Upload
roberto-balmer
View
169
Download
0
Embed Size (px)
Citation preview
Entry and Competition in local
Newspaper Retail Markets
When two are enough
Roberto Balmer
BAKOM
Disclaimer: The views presented here are those of the author and do not reflect those of BAKOM.
Swiss IO Day
University of Bern, 13 June 2014
2
Agenda
1. Introduction (the NEIO story, Bresnahan (1982))
2. Model (Bresnahan-Reiss (1991))
3. Variables
4. Results
1. Introduction 2. Model 3. Variables 4. Conclusions
3
Structure – Conduct – Performance
Source: Davis, Quantitive Techniques for Competition and Antitrust Analysis, Chap. 6
• The pre-Game Theory IO world was dominated by the SCP paradigm (50s)
• Causality between market structure, competition and performance (prices, profits, welfare)
• Bain (1950): Regresses profit on market structure (HHI) finding positive coefficients; interprets
as direct causality.
• SCP is obsolete. Conduct may depend on numerous factors.
• Example: A monopolist may non act like a monopolist but in the extreme case like a firm
under perfect competition in case of low barriers to entry.
• Structural parameters continue to play an important role in market analyses as indicators for
competition (e.g. number of firms, market shares, HHI). But challenge is to measure conduct.
1. Introduction 2. Model 3. Variables 4. Conclusions
4
New Empirical Industrial Organization
• SCP is «old IO», «NEIO» is coherent with game theory; Example
• Structural demand/supply model can allow to directly measure the level of
competition / conduct with sufficient data on prices, volumes, etc.
• Inverse demand: (1)
• Marginalkostenfunktion
• Supply; F.O.C. of profit maximisation problem. Parameter lamda can nest PC,
Cournot or Cartel:
F.O.C.
• Lamda=«conduct»
• Therefore supply relation: (2) where
Quelle: Bresnahan 1982. “The oligopoly solution concept is identified”
1. Introduction 2. Model 3. Variables 4. Conclusions
5
• Simultaneous equations (1) und (2)
(1)
(2)
• Identification as usual: An equation is identified if at least one variables is excluded of it. I.e.
if a variable only shifts demand (e.g. income), then supply is identified and vice versa (e.g.
input prices).
• In this case parameters can be estimated (including demand elasticities etc.). BUT:
Estimation of gamma does not allow any inference on Lamda (conduct/level of competition).
The Solution: Rotation
1) When Beta1 is known (i.e. the cost function), Gamma can be calculated. This is also the
case when MC are assumed to be constant (Beta1=0).
2) If next to cost and demand „shifters“ and demand „rotators“ (Z) are available (i.e. Variables
that change both the level as well as the slope of the demand function), estimation of Gamma
becomes possible in any case.
New Demand (1’)
New Supply includes now
interaction term (2’)
-Tests can now tell which form of competition / conduct best explains the data. E,g, econometric
test for Lamda=0 becomes possible (perfect competition)
- Such regressions are still rare in NRA/NCA practice, but approach is promising
New Empirical Industrial Organization
1. Introduction 2. Model 3. Variables 4. Conclusions
6
Newspaper and magazine publishers
Wholesaler 1 (e.g. Valora)
Subscribers
Distributor (e.g. Post)
R1 R2 R3 R4
Consumers
x
Advertising customers
Introduction: Local newspaper retail markets
1. Introduction 2. Model 3. Variables 4. Conclusions
• Motivation:
Show how to “measure” competition
in composite goods markets,
i.e. without individual price and volume data
• Local in-depth news in rural areas
• Newspaper selling points
• Wholesale: Valora is largely a monopolist: exclusive contracts with publishers,
delivery before 9 a.m. means high entry barriers (Post services 12 a.m.)
• Retail: Market “radius” may be limited. E.g. for basic shopping needs 10 minutes of
travel time (Valora/Cevanova). Lower for newspapers. 73% of communes have 1+
selling point
• Geo. Market = small communes.
• Retail: Subscription channel targets different customers (habitual readers)
7
• Largest newspaper retailers 2008: Valora (20%), Volg (11%), Coop (10%), Post (7%)
• All retailers should have non-discriminatory access to newspapers by Valora (in Valora-
Melisa the CompCom announces that it would otherwise intervene)
• Positive externality (foot traffic): Possible that even with unprofitable standalone
business continue to sell newspapers is reasonable
> Consider a virtual, composite good including complements such as food & near-food
items (any other goods of daily use).
• Coverage lowers entry barriers for local newspaper publishers (“experience good”).
• TV, radio, Internet content and Commuters dailies’ care not considered substitutes in
2004 (different range & depth of information, accessibility, habits)
• Way to quantify effects of entry on Competition & Competition–Coverage trade-off?
Introduction: Local newspaper retail markets
1. Introduction 2. Model 3. Variables 4. Conclusions
8
Bresnahan-Reiss 1991 Model (1/2)
• Broad range of products, impossible to take all individual prices & quantities into account
• NEIO: Estimate “conduct”. A monopolist does not necessarily show monopolist conduct
• BR1991 “Entry threshold” model: Estimate minimum demand necessary for N firms to enter
and development of market power with entry using only
1) Number of firms per commune (N)
2) Demand and supply shifters (Z, W) - income / real estate availability
3) Demand rotators: Market size S(Y)
• Market size S rotates per firm demand outwards; d(Z,P) being representative customer demand:
• Outward total demand rotation increases profits in the market but also profits of potential
entrants (market is split) up to entry of second firm (competition increases). Relate rotation of per
firm market demand S to the equilibrium number of firms (N(S)), N-th firm entry condition
-D1: Minimal per firm demand level that can sustain
profitable entry (P=AC) of a first entrant
-Q1: Level of per firm demand allowing entry of first
firm
-P1: Price of the virtual good at Q1
-DN: Level of per firm demand allowing entry of N
firms; margins decrease with competition.
-D∞: Further increase of per firm demand impossible
– MES: here any number of firms is sustainable.
Margins are 0.
P for newspapers ofter front printed and MC could be
decreasing for newspapers alone
)(),( YSPZdQd
1. Introduction 2. Model 3. Variables 4. Conclusions
9
Bresnahan-Reiss 1991 Model (2/2)
Formalisation• Minimum profit for N firms to enter
• Minimal total market size required for N-th firm to enter (“entry threshold”;
“Trigger” market size):
• Minimal market size per firm (“per firm entry threshold”) required for N-th
firm to enter:
• Idea: Define the “(Per firm) Entry threshold ratio”
• Assuming equal fixed costs, the entry threshold gives the fall in variable
profits with entry (from N to N+1 firms). A scale free measure of the
evolution of competition with entry
• Entry threshold ratio always above 1. A subsequent entrant needs more (at
best same) demand to break even as profitability per consumer decreases
with entry due to competition (also fixed costs increase with entry as
opportunities are rare)
• But: 1.00 may mean perfect competition or cartel!
N
Ss N
N
),(),(
)(
NNN
NPZdWqAVCP
WFs
customers per firm split symmetrically, so
0)(),(
),()( WFN
SPZdWqAVCPS NN
NNNN
N
PZdWqAVCP
WFS
NNN
N ),(),(
)(
1),(),(
),(),(
)(
)(/
111
11
NNN
NNN
N
NNN
PZdWqAVCP
PZdWqAVCP
WF
WFss
1. Introduction 2. Model 3. Variables 4. Conclusions
10
• Number of sellers in a commune N (selling «Blick»)
• Market size of a commune:
where Y0 is population and Y1 is inbound commuters. =1 for
normalization
• Fixed costs of entrants with N firms
where W variables includes the available agricultural land per capita
(indicator for low real estate prices). 𝛄s expected to be positive
(opportunities are rare)
• Variable profits per customer in a market with N firms
where the X variables include income in the commune and socio-
demographic variables such as the number of aged people in the commune,
the average number of schooling years and foreign resident population. αs
expected to be positive and decreasing with N
Latent variable:
Variables and Market model 1/2
N
n
nN XV2
1ˆˆˆ
0
),(),,,(),( WFWZVYS NNNN
1100ˆ),( YYYS
N
n
nLN WF2
0ˆˆˆ
1. Introduction 2. Model 3. Variables 4. Conclusions
11
Variables and Market model 2/2
Estimation procedure
• N=0, 1, 2 is a ranking only: Ordered choice dependent variable, “ordered probit”
estimation (based on cumulative normal Φ).
• Likelihood, maximise product of probabilities,
• Maximising log likelihoods which is equivalent
• Maximise log likelihood
)(1)Pr(1)Pr()Pr()0Pr()0Pr()0Pr( 111111 N
)()(
)Pr()Pr()00Pr()1Pr(
21
212121
andandandN
)()Pr()Pr()0Pr()2Pr( 2222 N
1042
1
2
1042
1
12
1042
1
1 )(1)2(1)()()1(1))()0(1m
m
m
m
m
m nnnL
1042
1
22
1042
1
1122
11
1042
1
)),(),,,(),((1ln)2(1
)),(),,,(),(()),(),,,(),((ln)1(1
))),(),,,(),(((ln)0(1ln
i
m
i
m
i
m
WFWZVYSn
WFWZVYSWFWZVYSn
WFWZVYSnL
1. Introduction 2. Model 3. Variables 4. Conclusions
12
Estimates
• Problem? Nonlinear Oprobit. Needs
dedicated programming
• Most coefficients as expected and
significant
• α, γ, λ all as expected
• 1 commuter counts 0.78 of a
resident in terms of market size
β:
- Proportion of elderly and foreign
people boost demand
- Income negative (richer communes
have higher subscribership?)
- Education insignificant
γ: land availability impacts fixed costs
negatively as expected
1. Introduction 2. Model 3. Variables 4. Conclusions
13
Entry thresholds
Per firm entry threshold (market size)
strongly increases from N=1 to N=2 (ratio
1.9), i.e. strong competitive effect
Equality test (Wald tests for proportionality)
> s1 is “different”
> “Two are enough”
S1 S2 S3 S4 S5
Newspaperretailers, S
482 1,841 3,012 4,123 5,512
Newspaper retailers(state dep. sample
means), S*490 1,864 3,066 4,154 5,156
01,0002,0003,0004,0005,0006,000
s1 s2 s3 s4 s5
Newspaperretailers, s
482 921 1,004 1,031 1,102
Newspaper retailers(state dep. sample
means), s*490 932 1,022 1,039 1,031
0200400600800
1,0001,200
s2/s1 s3/s2 s4/s3 s5/s4
Newspaper retailers,S
1.91 1.09 1.03 1.07
Newspaper retailers(state dep. sample
means), S*1.90 1.10 1.02 0.99
0.00
0.50
1.00
1.50
2.00
2.50
1. Introduction 2. Model 3. Variables 4. Conclusions
14
Entry threshold ratios - Comparison
- Similar to U.S. Doctors
- Common that second entrant has strong competitive effect
- Most other estimates find that the 3rd entrant has a stronger effect on
competition
- On this basis, AVG2007 used results to suggest to block 3-to-2 local
hospital mergers
- Not the case here. “Two are enough”. May not apply to other retail
markets
0
0.5
1
1.5
2
2.5
s2/s1 s3/s2 s4/s3 s5/s4
Druggists (BR1991)
Plumbers (BR1991)
Dentists (BR1991)
Hospitals (AGV2007)
Tire dealers (BR1991)
Doctors (BR1991)
Newspaper retailers (Balmer 2013)
1. Introduction 2. Model 3. Variables 4. Conclusions
15
Policy implications 1/2
a) Only way to make inferences on conduct in this market
b) Competition much stronger with 2 or more players in
market
c) Coverage: “If a monopolist in a small commune would earn
only the variable profits a duopolist would earn”, 263
Communes with market size between 482 and 921 would no
longer be able to sustain any selling point (2.1% of
population).
Competition – Coverage trade-off. Welfare implications of
increased (potential) competition or regulation then unclear.
d) Optimal that 2nd entrant has strong competitive effect, as
then coverage is already sustained (no differentiation)
1. Introduction 2. Model 3. Variables 4. Conclusions
16
Policy implications 2/2
e) Direct public intervention: Government could instruct the
Swiss Post to sell local newspapers (again) in some post
offices. Ideally access to post sales infrastructure would be
granted at cost-oriented prices to editors.
> Only appropriate where competition is not distorted. In
communes representing market sizes below 482, where
entry would not be economically viable, direct market
intervention unproblematic (eventually also with N=1).
> Analysis shows that in past, intervention often in other
Communes risking distortion of competition and
investment.
Limits of the model:
- No travel costs
- Independence between communes
1. Introduction 2. Model 3. Variables 4. Conclusions
17
Questions?
Roberto Balmer
Bundesamt für Kommunikation
TC / Sektion Ökonomie
Zukunftstr. 44
2501 Biel
Tel. +41 32 327 56 43
linkedin.com/in/RobertoBalmer
slideshare.net/RobertoBalmer
amazon.com/author/roberto.balmer
ssrn.com/author=572707