The Role of Econometric Analysis in Antitrust
Download
Report
Transcript The Role of Econometric Analysis in Antitrust
Quantitative Benefit-cost
Analysis of Mergers
Luke Froeb
Oct. 26, 2001
Federal Trade Commission
Vanderbilt University
1
References
mba.vanderbilt.edu/luke.froeb/papers/
Coauthors, Tschantz & Werden
Simulating Merger Effects Among Capacityconstrained Firms
Pass Through rates and the Price Effects of
Mergers
Merger Effects When Firms Compete by
Choosing Both Price and Advertising
Does retail sector matter for manufacturing
mergers? [very preliminary]
Vanderbilt University
2
Quantitative benefit-cost analysis
Goal: quantitative estimate of merger
effect.
Necessary to weigh efficiencies against loss
of competition
Two methodologies
Empirical comparisons, e.g. Staples/Office
Depot
Model-based simulations
Vanderbilt University
3
Empirical Comparisons
e.g., Staples-Office Depot
Good natural
experiments or
comparisons
Benefit-cost analysis
still requires structural
estimate of pass
through
Depends on demand
curvature
big pass-through iff big
anticompetitive effect
Vanderbilt University
4
Model-based simulation
Model current competition
Estimate model parameters
Simulate loss of competition from
merger
Vanderbilt University
5
e.g. Parking Merger
4
y
Key parameters
cost of walking
Sensitivity of
choice to price
location of
merging lots
location of nonmerging lots
capacity of lots
location of office
buildings
3
B:$1 .29
2
1
A:$1 .4
C:$1 .46
1000
z
500
0
1
2
3
Vanderbilt University
x
4
6
Simple approach:
Bertrand Price-setting game
Static game
Price-setting competition
What about product, promotion, placement?
Unilateral Effects
What about dynamic strategies?
What about coordinated effects?
Does retail sector matter?
Kroger-Winn Dixie vs. Quaker-Pepsi
Vanderbilt University
7
Simple approach:
Modeling Critique
How well does model capture loss of
competition from merger?
Coke strategy is “share of throat”
MCI-Sprint
More about placement and product than price
Tele-market new plans to rivals’ customers
More about promotion than price
Is Bertrand a good metaphor for loss of
competition?
Vanderbilt University
8
Simple approach:
Does retail sector matter?
When is retail sector transparent?
Constant or constant percentage markup
two-part tariffs, and retail sector must
carry profitable products
Retail sector earns no profit
When does it matter?
Double marginalizationprice effect
Two-part tariffs, and option of
exclusivityno price effect
Vanderbilt University
9
Simple approach:
What about advertising?
FOC’s if q=q(a,p)
FOC if q=q(a(p),p)
{0=q+(p-mc)dq/dp, 0=-1+(p-mc)dq/da}
0=q+(p-mc’)dq/dp;
mc’=mc+(da/dp)/(dq/dp)
Pre-merger: Price-only model with mc’
≈ price+advertising model
Does advertising increase with quantity?
Vanderbilt University
10
Simple approach:
Implementation
Estimate AIDS demand
Scanner data
Instruments
None needed for weekly data
Prices in other cities
LR vs. SR elasticities (Nevo & Hendel)
Correlated through costs
Results
High variance
Inelastic demand?
Goods are complements?
Vanderbilt University
11
Implementation Critique:
too many parameters
AIDS has too many parameters
Confidence intervals include both pro- and anti- scenarios.
Elasticity matrix for merging products is most important.
Alternatives: Logit, nested logit, PD GEV
(Bres.&Stern), mixed logit (BLP) + census data
(Nevo)
But all goods are substitutes
Only fool would admit post-merger price rise to FTC
Agencies discount efficiencies as not merger-specific
So parties are reluctant to admit even small price increase.
Proposal: assume 5% MC reduction
Then simulate post-merger prices
Vanderbilt University
12
PD GEV
Bresnahan & Stern
(multiple) dimensions of differentiation
PDGEVDimension 0.8
,
List prod1, prod2
prod1 Q
prod2 Q
prod3 Q
prod4 Q
COLUMN
List prod3, prod4
Implies substitution patterns
prod1 P
1.56
0.313
0.125
0.125
0.25
Vanderbilt University
prod2 P
0.313
1.56
0.125
0.125
0.25
prod3 P
0.125
0.125
1.56
0.313
0.25
prod4 P
0.125
0.125
0.313
1.56
0.25
ROW
1.
1.
1.
1.
1.
13
Implementation Critique:
Higher derivatives of demand
f(x),f’(x), and f’’(x) influence predicted price rise.
Need location, velocity, and acceleration
If we cannot estimate f’(x)
but observe only location
Product margins
Hall vs. Hausman in MCI-Sprint
If we cannot estimate f’’(x)
Sensitivity analysis; or
Use linear or logit for extrapolation to be conservative; or
compensating cost differentials don’t depend on acceleration
Vanderbilt University
14
Implementation Critique:
Average revenue instead of price
Average revenue is quantity share-weighted
price index.
Leads to inelasticity bias
Price changes cause weights to change.
Use fixed weight index when possible.
Or use disaggregated data
store-level data exist
but we don’t use them
Individual choice data exist
but we don’t use them
Vanderbilt University
15