Transcript Slides - Competition Policy International

ANTITRUST ECONOMICS 2013
David S. Evans
University of Chicago, Global Economics Group
Elisa Mariscal
CIDE, Global Economics Group
TOPIC 12: HORIZONTAL MERGERS AND THE ANALYSIS
OF COMPETITIVE EFFECTS
Date
Topic 12| Part 2
22 October 2013
Overview
2
Part 1
Part 2
Legal and
Economic
Background of
Mergers
Quantitative
Techniques for
Estimating Price
Effects
Merger
Screening
Mergers in TwoSided Markets
Unilateral
Effects:
Economic
Theory
Coordinated
Effects: Economic
Theory and
Evidence
3
Quantitative Techniques for
Estimating Price Effects
Methods for Estimating Price Effects of a Merger
4
Critical loss and diversion analysis and other “light” merger simulation
which requires estimates of marginal costs and demand substitution.
Natural experiments that examine what happened to prices in similar
circumstances such as in another geographic market or in another
merger.
“Heavy” merger simulation which uses econometric estimates of
demand and supply to “simulate” market before and after the
merger.
The LOVEFiLM Acquisition and Light Merger
Simulation
5
LOVEFiLM online DVD rental subscription service, which
operated in the UK, wanted to acquire in 2008 Amazon’s
online DVD rental subscription service in the UK.
Overlap product is ODR service in which customers pay a
fixed monthly fee for receiving by mail DVDs they have
selected online.
ODR is one of many channels for accessing film and TV
video content such as brick and mortar, DVD retail, Pay TV,
Video on Demand, Free TV, Internet, and Premium TV
channels.
OFT used a simple back-of-the envelope “light” merger
simulation to analyze market definition.
“Light” Simulation of Unilateral Effects
6
Measure
Implication for Price of Firm A
Gross margin of firm B
Higher margin indicates larger gain
from sale diverted to firm B
Diversion ratio for firm B when firm A
raises its price
Higher diversion ratio indicates more
sales lost pre-merger by firm A could
be kept post-merger by the
combined firm
Gross Margin Estimates
7
Company
Gross Margin
Amazon
[20-30%]
LOVEFiLM
[30-40%]
OFT’s estimate of gross margins where variable costs include retention
marketing, collection costs, costs of exchanges, customer service,
and library expenses.
Diversion Ratio Analysis
8
Diversion ratio is share of sales lost as a result of a small increase in
price obtained by substitute products offered by other firms.
Diversion ratios estimated from survey of consumers.
Consumers were asked what video service they would switch to if the
price of their existing ODR provider increased by 10%.
Postpone discussion of potential issues with survey design.
Diversion Ratio Estimates
9
Including Don’t Knows
To
Apportioning Don’t Knows
From LOVEFiLM
From
Amazon
From LOVEFiLM
From
Amazon
LOVEFiLM
-
[30-40]
-
[70-80]
Amazon
[0-10]
-
[30-40]
-
Blockbuster
[0-10]
[0-10]
[30-40]
[0-10]
Other
[0-10]
[0-10]
[0-10]
[0-10]
Don’t Know
[50-60]
[30-40]
Price Increase Estimates
10
LOVEFiLM
Price Increase
Amazon
Price Increase
Including Don’t Knows
From LOVEFiLM to Amazon
[0-10]
[0-10]
From Amazon to LOVEFiLM
Apportioning Don’t Knows
From LOVEFiLM to Amazon
From Amazon to LOVEFiLM
[0-10]
[40-50]
Comments on Use of Surveys
11
Good source of data for diversion ratios are company won-loss
reports. Often available for B2B businesses but not for B2C businesses.
For B2C businesses need to do surveys of consumers (or rely on
surveys conducted in the normal course of business) to determine
diversion.
Common survey method involves hypothetical questions (like survey
OFT relied on). Consumer answer to hypothetical question not
necessarily what they would do in the actual situation.
Reliability also depends on how survey question is posed and what
the consumer is asked to assume.
“Heavy” Merger Simulation Based on Econometric
Models
12
“Structural model” assumes shape of demand schedules (linear,
logistic, etc.), consumer decision making, product differentiation,
nature of competition (Bertrand, Cournot, etc.), costs, and other
features; obtains estimates of demand and competitive interactions
and marginal cost.
Merger simulation uses the estimated structural model to simulate the
effect of a merger of firms.
Merger simulation can also be used to model synergies and other
efficiencies of merger.
Generally estimated from historical data using sophisticated
econometric techniques.
Results are highly assumption driven.
Elasticities for Ready to Eat Cereals (Nevo 2000
Study)
13
Results suggest that individual price sensitivity
is heterogeneous. Most of the heterogeneity is
explained by demographics.
Own-price elasticities are not linear in price.
This is due to heterogeneity in price sensitivity.
Consumers who purchase different products
have different price sensitivities.
In addition, substitution patterns across brands
are driven by product characteristics.
Median and Cross-Price Elasticities of Ready to Eat
Cereals
14
K Rice
Krispies
GM
Cheerios
GM Lucky
Charms
P Grape
Nuts
Q Life
R Chex
N Shredded
Wheat
K Rice
Krispies
1.320
0.069
0.041
0.050
0.048
0.081
0.049
GM
Cheerios
0.106
1.709
0.049
0.089
0.08
0.106
0.099
GM Lucky
Charms
0.025
0.02
1.945
0.025
0.072
0.024
0.099
P Grape
Nuts
0.03
0.037
0.026
2.096
0.028
0.027
0.115
Q Life
0.033
0.028
0.149
0.032
0.103
0.031
0.02
R Chex
0.024
0.021
0.011
0.013
0.014
1.749
0.014
N Shredded
Wheat
0.018
0.024
0.009
0.07
0.015
0.017
2.268
Predicted percent change in price as a result of a
merger
15
Post and
Nabisco
GM and
Nabisco
GM and Chex
Kellogg and
Quaker Oats
GM and
Quaker Oats
P
Q
P
Q
P
Q
P
Q
P
Q
K Rice
Krispies
0.0
0.1
0.1
0.2
0.1
0.4
5.1
-4.1
0.7
2.0
GM
Cheerios
0.0
0.2
0.7
-0.9
1.1
-1.3
0.5
1.3
4.1
-3.5
GM Lucky
Charms
0.0
0.1
0.3
-0.4
0.7
-0.8
0.8
3.3
9.3
-10.6
P Grape
Nuts
1.5
-2.8
0.1
0.7
0.0
0.4
0.1
2.3
0.1
3.0
Q Life
0.0
0.1
0.0
0.3
0.1
0.5
15.5
-16.7
23.8
-25.3
R Chex
0.0
0.2
0.0
0.3
12.2
-19.0
0.0
2.1
0.1
3.4
N Shredded
Wheat
3.1
-8.6
7.5
-18.8
0.0
0.4
0.0
1.9
0.0
2.5
Percent Reduction in Marginal Cost Required for
No Change in Predicted Post-Merger Price
16
Post and
Nabisco
GM and
Nabisco
GM and
Chex
Kellogg
and
Quaker
Oats
GM and
Quaker
Oats
K Rice Krispies
0
0
0
16.5
0
GM Cheerios
0
2.1
3.4
0
12.1
GM Lucky Charms
0
0.9
1.6
0
19.2
2.6
0
0
0
0
Q Life
0
0
0
16.8
20.1
R Chex
0
0
22.1
0
0
5.1
10.4
0
0
0
P Grape Nuts
N Shredded Wheat
“Natural Experiments”
17
Basic idea is to find real-world analogies to the
world with the merger and then compare prices
and other competitive conditions to actual data on
prices and competitive conditions pre-merger.
Was there a similar change in concentration in the
part from which one can infer competitive effects?
Is it possible to compare geographic areas that look
like result “post merger” and compare to situation
“pre-merger”.
May need to use statistical methods to control for
other differences between situations so that the
comparison is “all else equal” except whether of
not there is a difference in market structure.
Staples Office Depot Merger Background
18
Office superstores provide one-stop shopping
for small businesses and home-office
customers.
By mid 1990s in US Staples, Office Depot, and
OfficeMax were leading office superstore
competitors.
Staples and Office Depot competed directly
in 40 metropolitan areas.
September 1996 Staples and Office Depot
announced plan to merge.
Evidence on Prices from Documents
19
FTC’s Econometric Estimate of Price Changes
20
TABLE 1
PX-400: SIMULATED IMPACT ON STAPLES OFFICE PRODUCTS PRICES OF ELIMINATING OFFICE DEPOT:
Staples Stores with Some Office Depot Competition
Model 1
Simulated Price Change
t-Statistic
Observations in Simulation
Sample is:
Parties Sample
Complete Sample
Unit of Observation:
Weekly/Stores
Monthly/Stores
Dependent variable is:
Parties Price Index
Recalculated Price Index
Competitor Variables:
Circle-based**
MSA-based***
Model 2
1.10%
11.19
6,896
Yes
Yes
Model 3*
Model 4*
Model 5
Model 6
Model 7
0.80%
2.90%
3.70%
4.00%
3.70%
8.60%
4.79
8.88
9.16
10.33
9.12
14.99
1,685
1,817
1,315
1,465
1,395
3,038
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
*Models 3 and 4 are based on the same regression model. Model 3 reports the simulated impact of eliminating Office Depot in markets
where either the MSA-based competition data or the Circle-based competition data indicate that a Staples store faces Office Depot
competition. Model 4 reports the simulated impact of eliminating Office Depot in markets where both the MSA-based competition data and
the Circle-based competition data indicate that a Staples Store faces Office Depot Completion.
**Variables which control for the number of Office Depot, OfficeMax, computer superstores and warehouse clubs within 5 miles, 5-10 miles,
and 10-20 miles of the Staples store.
***Variables which control for the number of Staples, Office Depot, OfficeMax, Wal-mart, Sam’s Club, Computer City, BestBuy,
Office1Superstore, Costco, BJ’s, CompUSA, Kmart and Target stores in the MSA.
Economic Evidence Not the End of the Story
21
The FTC sought to enjoin the merger and the parties decide to fight
it out in court.
The parties presented econometric evidence that rebutted the
FTC’s econometric evidence on the grounds that it failed to control
for differences in stores and markets and ignored efficiencies.
The judge ignored econometric evidence. Instead he relied on
company documents that group metropolitan areas into price
zones based on superstore competitors and on simple
comparisons.
Judge agreed to block the merger.
Postscript: Office Depot and Office Max announced merger in
early 2013 which the FTC is reported to be likely to clear
22
Mergers in Two-Sided Markets
What’s Different When Markets Are Multi-Sided?
23
Unilateral effects analysis needs to consider total price effect of
merger recognizing that one side could down and another up.
Consumer impact of merger depends on total price.
Simple merger simulation formulas for price effects are wrong (e.g.
the LOVEFiLM framework cannot be applied as is). Can be
modified but simple formulas replaced by complex and hard to
estimate ones.
Structural econometric models work so long as they are modified
to account for interdependent demand.
There is no presumption that prices will increase post merger even
ignoring usual cost efficiencies since increased demand-side
network effects can counter increased market power effects.
Non-Econometric Approaches
24
Traditional unilateral effects analysis on each side and assess
biases.
• Suppose one side is free and unlikely to change post-merger. Then
analyze impact of price change on paid side and assess whether
cross-demand effects will alter conclusion.
• Eg. if profitable to raise 3% without considering other side then less
than 3% once considered.
• Eg. if profitable to raise 10% without considering other side then
question is whether considering other side would reduce estimate
enough to allay concerns.
Simple natural experiments looking at different platform market
structure configurations in past or in other markets.
Two-Sided Econometric Analysis of Dutch
Newspapers
25
“In our case, the effects of the hypothetical merger
on subscription prices and readers’ welfare are
found to be small. Concerns mainly arise with
respect to the advertising side. Importantly to this
regard, with the exception of market concentration
analysis, there does not seem to be a significant
difference between the different methods used to
assess the unilateral effects of the hypothetical
merger we analyzed. This is because we used SSNIP
and UPP formulas adjusted for two-sided platforms,
so that only the HHI-based analysis did not take the
two-sided nature of the market into account. So, for
the example studied here, we find that commonly
used methods to assess mergers work well in twosided markets as long as one properly adjusts
them—in the way we have described above—for
the two-sided nature of the market.”
Filistrucchi, Klein, and Michielsen, “Assessing
Unilaterial Merger Effects in a Two-Sided Market: An
Application to the Dutch Daily Newspaper Industry.
Comparison of One Sided and Two-Sided
26
Analysis of HHIs shows concern on reader side. However, in
addition to market definition issues use of HHI for market power
likely to overstate market power.
Single-sided UPP finds no pressure on advertising side but two-sided
does as a result of accounting for cross demand effects. Single and
two-sided similar for reader side.
SSNIP shows higher increase in prices when firm can adjust prices
on both sides. Concern primarily on the advertising side.
Full econometric analysis shows no change in reader welfare but
higher per-reader prices per advertiser. Consistent with two-sided
SSNIP and UPP but not with HHI analysis.
27
Coordinated Effects: Economic
Theory and Evidence
Coordinated effects
28
Merger can increase the potential for coordination/tacit collusion.
• Co-ordination over prices.
• Market sharing.
Issues to consider.
• Is co-ordination feasible or likely in the relevant market?.
• Ability to coordinate.
• Ability to monitor competitors.
• Ability to discipline deviations.
• Lack of external constraints.
• Does the merger increase the risk of co-ordination?.
Ability to coordinate
29
Small number of competitors.
• Coordination more difficult when there is a large number of firms.
Homogeneous products.
• Coordination more difficult with differentiated products or wide product
ranges.
• Unless there exist focal products or prices.
Symmetric firms.
• Similar sizes and costs structures.
• More difficult to (tacitly) agree on the profit maximising price if firms
have different cost structures.
Ability to monitor competitors
30
Stable demand.
• High incentive to cheat if demand is rising.
• More difficult to monitor competitors’ behavior if demand is increasing
and output changing.
Transparent pricing.
• Makes it easier to monitor competitors behavior.
• Can be facilitated by trade associations, or third-party agencies.
• (Price transparency is often also good for consumers and can increase
competition!).
Ability to Discipline Deviations
31
There is usually a short-term cost to punishing cheaters.
• Cut prices to punish cheat, but everyone suffers (in the short run).
Profit margins.
• High margins increase the incentives to cheat and increase the cost of
punishing deviations.
Multi-market contacts.
• Cheating in one market can be punished in a different market.
• Less incentive to cheat in one market if it leads to retaliation in many
others.
Lack of External Constraints
32
Lack of substitute products/Low elasticity of demand.
Barriers to entry.
Lack of buyer power.
End of Part 2, Next Class Topic 13
33
Part 1
Part 2
Price Discrimination
and Other Complex
Pricing
Predatory Pricing
Limit Pricing
Loyalty Rebates