Transcript Markku Stenborg Economics of Competition and Antitrust

TMP 38E050 Advanced Topics
Economics of Competition and Antitrust
Markku Stenborg, PhD (Penn State)
• Currently at Bank of Finland, Research Dept
• April 1st at Ministry of Finance, Economics Dept
– Consultant at CEA
– Previously: Assistant Prof, Turku Business School; Senior
Adviser, Finnish Competition Authority; Senior Manager,
KPMG Transaction Services; Research Fellow, ETLA
– Docent of Law and Economics, University of Joensuu
Course homepage www.cea.fi/hkkk.htm (soon)
HKKK TMP 38E050
1
© Markku Stenborg 2005
• This
course coversof
theoretical
and empirical
issues related to
Economics
Competition
and Antitrust
competition policy such as
– Market definition
– Market power
– Mergers
– Coordination of market conduct
– Predation
– IPRs and high technology
• We shall also review some economics of competition strategy
relevant for analyzing competition policy
• Textbook
– Motta (2004) Competition Policy, Cambridge UP
– Recent articles
– Slides, articles and some notes will appear on homepage
HKKK TMP 38E050
2
© Markku Stenborg 2005
• I assume you master Intermediate Micro, Managerial Econ,
IO Economics
or similar, including
related basic Game
such as
of Competition
andTheory
Antitrust
Nash Equil, and understand basic Econometrics
• Grade will be based on
– Assignment
• Competition Law case, will be handed out during the
course
• Weight of assignment is one quarter toward final
grade, but:
• You need to get a passing grade for the assignment
to get a passing grade for the course.
– Final exam
• Four questions, two answers
• Expect to have (at least) one applied and one
technical question
HKKK TMP 38E050
3
© Markku Stenborg 2005
1. Intro: Competition Law and Policy
• You should read intro material: Motta, Chaps 1 and 2, Intro
from my Lecture Notes, and browse Kovacic & Shapiro and
Commissions publications to get feel for competition policy
– http://europa.eu.int/comm/competition/
• Why do we need competition law and policy?
– Competition promotes efficiency in many activities in
society
• Efficient: maximize surplus generated by production
and exchange, from the asset possessed in society
• Allocative efficiency
• Productive or X-efficiency
• Dynamic efficiency
– Market power reduces efficiency and restricts selfguidance of markets
HKKK TMP 38E050
4
© Markku Stenborg 2005
– Restraints on competition disturb the market process
– Market competition is self-guiding process
Goals of competition laws
• Promote efficiency?
– Yes, but with nonprice competition, simple formulaes (eg.
consumer + producer surplus) for efficiency are deceptive
and misleading
– With non-price competition, customer welfare becomes
more multi-dimensional
• quality of product, speed and security of supply,
introduction of new products and services, etc.
• these aspects may not be measurable and value
judgments are necessary
– Do not strive for perfect competition but promote
”workable competition”
HKKK TMP 38E050
5
© Markku Stenborg 2005
• Protect economic freedom and opportunity by promoting
competition, so that competition can create
– lower prices
– better quality
– greater choice
– more innovation
• Sometimes competition laws have also other goals:
– In EU, competition laws are used to promote single
market within EU
– Sometimes competition laws also protect SMEs
– These other goals can conflict with the main goal of
protecting economic freedom and opportunity
HKKK TMP 38E050
6
© Markku Stenborg 2005
Competition Laws
1. Cartels and such
Article 81 of EU Treaty statest that
[…] agreements between undertakings, decisions by
associations of undertakings and concerted practices […]
which have as their object or effect the prevention,
restriction or distortion of competition […] shall be prohibited
• Article 81 covers much more than formal cartels
• Not only collusion, but also many beneficial forms of
horizontal and vertical cooperation are prohibited
• In US and many other legislations there are similar
paragraphs
2. Monopolization and such
Article 82 of EU Treaty states that
HKKK TMP 38E050
7
© Markku Stenborg 2005
Any abuse […] of a dominant position […] shall be prohibited
[…]. Such abuse may, in particular, consist in:
– imposing unfair purchase or selling prices or other unfair
Competition Laws
trading conditions;
– limiting production, markets or technical development to
the prejudice of consumers;
– applying dissimilar conditions to equivalent transactions
with other trading parties, thereby placing them at a
competitive disadvantage;
– making the conclusion of contracts subject to acceptance
by the other parties of supplementary obligations which,
by their nature or according to commercial usage, have
no connection with the subject of such contracts
• Per se: conduct is prohibited if it meets the legal test
regardless of other issues
HKKK TMP 38E050
8
© Markku Stenborg 2005
• Rule of reason: conduct is prohibited if its negative
consequenses outweigh the positive
• Articles 81 and 82 of EU Treaty are per se prohibitions
• But to prove that firm has abused its dominant position,
authorities must
– show that the firm has dominant position
– conduct was abusive
– In practice, Article 82 has flavor of rule of reason analysis
• In some legal systems, many vertical restraints are dealt
with rule of reason
– Effects of vertical restraints to competition and efficiency
are ambiguous
– Many vertical restraints are solutions to problems, not a
problems for competition
• Vertical restraints can align private incentives in supply
and distribution
HKKK TMP 38E050
9
© Markku Stenborg 2005
3. Mergers
• Transactions that lead to increase in market power or to
some other competition problems may be prohibited
– Illegal to monopolize markets by M&As
• EU: ”A concentration which would significantly impede
effective competition […], in particular as a result of the
creation or strengthening of a dominant position, shall be
declared incompatible”
• US:”the effect of such acquisition may be substantially to
lessen competition, or to tend to create a monopoly”
– Formely EU had dominance test
– Usually SLC-test poses lower threshold for intervention
– Here, monopoly is US legal jargon  strong market power,
dominant position  monopoly in Econ textbooks
HKKK TMP 38E050
10
© Markku Stenborg 2005
Market Power in Case Law
• Read Motta, Ch 2 & 3, Rubinfeld, browse Nevo, and read
Volvo/Scania decision, market definition pp 5-21
Assessment of market power in abuse and merger cases
1. Define so-called relevant antitrust markets
2. Evaluate market power within the relevant markets
– SCP paradigm
• Relevant markets are defined basically by demand
substitution
– Only those goods that provide immediate and intense
competitive constraints to each other belong to the same
relevant market
– In some instances, also supply substitution and entry by
potential competitors are taken into account in market
delineation
HKKK TMP 38E050
11
© Markku Stenborg 2005
• Market power on relevant markets is analyzed:
– calculate market shares
Market Power in Case Law
– analyze competitive strengths of firms
– evaluate degree of actual competitive pressure firm faces
– evaluate entry barriers
– evaluate supply substitution
• In abuse cases, analyze whether conduct of dominant firm
was misuse of market power
– In EU, dominant firms have special obligations
– Dominant firms cannot use their market power to impair
conditions of competition
– Idea is to protect competition, not competitors
• Extra class on Mon. 14.3. at 14 pm, Room A-401
• No class on Thu. 17.3.
HKKK TMP 38E050
12
© Markku Stenborg 2005
2. Economics of Market Definition
• Read Motta Ch 3
• Why we need to define markets in case law?
– Calculate meaningfull market shares
• Diet Coke on cola or on beverage markets? BMW on
luxury car or vehicle markets, say?
– Old SCP idea: market shares tell us something about
market power
• More on this in Oligopoly and Merger sections
– Identify competitors and main competitive constraints
– We are interested in market definition only to extent it
helps in analyzing market power
– Sometimes we can identify and measure market power
w/o defining markets
• More on this later on
HKKK TMP 38E050
13
© Markku Stenborg 2005
How to define markets? — SSNIP
• Read Market Definition Guidelines
• Market is something that can be monopolized
• If it can’t be monopolized, important competitive pressures
out of ”candidate market”
– Take a small set of substitute goods and geographic area
– All produced by hypothetical monopoly
– Incentive to permanently increase prices by 5-10 %?
– Yes: Candidate market = relevant market
• Proceed to analyze market power etc
– No: Candidate market < relevant market
• Include more goods or geographic areas and repeat
• Logic: Goods on relevant market create intense competition
to each other
HKKK TMP 38E050
14
© Markku Stenborg 2005
to defineismarkets?
— SSNIP
– OnceHow
this competition
removed, incentive
to increase
price
– If strong competition remains, price increase is not
possible, and goods do not constitute relevant market
– Leave out significant constraints on market power,
candidate market is too small
– Keep in firms and products that are not significant
constraints, market is too large
• Price increase:
– Consumers substitute away
– Outside producers increase output or enter
• SSNIP asks: how much demand shifts away for a price
increase?
– SSNIP in economic jargon: What is demand elasticity for
this set of goods?
HKKK TMP 38E050
15
© Markku Stenborg 2005
• Residual
demand
curve is the
demand curve
faced by an
How
to define
markets?
— SSNIP
individual firm
– Residual demand = total market demand curve - supply
of all other firms in market
– Residual demand curve incorporates effects of changes in
prices of other products in response to changes in given
product’s price
– Residual demand is tool to evaluate relevant market as it
is relatively easy to estimate
• Homework: How do you derive residual demand?
HKKK TMP 38E050
16
© Markku Stenborg 2005
• Marshallian demand based on ceteris paribus assumption and
How to define markets? — SSNIP
measures effect of price change by keeping all other prices
constant
– Merger Guidelines assume that “the terms of sale of all
other products are held constant” = Marshallian demand
– Direct demands are hard to estimate
• Suppose condidate for relevant market has n goods
and their demand depend on each others prices
• Need to estimate at least n2 parameters to get any info
on Marshallian demand
• Price change in some goods do not leave all other
prices constant
• Prices and quantities are determined jointly in equilibrium
– Can one identify demand and supply?
HKKK TMP 38E050
17
© Markku Stenborg 2005
Critical Elasticity of Demand and Critical Loss
• SSNIP-test should be applied by estimating own elasticity of
demand
• What value of elasticity is large enough for concluding that
given set of goods comprise relevant market?
• P0 = Current price
• P1 = P0 plus some specified price increase t
• C = Short run marginal cost
• L = Current price-cost margin or Lerner-index:
(P0 – C)/P0 = 1 – (C/P0)
• T = Minimum price increase deemed significant (.05 or .1)
T = (P1 – P0)/P0 = (P1/P0) – 1
• e(P) = (dQ/Q)/(dP/P) elasticity of demand
HKKK TMP 38E050
18
© Markku Stenborg 2005
• AssumeHow
C is constant,
profits
are then (P-C)Q
to define
markets?
— SSNIP
• For profitable price increase, ex post profits must at least
equal profits from selling more at lower price
• Break-even condition is
Q(P0)(P0 – C) = Q(P1)(P1 – C)
where P1 = break-even price
• Rearranging: Q(P1)/Q(P0) = (P0 – C)/(P1 – C)
• Using definitions of T and L:
[(P0 – C)/P0]/[(P1 – C)/P1)] = L/(L+T)
• For linear demand Q = (A - P)/B:
Q(P1)/Q(P0) = (A - P1)/(A – P0)
= 1 – [(P1 – P0)/P0][P0/(a – P0)]
• Recall elasticity of demand here is e(P0) = P0/(A - P0), which
gives Q(P1)/Q(P0) = 1 - Le(P0)
HKKK TMP 38E050
19
© Markku Stenborg 2005
How
to define markets? — SSNIP
• Break-even
requires
Q(P1)/Q(P0) = L/(L+T)
so this gives us
L/(M+T) = 1 – Te(P0)
and solving gives us critical elasticity
e(P0) = 1/(L+T)
• When demand is isoelastic, break-even elasticity is
e(P0)[log(L+T) – log(L)]/log(1+T)
• Critical sales loss for a price increase = proportionate
decrease in quantity sold as a result of the price increase
large enough to make price increase unprofitable
• Sales loss resulting from price P0  P1 is
1 - Q(P1)/Q(P0)
HKKK TMP 38E050
20
© Markku Stenborg 2005
How
to define
markets?
— SSNIP
• For linear
demand
Q = (A–P)/B
we can write
this as
1 – Q(P1)/Q(P0) = 1 – (A – P1)/(A – P0)
= [(P1 – P0)/P0][P0/(A – P0)] = Te(P0)
• Applying break-even value of e(P0) derived above gives value
for break even critical sales loss Y = T/(L+T)
• If actual sales-loss after price increase is less than Y, it is
profitable to increase prices
• The break-even value of the critical sales loss is the same for
both linear and isoelastic demand curves
• Relationship between market power index L, and critical e
and Y for 5 % price increase:
L%
e
Y%
50 1.82
9.1
40 2.22
11.1
30 2.86
14.3
20 4.00
20.0
10 6.67
33.3
HKKK TMP 38E050
21
© Markku Stenborg 2005
Cross-Price Elasticity
• Sometimes in case law market definition is based on crossprice elasticity of demand
• Cross price elasticity of demand = (dQi/dPj)/(Qi/Pj), where Qi
and Pi denote the quantity and price of products i and j
• Cross price elasticity = How demand for good i reacts to
price increase of good j?
• Sounds like nice idea to delineate markets: goods belong to
same relevant market if they are good-enough substitutes
• i and j are on same market if cross-price elasticity is large
enough and otherwise are on different markets
• Cross price elasticity is not a good measure for market
delineation
• Market delineation is not question of how much demand will
flow from i to j as Pi increases
HKKK TMP 38E050
22
© Markku Stenborg 2005
• Cross-price
elasticity
is notmarkets?
usually symmetric,
eij  eji
How
to define
— SSNIP
• eij: ”i and j on same mkt” and eij: ”i and j on different mkt” is
possible
• Even if cross price elasticity is small, market need not be
narrow, as there may be many other goods that restrict the
market power of hypothetical monopolist
• If there are many good substitutes, price increase will divert
demand to many goods
• Then cross price elasticity must be small
• Price and cross price elasticities are connected:
• Price elasticity = 1 + weighted average of all cross price
elasticities
• Weight: share of income in relation to share of income of the
good in question
HKKK TMP 38E050
23
© Markku Stenborg 2005
”Cellophane Fallacy”
• US Supreme Court: high cross price elasticy between
cellophane and paper wrapping  relevant market wider
than cellophane  Du Pont not dominant
– Also SSNIP test ignores fact that firm may already have
market power
• Firm with market power wants to increase price to level
where competitive constraints start to bite
– Demand usually turns more elastic as price increases
• Then goods actually outside of relevant market seem to be
substitutes
• High cross price elasticy indication of use of market power,
not an indication of wide market
• Cellophane fallacy means that a different approach is
required in abuse of dominance cases
HKKK TMP 38E050
24
© Markku Stenborg 2005
3. Market Power
• Market power = ability to profitably charge P > MC
• Sources of market power
– Only few firms active in the market
– Products are differentiated, and some customers prefer
one firm’s product to other
• Firm that attempts to ”steal” customers from its
competitor must reduce price a lot
• Firms have less incentives to lower their prices
– Capacity constraints
• Firm have less incentive to win more customers
– Customers are not informed of all firms’ prices
• Incentive to lower price is reduced
– Switching costs
– 38E050
Cartel or collusion (later in Oligopoly section)
HKKK TMP
© Markku Stenborg 2005
25
• Market How
power to
is source
of inefficiency
define
markets? — SSNIP
– Allocative inefficiency
• Harberger triangle
• Rent seeking
– X-inefficiency
• Less need to control costs or to concentrate on key
capabilities
• Less need to provide value to customers
– Dynamic inefficiency
• Less incentive to innovate
• Market power allows restrictions on competition
– Entry deterrence
– Predation
– Price squeeze
– Cartel or collusion
HKKK TMP 38E050
26
© Markku Stenborg 2005
Monopoly’s
profit
How
to maximization
define markets? — SSNIP
• Assuming constant MC, profit  = [P(Q) - C] Q - F
• To maximize profits, set d/dQ = 0:
d/dQ = P(Q) + Q dP/dQ - C = 0 
P*(Q) - C = -Q dP/dQ
• Divide both sides by P*
(P* - C)/P* = -(Q/P*)(dP/dQ)
• Rewrite this as L = 1/e
• L = Lerner Index, e = elasticity of market demand
• Under perfect competition or perfectly elastic demand:
P = C, hence L = 0
• L is measure of market power: 0 < L < 1/e
• If we can estimate P, e and C, we can estimate market
power L
– Not practical
HKKK TMP 38E050
27
© Markku Stenborg 2005
Estimation of market power
Read Motta Ch 3, browse Nevo (2001) and Slade (2002)
Iwata-Breshanan-Lau method
• General idea: firms reveal market power through market
behavior
• In equilibrium, MC = Perceived Marginal Revenue, or
– Perfect competition: P = MC
– Monopoly:
P = MC + Q dP/dQ
– In general:
P = MC +  Q dP/dQ
• Then  seems to measure market power: 0 <  < 1
– Suppose MC stable and demand fluctuates, can we
identify  from observed data?
•  = (P-MC)/(Q dP/dQ)
• In general: no
HKKK TMP 38E050
28
© Markku Stenborg 2005
•
•
•
•
•
•
– DrawHow
picture
monopoly
and assume
observed
tofor
define
markets?
— only
SSNIP
data is known: problem
Only when demand curve rotates can  be identified
Firm i chooses qi to maximize P(Q,Z)qi – C(qi,.), where P is
market or average price and Z vector of variables that affect
demand
FOC: pi = MCi – [(Q/qi)(qi/Q)(P/Q)Q]
Term (Q/qi)(qi/Q) is Conjectural Variation coefficient
– How average firm expects market output will react to
changes on its output
• This is comparative statics, not dynamics
– FOC represents firms supply relationship, not supply curve
Suppose demand Q = a0 + a1P + a2ZP + v, v is error term
Substituting into FOC yields
(1) pi = MCi – P/(a1 + a2Z) + u
HKKK TMP 38E050
29
© Markku Stenborg 2005
• (1) canHow
be estimated:
estimate
market —
demand,
multiply P
to define
markets?
SSNIP
by a1 + a2Z, and regress
• This works for homogenous goods, eg bank loans
• To measure market power with differentiated goods, we need
to estimate demand consistently
• Basic idea: how well other goods substitute for goods of firm
i and constrain her market power?
• Answer: elasticity of residual demand
– Residual demand does not tell who or what constrains
market power
• Straight-forward approach is to specify system of demand
equations q = D(p;r), where q is J-vector of quantities
demanded from J commodities, p is a J-vector of prices, and
r is a vector of exogenous variables that shift demand
• Need to define D(.) in a way that is both flexible and
consistent with economic theory
HKKK TMP 38E050
30
© Markku Stenborg 2005
Aside: Residual Demand
of for
market
power
• The inverseEstimation
demand function
firm 1 is
(1)
P1 = P1(Q1,Q-1,Y;)
where P1 and Q1 are price and quantity for firm 1, Q-1 is
vector of other firms’ quantities, Y is vector of demand
shifters and  is vector of parameters
• Inverse demand functions for all other relevant products are
given by
Pj = Pj(Q1,Q-1,Y;)
– Notation continues to treat product 1 asymmetrically.
• Supply behavior of other firms are marginal cost = perceived
marginal revenue
(2)
(3)
MC-1(Q,W,W-1,-1) = PMR-1(Q1,Q-1,Y;,-1), where
HKKK TMP 38E050
31
© Markku Stenborg 2005
P 

 j  Qi
 market power
Estimation
 of
PMR j  Pj   


i
 Qi  Q j 
–
–
–
–
•
W is vector of industry-wide factor prices
Wj is vector of firm-specific factor prices
j are cost function parameters
j are conduct parameters describing conjectures ∂Qi/∂Qj,
ie expectations how rivals react to changes in output
Solve (3) and (2) for Q-1 and P holding Q1 fixed; solution is
(4)
Q-1 = F(Q1,Y,W,Wj; ,j,j)
• Next substitute (4) into (1) to get the residual demand curve
facing firm 1 and simplify:
P1 = R(Q1,Y,W,Wj; ,j,j)
– This is what we want to estimate
HKKK TMP 38E050
32
© Markku Stenborg 2005
Modeling Demand
Estimation of market power
• Traditionally: Linear Expenditure, Rotterdam, Translog
model, and Almost Ideal Demand System (AIDS)
• Problem 1: Number of parameters estimated increases with
square number of products: 20 firms with 20 brands each 
40 000 elasticities with straigth approach
• Problem 2: Multicollinearity of prices and need for an
instrumental variable for each of them
– Equilibrium price and quantity determined jointly by
demand and supply schedule
– Price increase by i is followed by rivals  need to take
into account how rivals react  residual, not market
demand
• Problem 3: Simple direct approach ignores heterogeneity
among consumers
– Next we turn to potential solutions
HKKK TMP 38E050
33
© Markku Stenborg 2005
1. Avoid problem
Estimation
of market power
– Focus on
aggregate demand
• Bank loans
– Focus on narrowly defined product
• Self service 95 octane
– Focus on sub-markets
• Particular segment in beer industry
• This is enough in some cases
2. Symmetric representative consumer
• Constant Elasticity of Substitution (CES) utility function
1/ r

r
U(q)    qi 
 i 1 
J
where r is a constant that measures substitution across
products
HKKK TMP 38E050
34
© Markku Stenborg 2005
• Demand of representative consumer obtained from CES is
Estimation of market power
qk 
pk1 /(1 r )
J
p
r /(1 r )
I
i
i 1
•
•
•
•
where I is income of representative consumer
Dimensionality problem is solved by imposing symmetry
between products
Estimation involves a single parameter, regardless of number
of products, and can be achieved using simple (non-linear)
estimation methods
Cross-price elasticities are restricted to be equal, regardless
of how “close” the products are in some attribute space
This restriction can have important implications and in many
cases would lead to wrong conclusions
HKKK TMP 38E050
35
© Markku Stenborg 2005
• Symmetry condition is restrictive and for this model implies
Estimation of market power
qi p j qk p j

, for all i, j, k
p j qi
p j qk
Alternative CES specification
U(q) 
J
J
 q   q
j
j 1
j
j
ln q j
j 1
yields Logit demand (more below)
– First term suggests consumer will consume only good with
highest j
– Second is entropy term and expresses variety-seeking
behavior
• Estimation of this model involves J, not J2, parameters and
allows for somewhat richer substitution patterns
HKKK TMP 38E050
36
© Markku Stenborg 2005
• Substitution patterns in Logit model are function of market
oftomarket
power
shares only,Estimation
and not related
product characteristics
• Market share here equivalent to quantities consumed by
representative consumer
• If price of good i increases, consumer is assumed to keep
same ratio qj/qk instead of consuming relatively more of
products that are similar to product i
• Models impose symmetry conditions which implicitly suggest
extreme form of ”non-local” (in attribute space) competition
• For some industries or cases this model of differentiation is
adequate, for most markets this is not
3. Separable utility and multi-stage budgeting
• Divide products into smaller groups and allow for flexible
functional form within each group
HKKK TMP 38E050
37
© Markku Stenborg 2005
• Justification: Separable preferences & multi-stage budgeting
• (Additive) Separable
preferences:
Estimation
of market power
U(q1,q2, ...,qJ) = v1(q1,q2) + v2(q3,q4) +…+ vG(qJ’,...,qJ)
where v1, ..., vG are sub-utility functions associated with
separate groups
– Groups could be broad categories such as food and wine,
and each group can be divided into more sub-groups
• Multi-stage budgeting
– Consumer allocates expenditure in stages
• at highest stage expenditure is allocated to broad
groups (food, housing, clothing, transportation…)
• at lower stages group expenditure is allocated to subgroups (beer, bread, cheese, ..) …
• until expenditures are allocated to individual products
(beer i)
HKKK TMP 38E050
38
© Markku Stenborg 2005
– At each stage, allocation decision is function of only that
group total
expenditureof
and
prices ofpower
commodities in that
Estimation
market
group (or price indexes for the sub-groupings)
– Cross elasticities between Opel sedan and VW sedan,
between Opel van and VW van, and between sedan and
van categories, but not between Opel sedan and VW van
– Reduces number of parameters to be estimated
• Three stage system
– Top level: overall demand for the product (cars or beer)
– Middle level: demand for different segments (sedan, suv,
stw, minivan; or lager, ale, stout, …)
– Bottom level: brand demand corresponding to competition
between different brands within each segment
• Typical application has AIDS at brand level: demand for beer i
within segment g in city c at quarter t is
HKKK TMP 38E050
39
© Markku Stenborg 2005
sjct = jc + j log(ygct/Pgtc) + kjk log pkct
market
power ygct is overall
where sjct isEstimation
share of total of
segment
expenditure,
per capita segment expenditure, Pgct is price index and pkct is
price of the kth brand in city c at quarter t
• Middle level of demand captures allocation between segments
of beer
– Often lso modeled using AIDS model
– Then demand equation above is used with expenditure
shares and prices aggregated to segment level
• Top level demand for whole category is
log qqct = 0 + 1log yct + 2pct + Zct
where qct is overall consumption of beer in city c at quarter t,
yct is real income, pct is price index for cereal and Zct are
variables that shift demand (eg demographics)
HKKK TMP 38E050
40
© Markku Stenborg 2005
4. Discrete Choice Models
Estimation
power
• Model products
as bundlesof
of market
characteristics
– sweetness, fiber content, …
– alcohol content, bitterness, ...
– horsepower, length, ...
• Preferences are defined over characteristics space
• Each consumers chooses the product with best characteristics
for her
– use bus if U(bus) > U(car), U(train), U(walk), ...
• Discrete choice models yield Logit demands (under some
assumptions)
– prob that agent n chooses i has logistic distribution
• Dimension of characteristics relevant dimension for empirical
work, not number of brands
• Heterogeneity is modeled and estimated explicitly
•HKKKCan
be estimated using individual or aggregate market data
TMP 38E050
41
© Markku Stenborg 2005
Comparison
of market
power
• Symmetric Estimation
average consumer
models least
adequate for
modeling demand for differentiated products
– Problem: all goods are assumed to be equally good
substitutes
• Logit models widely used because they are simple
• Multi-level model requires a priori segmentation of market
into relatively small groups, which might be hard to define
• AIDS assumes all consumers consume all products
– For broad categories like food and shelter reasonable
– For differentiated products, it is unlikely that all consumers
consume all varieties
• AIDS is closer to classical estimation methods and neoclassical theory, and more intuitive to understand
• Discrete choice models require characteristics of products, are
more technical to use, and rely on distributional assumptions
functional forms
HKKKand
TMP 38E050
42
© Markku Stenborg 2005
• Nevo 2001 uses discrete choice models succesfully
market
powerof cereal in 65
– Panel of Estimation
quantities and of
prices
for 25 brands
U.S. cities over 20 quarters, using scanner data
– Estimate own price and cross-price demand elasticities
– Compute price-cost margins implied by three industry
structures: each brand on its own, actual structure of few
multi-product firms, and monopoly or collusion
– Markups implied by current industry structure and
imperfect competition match observed price-cost margins
– High margins due to consumers' willingness to pay for
favorite brand, and to pricing decisions that take into
account substitution between own brands
– Margins not due to lack of price competition nor collusion
– Market power entirely due to the firms' ability to maintain
portfolio of differentiated products and influence perceived
product quality through advertising
HKKK TMP 38E050
43
© Markku Stenborg 2005