality level shifts the demand curve upwards. Hence the …rm
can charge a higher price without losing any customer. Note the following:
if the …rm increases s it can extract the marginal consumer’s additional
valuation associated with the quality improvement. The fact that it is the
marginal consumer’s valuation that matters has e¢ ciency implications. Let
us compare the following two situations.
38
In the two situations, the marginal return from investments in qualities
are identical from the monopoly’s point of view, hence the …rm’s prospects
of extracting pro…t from quality improvements are exactly the same.
But the two situations are not identical from a welfare point of view.
In the left …gure, inframarginal consumers’ valuations increase a lot from
the investment, in the …gure to the right, very little. As every consumers
valuation of quality matters to the social planner, it follows that the quality
improvement creates much more additional value in case a compared to case
b.
Formally, we show this in the next section.
p(q; z) cq (q; s) = 0
Z q
dp(z; s)
dz cs (q; s) = 0
0 ds
These two conditions simultaneously determine the e¢ cient price and quality
level, p and q .
39
The …rst condition is well known - the planner wants to set the price
equal to marginal cost. The second condition can we write as follows (we
just multiply and divide by q ):
R q dp(z;s)
0 ds
dz
q cs (q ; s) = 0
q
We have the following interpretation
Rq dp(z;s)
dz
0
q
ds
is the average consumer’s marginal valuation of quality
As indicated above, the planner takes the interest of all consumers into
account, hence the e¢ cient level balances marginal quality cost with the
average valuation of the improvement.
Thus there is a source of distortion associated with quality choice, but, as
we will see, it is not clear if the equilibrium quality ends up being to high or to
low from a social perspective. Furthermore, there are some methodological
issues, which the text book suppresses, and which are important for the
interpretation of the results.
40
R qM
than the average consumer’s valuation, 0 dp(z;s)
ds
dz=q M . It follows that the
monopolist preferred quality is much higher that what would have been the
socially optimal level: sM > s . It is even possible to think about this
e¤ect so strong that the monopoly quantity q M exceeds the socially optimal
production level q , see the …gure below.
Note that the higher quality shifts the demand curve outwards, and if
this shift is su¢ ciently large, it is certainly possible to construct a situation
where q M exceeds q . Does this mean that the monopoly produces too much
compared to social e¢ ciency? No that would be an absurd conclusion. Given
its quality level sM , the monopoly …rm sure produces to little, production
in a monopoly is always downward distorted. However, this methodological
problem arises in the text book:
c(q; s) = (c + (s)) q
q
Q(p; s) = s (A p) ! p=A
s
which yields the integrated monopoly’s choice of quality
dp(q M ; s) 0
(s) = 0
ds
41
First best quality satis…es
Rq dp(z;s)
0 ds
dz 0
(s) = 0 (7)
q
Given the speci…cation of demand, we can explicitly solve for s and sM .
It follows that s = sM . The authors conclude:
"..the speci…c result that the integrated monopolist chooses the socially
optimum service level re‡ect the particular demand and cost relationships
that we have assumed, and is not fully general. Nevertheless, the result
is useful because it does show a rather general interest that the manufac-
turer has in providing retail services" (page 468) This is misleading. As we
will see below, the speci…cation of the demand curve in the book represents
preferences for which the marginal consumer valuates quality more than the
average consumer. Hence the quality level ends up being too high from an
e¢ ciency standard.
To make the comparison correct, we have to introduce the concept "con-
strained e¢ ciency".
42
We reach the following conclusions:
If R qM dp(z;s)
0 ds
dz dp(q M ; s)
>
qM ds
the constrained e¢ cient quality level is higher than the equilibrium level.
If R qM dp(z;s)
0 ds
dz dp(q M ; s)
<
qM ds
the constrained e¢ cient quality level is lower than the equilibrium level.
D = (p(q; s) r) q (s)
The …rst order conditions will be
m(q D ) = r
dp(q D ; sD ) D
q = 0 (sD )
ds
We can compare the quality level chosen by the downstream …rm with
the quality level an integrated monopolist would choose. Say the marginal
cost upstream is constant equal to c, we have the following two equations:
m(q M ) = c
dp(q M ; sM ) M
q = 0 (sM )
ds
As r > c (double marginalization) we have q M > q D . Regarding quality
choice, which one is largest depends on whether
43
The downstream …rm’s quality choice may deviate from the hypothetical
integrated monopolist for two reasons. First, due to the double marginal-
ization problem, everything else equal, the integrated monopoly’s produc-
tion, q M , exceeds q D . The second e¤ect is of less signi…cance, the marginal
consumer’s valution of quality may di¤er between production level q M and
production level q D . Note that the quality dimension enhances the negative
e¤ect of double marginalization, due to the interaction between the two in-
struments. Because production is low, due to double marginalization, the
equilbrium quality level tend to be low. Low quality reduces demand, and
hence yields a lower marginal revenue.
Take the following speci…cation of deman as an example:
p(q; s) = D + s q
dp(q; s)
=1
ds
which means that the …rst order condtion for quality is
0
qD = (sD )
U = (r c) q D
44
5.5.1 The case of competitive retailing: No RPM
Chapter 17.4.3 in the textbook introduces competitive retailing. Compet-
itive retailing is there de…ned in the standard way as the situation where
competition between many retailers bring the retail price-cost margin down
to zero. Hence the double marginalization problem disappears.
Competitive retailing, as de…ned in the book, requires that the service cost
function is per unit sale, otherwise the cost function will exhibit increasing
returns to scale5 . Hence in their model, total service cost is
(s)q
45
Exercise: Can you specify a demand function for which the marginal
valuation of quality always equals the average valuation?
Hence to analyze downstream competition in a model with quality choice,
we have to allow for equilibria with di¤erentiation in quality. This brings us
to the next topic, price discrimination and quality choice.
Suppose now that the manufacturer o¤ers the retailers a whole sale price r
as well as a binding retail price p. In this case, competition together with
the zero pro…t constraint of the retail sellers imply that the quality o¤ered
in the market is given by
(s) = p r
It follows that the manufacturer can perfectly govern the quality level of the
retailers by varying the price margin p r . The pro…t to the manufacturer
is given by
m
= q(p(q; s) r c)
= q(p(q; s) (s) c)
This is exactly the same expression as the expression for the pro…t of the
integrated monopolist. It follows that the manufacturer can and will set the
rpm margin p r so as to implement the quality level s that maximizes total
pro…t in the chain, equl to her own pro…ts. Note, however, that in this case,
all the consumers get the same quality level, independently of their indivud-
ual willingness to pay for quality. Hence the quality level is not tailored to
the preferences of di¤erent groups of consumers. In order to achieve such tai-
loring, the manufacturer needs more instruments. For instance, if there are
two groups of customers, and the optimal (pro…t-maximizing) quality level
is s1 when selling to customers of type 1 and s2 when selling to customers of
type 2, this can be obtained by giving the retailers serving market segment
1 a margin of (s1 ) and those serving market segment 2 a margin of (s2 ).
Price and quality discrimination will be discussed in detail the next section.
46
6 Price discrimination and quality choice
Price discrimination generally is a important topic in the literature on reg-
ulation and competitive structure. In many of the cases discussed in the
course, it is claimed that a dominant …rm discriminates in price in favor of
some …rms and at the cost of others. We will come back to other forms of
price discrimination, here I introduce a model of price discrimination of sec-
ond degree, a topic discussed in the text book, and relate this to the choice
of quality.
Price discrimination is often referred to in the policy discussions. A com-
mon claim is that price discrimination is used to foreclose competitors from
the market, and to build up entry barriers, in order to protect own market
power. In contrast, it is a well established result that price discrimination
is important to enhance e¢ ciency in markets with large …xed costs, a cost
structure which is a common characteristics in the cases discussed in the
course.
The IO literature distinguishes between price discrimination of …rst, sec-
ond and third degree. First degree (perfect) price discrimination has limited
empirical relevance. However it points at one important reason why price dis-
crimination may be important for e¢ ciency. If prices can be individualized,
the dominant …rm has an incentive to maximize value creation and extract
all gains from trade associated with each customer7 . Hence …rst order price
discrimination yields e¢ cient allocation, see …gure below
7
You can literally think about …rst degree price discrimination as the monopoly ap-
proching each single consumer, negotiating individiualized contracts. If the the …rm as all
bargaing power, the equilibrium as characterized above occurs. If the single consumer has
some bargaining power, e¢ ciency is still obtained with a di¤erent split of total value (see
the section on bargaining).
47
Price discrimination of second degree is well known from telecommuni-
cation markets and the transport sector, where there is huge variability in
price models and contract terms . Each …rm o¤ers price menus from which
their customers choose.
We will explore second degree price discrimination within a model which
also includes choice of quality. We can then utilize the insights from the
previous section, and also capture some of the e¤ects associated with choice
of price models.
One important conclusions follow from these models: price discrimination
of second degree (as price discrimination in general) has potentially strong
e¢ ciency enhancing e¤ects in infrastructure and network markets (telecom-
munication, air industry etc). In those markets a substantial fraction of
total cost is common and …xed (sunk), hence …rms must charge a mark-up
on marginal cost in order to cover total cost. Price discrimination allows
…rms to expand faster, and to expand into market segments which would not
be pro…table under the restriction of uniform prices.
We conduct the analysis within the simplest possible setup. Let’s begin
with one monopolist.
Assume there are two types of consumers, "rich" and "poor", and each
consumer buys (at most) one unit of the product. To capture the logic from
the previous section (competitive retailing) we assume that the …rm can o¤er
a menu of contracts. The …rm can restrict the o¤er to one contract type,
consisting of a price p and quality s. If if o¤ers a menu, we refer to contract
H as the contract which is targeted consumer type H, and to contract L the
other one.
48
Utility functions
UH (pH ; sH ) = uH (sH ) pH
UL (pL ; sL ) = uL (sL ) pL
(si )
U (s; p) = 2 + ln s p
(s) = s
(p; s) = p s
49
In a diagram we can illustrate the consumer’s indi¤erence curves in a
diagram with p on the vertical axis. With the speci…cation of preferences we
have the slope of the indi¤erence curves:
dp 1
M RS = =
ds s
Note that utility increases for any movement in south east direction, as
indicated by the arrow. The shaded area shows the combinations of price
and quality which provide the consumer with a utility at least as high as the
reservation utility.
In the same diagram we can illustrate the …rm’s iso pro…t curves. As the
pro…t is
(p; s) = p s
we …nd the slope of the isopro…t curve as:
dp @ (p; s)=@s
= =1
ds @ (p; s)=@p
We can illustrate the isopro…t curves in the same diagram as the con-
sumer’s indi¤erence curves
50
I have drawn two iso pro…t curves. Note that pro…t increases in north
west direction.
Assuming that the …rm provides an o¤er, the pro…t maximizing o¤er is
given by p ; s in the …gure. s maximizes the value created by the con-
tract, and by charging p the …rms extracts the entire surplus. Note that
price quality combination is on the highest iso-pro…t curve compatible with
the consumers reservation utility. Formally the …rm solves the constrained
maximization problem:
maxp s
p;s
s:t:
2 + ln s p 0
where the constraint is known as the participation constraint.
We can solve the maximization problem using the Lagrange technique,
though it is simpler to just substitute out p in the pro…t function from the
constraint, assuming that the constraint binds. The latter follows directly
from the observation that the …rm will always raise the price so that the
consumer’s utility equals the reservation utility (extract all rents).
Hence the …rm solves the following problem
max 2 + ln s s
s
2 + ln sM p=2 p=0
hence
pM = 2
What can be said about the quality level? It follows that quality level is
…rst best, the same quality as the social planner would choose. Let’s provide
three arguments for this conclusion.
52
two parties. Say that the two parties, the consumer and the …rm, negotiated
over price and quality. The more bargaining power the consumer has, the
lower will be the price, but they will always agree on the same quality level. If
there were income e¤ects in the consumer’s preferences, the e¢ cient quality
will depend on the parties respective bargaining strengths.
UH (sH ; pH ) = 4 + ln sH pH
and
UL (sL ; pL ) = 2 + ln sL pL
We consider two cases:
Case 1: = 0:5
Case 2: =2
Note that the two types di¤er in two respects. First, the rich consumer
has a higher utility level, and thus a higher willingness to pay. Second, as we
will show, the marginal valuation of quality di¤er. The latter implies that the
marginal consumer’s valuation of quality di¤er from the average consumer’s
valuation, which may lead to distortions.
The …rm’s pro…t function is now
(pL ; pH ; sL ; sH ) = pL sL + p H sH
53
We …nd the rich consumer’s MRS:
1
dUH (sH ; pH ) = 0 = dsH dpH = 0
sH
hence
dpH
=
dsH sH
Obviously, we have two possible situations. Either H’s indi¤erence curves
are steeper than L’s (if > 1) or L’s are steeper than H’s (if < 1), as
illustrated.
Step 2: Consider the contract o¤er [s; p] = [1; 1]. Let A be the set of
contract o¤ers the poor consumer prefers to [s; p] = [1; 1]. Let B be the set
of contract o¤ers the rich consumer prefers to [s; p] = [1; 1]. Illustrate A and
B graphically.
54
Note that everything below the indi¤erence curve passing through [s; p] =
[1; 1] is preferred to [1; 1].
Use these two sets to explain how the …rm can di¤erentiate the contract
o¤ers under the constraint that both agents incentive constraints must be
satis…ed. Hence one contract is preferred by L and the other contract by H.
55
Consider the …gure to the left, in which H’s indi¤erence curves are steepest
( > 1). Suppose the …rm o¤ers two contracts, a and b. Note that H prefers
a to b, whereas L prefers b to a. Hence the contracts yield separation in the
market. Note that unless the slopes of the indi¤erence curves are identical,
it is always possible to specify contract terms such that a separation between
the two types is obtained. Another example is a and c. However, if the …rm
o¤ers contract a and d, separation will not occur, since both types prefer the
d-contract to the a-contract. Hence to obtain separation, you have to …x one
contract (say a) and then pick an alternative contract in either of the two
segments 1 and 2.
The …gure illustrates the two situations. Consider the …gure to the left,
> 1. The contract p ; s is the solution which maximizes the …rm pro…ts
56
if type L is the only type in the market. Recall to two characteristics of the
solution: i) the consumer is on his reservation utility, and ii) the marginal
rate of substitution equals the slope of the iso-pro…t curve. The latter means
that the contract is e¢ cient.
Suppose type H is present i the market. The …rm can choose to continue
o¤ering the contract p ; s , that is have the same contract terms for both
types. However the …rm can do better. Suppose the …rm continues o¤ering
contract p ; s , but in addition o¤ers a contract pH ; sH which it hopes will
be chosen by H. For H to prefer the new o¤er to p ; s it must belong to the
segment 2 (ref previous exercise). Among all contracts in segment 1, the …rm
prefers the one which maximizes pro…t - that is a contract which is on the
highest possible indi¤erence curve. Obviously that is pH ; sH , the tangency
point of the iso-pro…t curve and the indi¤erence curve.
Note that the …rm cannot raise p further, if it did, type H will prefer
p ; s to pH ; sH .
Furthermore, whenever 6= 1, it is possible to gain from market separa-
tion (see the …gure to the right illustrating the situation with < 1).
Step 5: The …nal step is to show that the …rm can do better by modifying
the contract o¤er given to L. You can show the result graphically, but
try also to solve the problem analytically - with the simple speci…cations of
preferences and pro…t functions, it should be possible.
57
The contract indicated by a and b corresponds to the solution discussed in
the previous question. As mentioned, the …rm cannot raise the price further
in the contract pH ; sH , as the high type then will switch to the low type’s
contract p ; s . However the …rm can do better. It can modify the contract
o¤er the low type buys, by reducing both s and p, as indicated by contract
o¤er c in the …gure. Doing that lower’s the pro…t obtained by selling to the
poor consumer (note that c is on a lower iso pro…t curve than a), but makes
it possible to raise the price pH in the rich consumer’s contract (contract
terms d in the …gure). The latter raises pro…t.
Look again at L’s new contract. Note that it represents a downward
movement along a given indi¤erence curve uL = u. The impact of this
movement on the pro…t is in the beginning negligible, since the slope of the
iso pro…t curve is almost the same as the slope of indi¤erence curve. Hence it
costs very little to make a small distortion in L’s contract (distortion, as we
deviate from an e¢ cient contract). However, the modi…cation of L’s contract
makes it much less attractive to the rich consumer (as his relative valuation
of quality s is higher). Hence the …rm can raise pH a lot. This is known as
rent extraction.
The optimal set of contracts balances these two e¤ects.
To conclude: the optimal contracts satisfy the following (general feature):
58
yield in total four constraints. However, we know from the discussion above
(see the bullet points) that two of the constraints will not bind: the high type
obtain a utility strictly above his reservation utility, and the poor consumer
strictly prefers pL ; sL to pH ; sH . Hence we have two binding constraints, and
we can specify the problem as follows:
M ax pL sL + pH sH
pL ;sL ;pH ;sH
s:t:
2 + ln sL pL = u
4 + ln sH pH = 4 + ln sL pL
L = pL sL +pH sH (2 + ln sL pL u) ( ln sH pH ln sL + pL )
LpL = 1 + =0
1 1
LsL = 1 + =0
sL sL
LpH = 1 + =0
1
LsH = 1 =0
sH
ln sL pL = u
ln sH pH = ln sL pL
59
Since = 1 we get = 2 from the …rst equation. Inserted in the
second equation yields
1 1
1+2 = 0
sL sL
1 1
=
sL 2
We know the the e¢ cient quality level for the low type is
1
=1
sL
we see that the quality level is ine¢ ciently high if < 1 and ine¢ ciently low
if > 1, exactly as we know from the graphical elaboration.
The …nal step is to evaluate the equilibrium from a social perspective. In
particular, are the quality choices socially optimal? If not, what can you say
about the distortions.
As is clear from the …gure H’s contract is e¢ cient, but not L’s. The …rm
has an incentive to distort Ls contract in a way that makes it less attractive
to H. The reason is that H’s incentive constrained binds, and by making L’s
contract a less attractive choice, the …rm can extract more rent from the
rich consumer. However, the …rm will never make quality distortions in H’s
contract - the more attractive the contract is to H, the more rent the …rm
can extract from o¤ering the contract.
60
In the …rst case, the best the …rm can do is to choose contract a in
the …gure. The …rm extracts all rents from the rich consumer (he is on
his reservation utility), and the contract maximizes value creation (e¢ cient
contract).
If the …rm sells to both, the best it can do is o¤er contract b in the …gure
to the right. Note that the contract o¤er must be on the reservation utility
uL = u, or below. The highest pro…t is then represented by the tangency
point.
61
of consumer demand, with downward sloping demand curves, price discrimi-
nation distorts the distribution of consumer goods across consumers. Recall
that an e¢ cient distribution of consumer goods requires that marginal rate
of substitution between any two goods must be the same for all consumers.
This condition is not satis…ed under price discrimination.
6.7 Competition
In the previous section we assumed that the market were monopolized. Sup-
pose instead that the market is competitive with free entry. The model then
corresponds to section 17.4.3. in the text book, "competitive retailing".
In equilibrium, any active …rm gets zero pro…t. If not, it would be possible
for a new …rm to enter, and make a pro…t from o¤ering a slightly more
favorable contract.
The equilibrium has the following characteristics:
1) Both consumers get access to their respective preferred qualities. The
argument is exactly the same as the argument we used show that the equi-
librium as described in the book, is not really an equilibrium. Suppose that
one of the consumers does not get access to her preferred quality. Then a
…rm can enter, optimize the quality, and make a pro…t.
2) Active …rms gets zero pro…t.
62
Hence, the rich consumer’s quality will be sH = , whereas the poor
consumer’s quality will be sL = 1. Finally, zero pro…t implies:
pL = sL = 1
pH = sH =
dp(q M ; s) M
q cs (q M ; s) = 0
ds
We also know that the constrained e¢ cient quality, given q M , is
R qM dp(z;s)
dz M
0 ds
M
q cs (q M ; s) = 0
q
1
p(q; s) = D q
s
p(q; s) = D + s q
p(q; s) = s (D q)
Note that the …rst speci…cation is the one used in the textbook.
Solution: Note the following, you are supposed to evaluate the deriva-
tives for quality level q = q M . Let’s …rst calculate the monopoly pro…t with
the three speci…cations of demand.
1 D c
p(q; s) = D q ! qM = s
s 2
M D+s c
p(q; s) = D + s q ! q =
2
c
D
p(q; s) = s (D q) ! q M = s
2
63
Note that q M is increasing in s in all three cases.
We di¤erentiate with respect to s
@p(q M ; s) qM
=
@s s2
@p(q; s)
= 1
@s
@p(q; s)
= D qM
@s
and insert for q M
dp(q M ; s) 1D c
=
ds s 2
dp(q M ; s)
= 1
ds
dp(q M ; s) D + sc
=
ds 2
We next calculate the marginal impact on the average consumer. We inte-
grate and divide by qM
R qM dp(z;s) R qM z
0 ds
dz 0 s2
dz 1 qM
M
= M
=
q q 2 s2
R qM dp(z;s) R qM
0 ds
dz 0
1dz
M
= =1
q qM
R qM dp(z;s) R qM
0 ds
dz 0
(D z) dz 1
M
= =D qM
q qM 2
Next, we insert qM
R qM dp(z;s)
0 ds
dz 1D c
=
qM s 4
R qM dp(z;s)
0 ds
dz
= 1
qM
R qM dp(z;s) c
0 ds
dz 3D + s
M
=
q 4
64
1
Let’s compare: if demand is p(q; s) = D s
q we have
dp(q M ; s) 1D c
=
ds s 2
R qM dp(z;s)
0 ds
dz 1D c
M
=
q s 4
We see that the monopolist has a stronger incentive to invest in quality than
the planner (quality is more valuable to the marginal than to the average
consumer).
If p(q; s) = D + s q then
dp(q M ; s)
= 1
ds
R qM dp(z;s)
0 ds
dz
= 1
qM
which is as expected, since consumers are homogenous with regard to the
preferences for quality.
Finally if p(q; s) = s (D q) we have
dp(q M ; s) D + sc
=
ds 2
R qM dp(z;s)
0 ds
dz 3D + sc
=
qM 4
Since the average consumer appreciate quality more than the marginal con-
sumer, the monopolist has to weak incentives to invest.
Exercise on constrained e¢ cient quality level
Suppose the government can regulate the quality level (which is actually
quite common - e.g. safety standards), but nothing else. Hence we have a
simple game where the planner chooses s and the …rm chooses p.
Let the …rm’s cost function be
c(q; s) = cq + (s)
p(q; s) = D + s q
65
We consider two alternative speci…cations of the game:
1: Simultaneous game
The …rm and the planner make their choices independently and simulta-
neously
2: Sequential game
Solve the two games, and compare the quality levels with the quality level
the government would have chosen, if the government could choose both price
p and quality q (which is the …rst best allocation).
Solutions:
Consider …rst the simultaneous game. The …rm chooses price p and the
government sets quality s. We use the Nash equilibrium concept: that is,
both choices are best replies.
The social planner maximizes social surplus, for a given q, that is solving
Z q
max (D + s z) dz c(q; s)
s 0
or Z q
max p(z; s)dz cq (s)
s 0
which yields …rst order condition:
Z q
dp(z; s) 0
dz (s) = 0
0 ds
Note that this is one equation which determines s as function of q : the
planner’s reaction function, or best response function.
The …rm solves the following optimization problem:
maxp(q)q c(q; s)
q
or written as
max (D + s q) q cq (s)
q
66
with …rst order condition:
max (D + s q) q cq (s)
q
The sign and size of the Lagrange multiplier are important for the
interpretation of the …rst order condition.
67
7 Ramsey pricing
A common feature of the cases discussed in the course is that …xed and sunk
costs represents a substantial part of total costs. Examples are telecommu-
nications networks, infrastructures, Internet markets etc.
With large common costs, …rst best prices (marginal cost) are incom-
patible with cost recovery, since average production cost exceeds marginal
cost.
Thus …rst best e¢ ciency requires transfers. Without subsidies, the pricing
model at best satis…es the condition for second best prices, that is price
structures which maximize welfare subject to the …nancial constraint. This
is in the literature known as the Ramsey problem.
68
The solution of the problem is given by the …rst order conditions:
@L
= 0
@q1
@L
= 0
@q2
and the constraint
69
where the last step follows from the de…nition of demand elasticity "1 :
p1 1
"1 = q10 (p1 ) =
q1 (p1 ) p01 (q1 ) p1q(q1 1 )
pi (qi ) ci 1
=
pi (qi ) "i 1 +
X
(pi (qi )qi ci q i ) = F
i
known as the "inverse elasticity rule". We see that the markup on marginal
cost relative to price should be higher the less price sensitive demand is. This
principle has a simple interpretation: if the price exceeds marginal cost, there
will be a deadweight loss (e¢ ciency loss), see the …gure
70
Hence, if price sensitivity di¤er between market segments, welfare is im-
proved if the mark-up is increased in segments with low price sensitivity, and
increased in segments with high. Think about the extreme case, a market
segment with completely inelastic demand. As the mark-up has no e¤ect on
demand, there is no associated deadweight loss - such segment would be the
perfect object of taxation.
The next step is to derive the monopoly solution, and compare it with
the Ramsey solution.
71
Note the following: the two price levels are equal if the monopoly prices
yields a revenue which exactly corresponds to total costs, in which case there
is no monopoly rent in equilibrium. In the equation above, this is repre-
sented by a value of approaching in…nite (for an explanation of this, see
the exercises below).
p1 ! 1 + p2 ! 2 p (9)
72
that (9) must hold. The …rm’s problem:
s:t: p1 ! 1 + p2 ! 2 p
The Lagrangian is
pi (qi ) ci 1
=
pi (qi ) "i 1 +
The obvious question is: how should the regulator set ! 1 and ! 2 ? Note
that the right hand side of (10) is independent of i if ! i is proportional to
qi . If so, the price pro…le chosen by the regulated …rm, corresponds to the
Ramsey pro…le. Certainly, the planner has limited information about qi , but
it seems reasonable to think that it is easier for the regulator to get access
to information about production levels than to get access about demand
elasticities.
Comment: Regarding the interpretation of . If the constraint is not
binding = 0, the price ceiling is higher than the monopoly prices, and thus
the regulation does not bind.
and 1
p2 (q2 ) = 2q2 2
73
The cost structure is
C(q1 ; q2 ) = c1 q1 + c2 q2 + F
where F is a …xed common cost.
Suppose the regulator can regulate the prices p1 and p2 . The regulator’s
target is to maximize social surplus, conditional on a break even constraint.
Thus the constraint is:
p1 q1 + p2 q2 c1 q 1 c2 q 2 F 0
Use the Lagrangian multiplier technique to solve the regulator’s optimiza-
tion problem. How large is the optimal mark-up in segment 1 (that is p1p1c1 )
relative to the mark-up i segment 2?
The Lagrange multiplier has the following interpretation:
dW
=
dF
that is: the marginal impact on social welfare, in optimum, from an increase
in the …xed cost, equals (thus is positive). Obviously, social welfare
declines if F increases. Try to show that social welfare does not decline
linearly in F , and provide intuition to this result.
3
21
q1 =
3 c1
2
1
q2 =
c2
We can calculate consumer surplus in each of the sectors:
Z q1 Z q1
1 1 1 13 3 23
p1 (z)dz = z 3 dz = jq01 z = q
0 0 1 13 2 1
and Z q2 Z q2
1 2 1 1
p2 (z)dz = 2z 2 dz = jq02 1 z1 2 = 4q22
0 0 1 2
We can also simplify sales revenue:
1 2
p1 q1 = q1 3 q1 = q13
1 1
p2 q2 = 2q2 2 q2 = 2q22
We got the following expression for the Lagrangian function:
3 23 1 2 1
q1 c1 q1 + 4q22 c2 q2 + q13 c1 q1 + 2q22 c2 q 2 F
2
We di¤erentiate w.r.t. q1 and q2
2
1 2 23 1
q13 c1 + q c1 = 0
3 1
1 1
1 1
2q22 c2 + q22 c2 = 0
that is
1 2 1
q1 3
c1 + q 3
c1 = 0
3 1
1 1
2q2 2
c2 + q2 2
c2 = 0
75
which yields solutions
1
q1 3
c1 1
1 =
q1 3 31+
1
2q2 2
c2 1
1 =
2q2 2 21+
C(q1 ; q2 ) = c1 q1 + c2 q2 + F
77
Inserting from the demand functions yields
2 1
= q13 c1 q1 + 2q22 c2 q 2 F
pM
1 c1 1
M
=
p1 3
M
p2 c2 1
M
=
p2 2
Hence
pM
1 c1 2 pM
2 c2
M
= M
p1 3 p2
In the Ramsey problem we calculated (* refers to social (second best)
optimum)
p1 c1 1
=
p1 31+
p2 c2 1
=
p2 21+
Note that the monopoly and the social planner have the same preferences
regarding how price should be balanced. However the price levels di¤er, since
1+
< 1, generally the monopoly prices exceed second best prices.
Note however that if is very large, the second best prices become close
to monopoly prices - which is intuitive: we showed that (see solutions set
78
4) is an increasing function of F . As F goes up, and conditional on the
…nancial constraint, prices must increase (otherwise there will be a de…cit).
Final observation: we could think about a situation where F is so high
that the unregulated monopoly’s net pro…t is exactly zero - total revenue
equals total cost, in which case pM = p . Since
pM
1 c1 1
M
=
p1 3
and
p1 c1 1
=
p1 31+
this obviously means that must converge to in…nity (which is the only way
1+
can be equal to 1). How can this result be interpreted?
Think about the Ramsey problem - in which case we maximize consumer
surplus conditional on a …nancial constraint. Clearly, if F is su¢ ciently
large, there is only one possible allocation compatible with the constraint -
the monopoly quantities. If F is increased even further, the optimization
problem ceases to have an solution, as there exists no allocation such total
costs can be covered by total revenue. Hence the market breaks down.
Think about the impact on social welfare associated with a small increase
in F , evaluated exactly at the level of F at which the …rm breaks even before
the increase, but after the increase runs a de…cit. Hence, before the increase,
social welfare is exactly equal to consumer surplus (since the monopoly breaks
even). After the increase, consumer surplus drops to zero. Since consumer
surplus is positive, there is a discontinuity in social welfare as function of F .
This discontinuity explains why a marginal increase in F has an in…nitely
negative e¤ect on welfare at this critical value of F .
8 Access pricing
79
which provides the regulator with means that go far beyond the competition
acts’. E.g., the regulator can intervene in the market if they …nd that the
market structure potentially may lead to abuse of dominance –whereas under
the competition law, the competition authorities (Konkurransetilsynet) must
prove that an actual abuse has occurred. In the literature, the former is often
referred to as "ex-ante regulation", the latter to "ex-post".
Competition in telecommunication has two "dimensions", as illustrated
in the …gure.
80
So-called "MVNOs" (TDC, Hello, Ventelo) are partly integrated (e.g.,
with their own sim-cards), but do not have their own network. Hence they
need access to one of three integrated networks. As pure service providers
(Teletopia, Fjordkraft, etc), MVNOs compete for customers in the …nal user
market. Note that some of the service providers are owned by the integrated
networks (e.g. Chess, Talkmore, One Call).
81
8.2.1 Fierce competition downstream (Bertrand competition)
Suppose we have a bottleneck network upstream, and a n service providers
downstream. Let consumers’ willingness to pay be P (Q). We have the
upstream marginal cost in the network equal to c, and a downstream marginal
cost cS (which is the same for all …rms). Refer to r as the whole sale price.
Consider …rst full integration, in which case the integrated network is the
sole supplier of the services. Hence there are no independent service providers
downstream. The integrated network’s pro…t is
I = (P (Q) cS c) Q (11)
Let us compare this situation with a market where there are no integration.
With no integration, the pro…t functions are:
U = (r c) Q(P )
S = (P cS r) qS
where X
qs = Q
In the latter case, downstream competition competes the margin down to
zero. Hence the market price P will be:
P = cS + r
U = (P cS c) Q(P ) (12)
Note that the pro…t functions (11) and (12) ends up being identical. With
full integration, the network sets P directly, with the aim of maximizing U .
With no integration, the network sets r, and given r, downstream …rms
compete. Competition downstream drives the margin down to zero, P =
cS + r, which maps the wholesale price r into a consumer price P . That is,
the network sets P indirectly, in that it can steer P using r as an instrument.
82
Let P M be the monopoly price, the price that maximizes U. With no
integration we …nd the optimal wholesale price as follows:
r = PM cS
Note that a single instrument, the wholesale price r, is su¢ cient for the
network to achieve the following: i) maximizise total value, and ii) extract
all value created as pro…ts. This is simply due to Bertrand competition
downstream. Since any pro…t downstream is competed away, all pro…t in the
vertical chain ends up with the upstream monopoly anyhow.
Consider next imperfect competition downstream.
r = c
T = M
where r is the variable part (the wholesale price) and T is a lump sum
transfer from the service provider to the network. Note the simple structure:
the network charges a wholesale price equal to marginal cost upstream -
hence the service provider’s marginal cost equals the true marginal cost in
the vertical chain. It follows that the service provider will charge a price that
maximizes total pro…t. The network then extracts this pro…t using the lump
sump transfer T .
If downstream competition has the characteristics of oligopolistic compe-
tition, the situation is slightly more complex - but the main conclusion goes
through. Note the following: the network sets the wholesale price r, which is
83
part of downstream …rms’marginal costs. Raising r increases marginal cost
downstream, which completly or partially will be passed on to consumers.
Hence consumer price P is an incrasing function of r. This will hold inde-
pendent of the details in downstream competition. Since P is increasing in
r, there exists an r such that P = P M , as illustrated in the …gure:
84
support P M . Since competition is imperfect, each service provider obtains
a pro…t (before transfer T ) equal to some pro…t level D (with magnitude
depending on the degree of competition), which will be extracted by the
network through the lump sum transfer: T = D .
Note that in all these market con…gurations, the network earns the monopoly
pro…t. We will now show the …rst important result: the network has no in-
centive to distort competition downstream.
85
8.3 Chicago school argument - modi…cations
1. Imperfect contract and the "hold-up" problem
2. Entry barriers, raising rivals costs (current and future rivals) by creat-
ing bottlenecks.
p(q; s);
p(q; s)
86
We consider two alternative speci…cations of the game:
We solve the models by backward induction, starting with the last stage.
Alternative A follows the same logic as before. The two parties have
incentives to sign an e¢ cient contract, which realizes all gains form trade.
Formally, given contract terms [r; T ] the downstream …rm solves the following
maximization problem:
87
Total pro…t, if they reach an agreement, is
+ D
U
= (r c) q + T + p(q; s)q rq (s) T
= p(q; s)q cq (s)
The …rms have a common interest in maximizing total pro…t. Note that
any contract proposal which yields a lower total pro…t than what is possible
to create, will be dominated by a proposal which maximizes total pro…t: they
can simply split the gain, and both be better o¤.
There is a con‡ict of interest in how to split the pie. Let denote the
downstream …rm’s relative bargaining strength, which means that the down-
stream …rm captures a share of total pro…t.
Total pro…t is maximized if (take the derivative of U + D with respect
to q and s):
dp(q; s)
p(q; s) + q c = 0 (15)
dq
dp(q; s) 0
q (s) = 0 (16)
ds
Comparing (23) and (24) with (21) and (22) you see directly that if the
variable part r is set equal to marginal cost c, then the downstream …rm’s
production decision, as well as its quality choice, correspond to what is opti-
mal for the two parties together. Since T is a lump sum transfer, it does not
impact the production and quality decisions.
Note: the two part tari¤ is essential for this result. With two instruments,
the role of the variable part r is to provide e¢ cient incentives, and the …xed
part T splits the pie. If the contract was a uniform price contract, with no
…xed fee, an e¢ cient solution would not be reachable. However, business
contracts are typically more sophisticated than uniform price contracts.
To conclude, with a two part tari¤, production and quality choices corre-
spond to an integrated monopoly’s optimal choices, refer to that as q M and
sM , and the two …rms’pro…t levels will be
D = p(q M ; sM )q M cq M (sM )
U = (1 ) p(q M ; sM )q M cq M (sM )
88
This also means that if the downstream supplier is more e¢ cient than the
upstream in retailing, the upstream monopoly has no incentive to foreclose
the more e¢ cient downstream …rm.
Consider next alternative B, in which there are three stages:
Stage 1: Downstream …rm invests in s. The investment is sunk.
Stage 2: Contract terms [r; T ] are negotiated
Stage 3: Downstream …rm chooses q
I referred to this alternative as a "short term contract". As we will see,
this sequence of stages creates an ine¢ ciency which is avoided if the parties
can sign a long term contract. We begin with stage 3:
Given contract terms [r; T ] and given s which is …xed in period 1, the
downstream …rm solves the following maximization problem:
U = I
D = (s)
89
The value of the contract is then split between the parties.
We have as before, joint pro…t:
+ D
U
= (r c) q + T + p(q; s)q rq (s) T
= p(q; s)q cq (s)
U+ D U ( (s))
= p(q; s)q cq I
The surplus is split between the parties, hence the downstream …rm gets a
share and the upstream a share 1 .
In stage 1, the downstream …rm invests in quality. The …rm’s optimization
problem is:
max (p(q; s)q cq I) (s)
s
with …rst order condition (if it is optimal for the downstream …rm to sign a
contract)
dp(q; s) 0
q (s) = 0
ds
which is one equation which determines sD as function of .
Note that, because quality investments are sunk before the contract is
negotiated, e¢ cient bargaining splits gross pro…t (before investment cost is
deducted), instead of net pro…t. Another way to say it is the following; the
two parties split the revenue ‡ows but not the investment cost.
Obviously, this has a negative impact on the downstream …rm’s invest-
ment incentives. Unless the downstream …rm has all bargaining power ( =
1), it is not possible to obtain an e¢ cient solution.
Furthermore, it is certainly possible that the downstream …rm …nd that
it cannot obtain a positive pro…t. If the downstream …rm invests sD it obains
pro…t
(p(q; sD )q cq I) (sD )
which can be positive or negative. In the latter case, the terms of access does
not provide the downstream …rm with a margin which covers the investment
cost.
90
The problem would have been solved if the two parties could negotiate
before the downstream …rm invests (that is signing a long term contract) -
however this requires not only that a long term contract is feasible and not
subject to renegotiations.
91
Suppose an independent retailer is foreclosed by an upstream bottle-
neck.The monopolist "protects its market power by raising entry barriers"
P =D q
(D qA q B ) qB cqB
92
which creates value
2
D qA c
B =
2
a pro…t ‡ow the monopolist can extract using the transfer T .
There is, however, one critical issue here: …rm B has no information about
qA , neither the content of the contract or whether a contract with A exists.
Thus, how can B calculate the pro…t level the contract eventually creates?
Something he has to do in order to make an assesment of the contract o¤er,
including the proposed transfer TB .
Since B does not know anything about As contract terms, he has to make
som conjecture, and suppose his beliefs are rational in the following sense: he
thinks through the problem and calculates what would be the optimal o¤er
to provide retailer A, qA , and the corresponding transfer TA . Given this, he
accepts the contract o¤er qB ; TB if it gives him a non-negative pro…t. Note
that A is in a symmetric position, she is unaware of the contract terms B
eventually obtains.
What they know is that the monopolist face two potential retailers, and
at any time has the incentive to negotiate, and renegotiate the contracts term
in order to maxize the value the contract creates. Hence, given the outcome
of the negotiations with A (that is given qA ) the monopolist maximizes
A = (D qA qB ) qB cqB
Similarily, given the outcome of the negotiations with B the monopolist max-
imizes
B = (D qA qB ) qA cqA
However, simultaneously maximizing these two pro…t ‡ows makes the mo-
nopolist acting as a cournot duopolist in the market. Formally, maximizing
A w.r.t. qA given qB , and maximizing B w.r.t. qB given qA , is mathemat-
ically equivalent to the cournot dupopoly problem. Hence, given that the
monopolist has all bargaining power, the best it can do is to extract the sum
of dupopoly pro…t which is something less than the monopoly pro…t.
The problem is that the monopolist cannot credibly commit not to rene-
gotiate the contract with either …rm. Hence negotiating with …rm A, the two
parties A and the monopolist have incentives to expand production at the
cost of …rm B.
93
The fact that the contracts are commercial and secret is essential. If
contract terms were public, the monopolist could write a set of general access
conditions into the agreements, and such that the two …rms total production
sum up to the monopoly quantity, and …nally extract using the transfers T.
How can the monopolist deal with the problem? One option is to o¤er
an exclusive contract, hence writing into the contract that the retailer has
exclusivity. Hence the monopolist will not sell through any other retailer.
If the two retailers are identical (as in the illustration above), there is no
e¢ ciency loss associated with using only one of them.
Exclusive contracts are, however, met with considerable skepticism from
competition authorities. In particular in EU due to their potentially adverse
e¤ects on internal trade. Seemingly a ban on exclusive contract may improve
e¢ ciency as the consumer price declines, due to the contract externality.
However, a ban may induce the monopolist to integrate downstream and see
only through its internal retailer.
Final comment: the contract externality problem is a potential source of
distortions. To illustrate this, suppose the two retailers are di¤erentiated,
each with a speci…c competitive advantage. Hence, maximizing total value
requires that the monopolist sign contracts with both. However due to the
contract externality problem, the monopolist is forced to o¤er one them ex-
clusivity, and hence the other retailer’s competitive advantage is not realized.
94
8.4.1 "Price-minus" principle
Suppose an integrated network, I, faces an independent service provider,
S. To make the illustration more realistic, assume the sercive provider is
a small …rm, with a limited capacity, qS , to serve customers. The service
provider’s marginal cost is cS whereas the integrated network’s retail cost is
cI . Consider the prospects for a pro…table business contract.
The pro…t functions are:
I = (P (Q) cI c) (Q qS ) + (r c) qS
S = (P (Q) cS r) qS
qS = qS if P (Q) cS r 0
qS = 0 otherwize (foreclosure)
I = (P (Q) cI c) Q + [r (P (Q) cI )] qS
r (P (Q) cI ) 0
P (Q) cS r 0
that is
cI cS :
The integrated network has an incentive to sign a contract with the service
provider, if and only if cI cS . This is also the situation where the contract
is socially e¢ cient.
95
Suppose now the regulator introduces an access price regulation, known
as a "margin squeese test", which is the current regulation regime in Norway.
A margin squeeze test compares the retail price of a dominant undertaking
P (q) with its wholesale price r and retail costs cI . The idea is that an equally
e¢ cient service provider (equally e¢ cient as the integrated network’s own
service division) obtains a positive pro…t, hence it must be that
r P (Q) cI
Logically, for e¢ ciency, a less e¢ cient service provider should have a negative
margin (ine¢ cient entry is not warranted), hence this narrows down the test
to
r = P (Q) cI
Note that r = P (Q) cI satis…es the constraint (18). With a perfectly
designed margin squeese test, the integrated network obtains the same pro…t
from providing access, as it obtains from running the retail activitities in-
house. Hence the independent service provider extracts the entire pro…t from
any superior e¢ ency.
Obviously, a perfectly designed margin squeeze regulation is e¢ cient. The
challenge is that there are regulatory costs which are not accounted for.
Furthermore, the model is stylized and there are other characteristics of the
market situation which a business contract takes into account, and which is
impossible to internalize in a regulatory regime. The next sections discusses
this.
96
I = (P (Q) cI c) (Q qS ) + (r c) qS
S = (P (Q) cS r) qS
qS = qS if P (Q) cS r 0
qS = 0 otherwize (foreclosure)
r [(1 )c + (P (Q) cI )] 0
P (Q) cS r 0
P (Q) cS r P (Q) cI
which exists if
cS < cI
If < 1 the interval for r is larger (everything else equal):
97
Obviously, regulating access terms is challenging if the crowding out factor
di¤er between …rms. A business contract is pro…table, and socially e¢ cient,
at a low access price if the crowding out factor is small. Di¤erentiation of
access prices would be possible as part of voluntary contract arrangements.
However, the information required to adjust the contract terms as part of a
regulatory regime, is not available to the government.
Furthermore, any uniform access price is ine¢ cient, as it may foreclose
innovative service providers, which builds up new markets at a higher cost,
and induce ine¢ cient entry from service providers for which the crowding
out factor is close to one.
cIA cSA
cIB cSB
that is if the independent service providers are more e¢ cient that the inte-
grated network.
It follows that e¢ ciency requires segment speci…c access prices. Take the
"price minus" rule as derived above:
rA = PA (Q) cIA
rB = PA (Q) cIB
Suppose the regulation is based on the average mark up. The …gure below
illustrates that this may lead to ine¢ ciencies.
98
We see that service provider B has an incentive to enter market segment
2, despite the fact that it is less e¢ cient than the network owner is. On the
other hand, the more e¢ cient service provider A, is foreclosed from market
segment 1 because of the non-discrimination rule.
99
In the telecommunication market, independent service providers are po-
tential business customers for Ice. The access regulation of Telenor harms
Ice in that it will be harder for Ice to compete for business customers. To
illustrate the e¤ects suppose the telecomunication market consist of business
market B and the retail market C.
Ice’pro…t is
100
of how long it is actually using the quay. The Hurtigruten has two ships which
arrive daily, with a total length of time at the quayside of 4.25 hours per 24 h.
The ship is then charged two quay teams per 24 h. The practice is surprising
when ships arriving several times a day have accumulated time at the dock
that is longer, only paying one quay fee"
Hurtigruten’s claimed that the correct pricing principle is a fee per "meter
hour": the length of the ship multiplied with the time it occupies the quay.
Note that harbour costs have the following characteristics:
– Maintenance costs
– Congestion costs
With no congestion costs, total costs are substantially larger than vari-
able costs
101
The inverse elasticity rule: higher mark ups in segments with inelastic
demand
Fixed cost F
Marginal costs c1 ; c2
Fixed capacity q
102
The planner maximizes social surplus W , subject to i) a …nancial con-
straint, ii) a capacity constraint. Hence the planner’s problem is:
Z q1 Z q2
max W = p1 (z)dz c1 q1 + p2 (z)dz c2 q2
0 0
s:t: p1 q1 c1 q 1 + p 2 q 2 c2 q 2 F 0
q2 q
Z q1 Z q2
L= p1 (z)dz c1 q 1 + p2 (z)dz c2 q 2
0 0
+ (p1 q1 c1 q1 + p2 q2 c2 q 2 F)
(q2 q)
(p1 q1 c1 q 1 + p 2 q 2 c2 q 2 F ) = 0
(q2 q) = 0
103
is zero. Recall the economic interpretation of as the value of increasing
capacity. It is clear that investing in a larger capacity has no value if the
port is never crowded. If however the capacity binds in peak periods, delays
occur, and a higher capacity would have been valuable: hence is a positive
number.
Thus we have four equations and four endogenous variables: q1 ; q2 , and
. Consider …rst the situation where capacity does not bind.
p1 (q1 ) c1 1
=
p1 (q1 ) "1 1 +
p2 (q2 ) c2 1
=
p2 (q2 ) "2 1 +
p1 q1 c1 q 1 + p 2 q 2 c2 q 2 F = 0
q2 < q
p1 (q1 ) c1 1
=
p1 (q1 ) "1 1 +
p2 (q2 ) c2 1
= +
p2 (q2 ) "2 p2 (q2 ) 1+
p1 q1 c1 q 1 + p 2 q 2 c2 q 2 F = 0
q2 = q
Note the following. Everything else equal, a higher increses the price
in peak periods, p2 . But remember that is endogenous - hence we cannot
draw a too simplistic conclusion.
Still, we can, by reasoning, …gure out how is impacted by q. To put it
into a story, say a section of the port is closed due to maintenance, which
means lower capacity. The immediate e¤ect is that capacity binds in peak
periods. An e¢ cient port will then raise the price such that the market is
104
cleared: this is the simple principle of using the price to allocate a scarce
resource across buyers, illustrated by the following …gure:
Suppose the initial capacity is q2old , which as illustrated is so high that the
capacity does not bind even in peak periods, q2 (P ) < q2old . Then the capacity
jumps down to q2 , at a level at which demand exceeds supply at the original
Ramsey price. Hence the quay fee has to be raised in order to clear the
market.
Obviously, the higher price in peak periods raises the port’s revenue, thus
improves the …nancial situation. Unless the port reduces the price in load
periods, total revenue will exceed total cost (since revenue in peak period has
increased), and hence the …nancial constraint will not bind. This cannot be
optimal. The better …nancial situation means that declines, and so does
p1 .
Is …rst best achievable? Yes, if net revenues during peak periods are suf-
…cient to cover total costs. Then we have = 0 and the optimum conditions
are:
p1 (q1 ) c1
= 0 ) p1 (q1 ) = c1
p1 (q1 )
p2 (q2 ) c2 =
p 2 q 2 c2 q 2 F > 0
q2 = q
105
Actually, given the constraint q2 = q, the allocation is e¢ cient, as it
corresponds exactly to the solution of a pricing problem with the …nancial
constraint removed.
However is positive, thus increasing the ports’ capacity is valuable.
As mentioned above, if exceeds the marginal cost of providing capacity,
increasing the capacity is socially optimal.
Note that, due to the …nancial constraint, the Ramsey price is so high
that the capacity does not bind. Hence = 0. This means investing in more
capacity is a pure waste. However if we remove the …nancial constraint, the
price will drop to P (q2 ) and will be positive. This illustrates a general
point: unless we are in …rst best, optimal allocation of resources in one
market segment cannot not be derived indepently of the distortions in other
segments. Public economics is di¢ cult.
106
9.3 Priority pricing
A related topic is priority pricing. Consider the follwoing example: Hur-
tigruten follows a strict time schedule, wheras as a cargo is more ‡exible. In
situations were there will be delays, e.g. because of bad weather, a market
for priority can be created - re‡ecting the di¤erences in delay costs. If a …rm
has priority, it will not experience any delay.
Take the following simple illustration: the value of using Bodø port for
Hurtigruten and the cargo is
On time Risk of delay
Hurtigruten 100 60
Cargo 60 60
and suppose bad weather occurs with probability 0.5. The port has
di¤erent alternatives. It can charge both companies a price 60, and in case
of bad weather either of them may experience delays.
An alternative scheme would be to o¤er priority contracts. Note that
the cargo has no willingness to pay for priority, hence the most the port
can extract from the cargo, is 60. Hurtigruten is willing to pay more, up to
100, if it is given priority. However, if the price for priority is set too high,
Hurtigruten would be better o¤ choosing a standard contract at 60. If so,
it experiences delays with probability 1/2. Hence the expected value of the
standard contract would be
1 1
100 + 60 = 80
2 2
Hence the port can charge Hurtigruten at most 80 for a priority contract,
if the price exceeds 80, Hurtigruten is better o¤ with the standard contract
(this constraint corresponds to the incentive constraint in contract theory).
In equilibrium the port charges Hurtigruten access with priority at a price
80, whereas the Cargo pays 60.
Obviously the market is e¢ cient. Priority, which means that the delay
is avoided, is allocated to the …rm with the highest willingness to pay. Any
other rationing mechanism would yield ine¢ ciencies.
Note that the market mechanism resembles price discrimination of second
degree except for the fact that there are no ine¢ ciencies associated with the
marginal customers contract: the delay is real.
A price discrimination story could be as follows. There are no risk of
delays from the beginning. However, as the the two customers marginal
rates of substitution between price and delay di¤er, the seller (the port) can
107
use delays to extract rent from Hurtigruten - thus using arti…cial delays as
an instrument.
qN = pN"
qD = pD"
where > 1 is a shift parameter. Note that the demand elasiticity, Elp q =
", is the same in the two periods.
Suppose service capacity does not bind.
2. Suppose increases. Explain how this impacts the optimal price level
and the Lagrangian multiplier. Provide intuition to the e¤ect has on
the Lagrangian multiplier in optimum.
Solution
We have the Lagrangian
Z qN Z qD
pN (z)dz cN qN + pD (z)dz cD qD + (( pN c) qN + (pD c) qD F)
0 0
108
Since the demand elasticity does not change over the 24 hours, the optimal
mark up is constant
p c 1
= (19)
p "1+
where
p = pN = p D
hence from the demand functions
qN = p "
qD = p "
( pN c) qN + (pD c) qD F
= ( p c) p " + (p c) p " F
= (1 + ) ( p c) p " F = 0
that is
" F
(p c) p = (20)
1+
Note that (19) and (20) are two equations that determine and p as
functions of the exogenous variables, F , , " and c. Note also the recursive
structure: (20) determines p and then (19) determines .
Consider an increase in . Since c, F and " are given, it follows from (20)
that p declines (note that the left hand side is strictly increasing in p). From
(19) it is clear that declines (the left hand side is monotonically increasing
in p and the right hand side is increasing in .)
Economically, a higher (which represents a positive shift in demand)
makes it possible for the port to cover the …xed cost with a smaller markup,
which is bene…cial as it makes the distortions associated with pricing above
marginal cost less severe.
Note the interpretation of :
@L dW
= =
@F dF
where W is social surplus. A lower , due to higher demand, re‡ects that the
distortions associated with a price which exceeds marginal is, on the margin,
109
smaller. Recall that the deadweight loss increases at the rate of the square
of the mark up, not linearly in the mark up.
Part 2:
Suppose maximum port capacity is q, and assume the capacity is binding
in day time.
2. Discuss how impacts prices and the two Lagrange multipliers. Ex-
plain the result. I suggest that you simplify the model as follows: as-
sume demand is iso-elastic (that is " = 1). With this simpli…cation you
solve explicitly for and . Note that the assumption " = 1 is incom-
patible with pro…t maximization (since the monopoly price approaches
in…nity), but compatible with welfare maximization.
Solutions:
We have the Lagrangian
Z qN Z qD
pN (z)dz cN q N + pD (z)dz cD q D
0 0
+ (( pN c) qN + (pD c) qD F)
(qD q)
pN c 1
= (21)
pN "1+
pD c 1
= + (22)
pD " 1+
(pN c) pN" + (pD c) q F = 0 (23)
q = pD" (24)
110
as function of the capacity q and the shift parameter . The interpretation
is the following: pD has the role of rationing demand such that total demand
is compatible with the limited capacity. If demand shifts outwards (higher
), then pD must increase, and increase more, the less elastic deman is. We
…nd that:
1
El pD =
"
Inserting pD in (23) yields one equation which determines pN . Since an
increase in yields a higher pD , then pN declines (ref 23), and thus declines
(from 21).
Finally you can use (22) to calculate . It is a bit messy to calculate how
impacts - to simplify, consider iso-elastic demand. With " = 1 we have
the following four conditions:
pN c
=
pN 1+
pD c 1
= ( + )
pD 1+
pN c
+ (pD c) q F = 0
pN
pD =
q
with solutions:
F + cq
=
1 (F + cq )
1 cq
= 1
1 (F + cq )
c
pN =
1 (F + cq )
pD =
q
which shows that an increase in yields a lower pN , a higher pD , a lower
and a higher .
Priority pricing
We take the example from the slides dated February 14. The story is:
111
Hurtigruten follows a strict time schedule
Solution: Suppose the port does not o¤er a contract with priority. Then
the best it can do is to charge 60 and serve both ships, which yields a revenue
of 120.
O¤ering
10 Damage compensation
How can we compute the damage from an illegally high price? Suppose a
cartel or a dominant …rm abuses market power, and charges an illegally high
price. Their customers are harmed. If the illegal action is detected, those who
are responsible will be punished. In addtion, anyone who is harmed by the
overcharge may claim compensation. According to the law, the compensation
should re‡ect the actual harm.
To estimate the harm, you have to evaluate the competitive structure in
the market and apply economic theory
We will brie‡y discuss:
112
Harm due to a larger overcharge (why this di¤erence?)
Does competitive structure matter (e.g. does it matter if the …rm
seeking compensation is a monopoly or competes …ercly)?
Does it matter if only one …rm in the market pays the overcharge or
all …rms do?
Are there other strategic e¤ects?
113
First order condition:
R0 (q)F1 w1 = 0
R0 (q)F2 w2 = 0
=) v1 (w1 ; w2 ); v2 (w1 ; w2 )
d (v1 ; v2 )
= v1
dw1
With a small overcharge the harm is the direct, the number of units of
the overpriced input.
Note that this does not depend on whether the …rm competes in a
market, or has monopoly power
The input price increases with the amount t. We can then decompose
the harm in two parts. With the high price the …rm purchases the amount
114
x1 . The overcharge is t, hence there is a direct cost equal to tx1 : that is, the
…rm pays to much for its actual volume. The second part has the following
explanation. Because of the overcharge, the …rm buys less than what it would
have bought in the contrafactual situation, that is the situation without the
overcharge. If the price had been p , the …rm would have bought the volume
x0 . The triangle B is known as the indirect loss.
Note the di¤erence between a small and a large overcharge. If you let t
be small, the triangle B converges to zero, and the loss is represented by the
direct loss xt only.
So far we have considered situations where there is one …rm that is
harmed. Let us now consider oligopolistic markets, and where all …rms pay
the overcharge.
115
10.4 Perfect competition and complete pass-on
If there is perfect competition downstream, the e¤ect described in the previ-
ous section eliminates the hart.
Suppose now that all …rms in the market pay the overcharge
Hence the marginal cost in the market shifts upwards with dc, and so
does the equilibrium price.
dp = dc
Generally yes
q1 = D p1 + ap2
q2 = D p2 + ap1
116
where a > 0.
If …rm A pays an overcharge, B sets a higher price if it knows about the
overcharge than if it is unaware of it. Hence from A’s point of view, the fact
that B is informed is an advantage.
The book price is set by the publisher and is …xed for a given period
of time.
– In Norway until May 1. the year after the book was published
117
11.1 How is competition impacted by RPM?
Resale price maintanence means that the publisher (the upstream …rm)
sets the price
Questions:
118
Hence the net value of increased sale for the vertical chain: 90 percent
of book price
Pro…t functions
119
11.4 Equilibrium sales e¤ort
d 1 dq1 (p1 ; p2 ; s1 ; s2 ) 0
= (p1 r) (s1 ) = 0
ds1 ds1
Two problems:
120
11.5 Optimal price response
Let’s think about the quality dimension in a slightly more dynamic setting.
Suppose store 1 invests a lot in quality, stimulates reading and build up the
market for new releases. Given store 1’s quality choice, what is store 2’s
optimal choice of price level? You can think about store 2 as just across the
street of store 1.
1 = (p r) q1 (p; p; s1 ; s2 ) (s1 )
2 = (p r) q2 (p; p; s2 ; s1 ) (s2 )
First order condition for choice of sales e¤ort:
d 1 dq1 (p; p; s1 ; s2 ) 0
= (p r) (s1 ) = 0
ds1 ds1
In addition, RPM impacts the content of sales e¤ort/ promotions
121
12 Appendix A
12.1 Dependence in demand
The most intuitive way to capture the idea of social welfare with two goods
is the following procedure. Suppose …rst that the second good does not exist
(or that it exists but o¤ered at a price that no one is willing to pay). In this
situation, social welfare is Z q1
p1 (z)dz
0
What is the additional gain of introducing good 2? Now we have to take
into account that good 1 is there, o¤ered at a price p1 . The additional value
of good 2 is Z q2
p2 (z; p1 )dz
0
Let us just illustrate the e¤ects, we will draw on this insight later, but we
will not go deep into this. There are dependencies in demand if the marginal
willingness to pay for one good depends on the price of the other good. Let
the marginal willingness to pay for good 1 be written:
p1 (q1 ; p2 )
122
As an illustration, think about two goods that are almost perfect sub-
stitutes. The second term is almost zero, and the total consumer surplus
is captured from the …rst term. The other extreme would be good that are
almost perfect complements, in which case the …rst term is almost zero.
123