Regression Analysis & Market Delineation

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 30

Regression Analysis

& Market Delineation


Luke M. Froeb
Vanderbilt University &
ERSGroup.com

26 March, 2008 8:45-11:45am


Antitrust Economics & Econometrics
ABA Spring Meetings
Acknowledgements
Henry McFarland, Economists, Inc.
David Scheffman, Vanderbilt & LECG
Gregory Werden, US Dept of Justice

Vanderbilt University 2
Take-away: economists can help,
but only if you understand what
they are doing
Regression creates experiments from non-
experimental data
What else could have accounted for estimated
effect?
How well does experiment mimic effect we are
trying to isolate?
Quantitative market delineation requires
careful thought about how to apply monopoly
model 3
Click & Learn Regression
<<pull up program>>
But for regression model.
Which functional Form?
How well does it fit?

4
Bid Rigging: Frozen Fish Conspiracy

5
1976 Folding Cartons Conspiracy
DOJ investigation resulted in indictment of 23
firms
Difficult to prove conspiracy or meeting of
the minds
But ring leader was compulsive note taker
Testified in exchange for no jail time
But judge thought outcome was unfair.

6
Follow-on Damage Estimation
Forecast showed big damages
Shift of intercept AND slope
Backcast showed negative damages
What to do?
<<Click&Learn backcast vs. forecast>>

7
Merger Analysis: Staples-Office Depot
Prices in two-office-superstore cities estimated to be
7% lower than in one-office-superstore city.
15% estimated pass-through (from cost to price)
85% reduction in costs to offset merger effect
Critique:
Could unobserved costs account for relationship?
How well does experiment mimic merger effect?
Did experts cancel each other out?
<<Click&Learn dummy variable regression>>

8
Consummated Mergers
Control Group: Pre-merger period
Experimental Group: Post-merger period
Did price increase?
BIG question: Compared to what?
Control cities hit by same demand and cost shocks
Differences-in-Differences Estimation
First difference: pre- vs. post-merger
Second difference: target vs. control cities
(Marathon/Ashland Joint Venture)

Combination of marketing and refining assets


of two major refiners in Midwest
First of recent wave of petroleum mergers
January 1998
Not Challenged by Antitrust Agencies
Change in concentration from combination of
assets less than subsequent mergers that were
modified by FTC
Merger Retrospective (cont.):
Marathon/Ashland Joint Venture
Examine pricing in a region with a large change in
concentration
Change in HHI of about 800, to 2260
Isolated region
uses Reformulated Gas
Difficulty of arbitrage makes price effect possible
Prices did NOT increase relative to other regions
using similar type of gasoline
Difference Between Louisville's Retail Price and Control Cities' Retail Price

25.00

Merger Date
20.00

15.00

10.00

5.00
Cents

0.00
1/1/1997

3/1/1997

5/1/1997

7/1/1997

9/1/1997

11/1/1997

1/1/1998

3/1/1998

5/1/1998

7/1/1998

9/1/1998

11/1/1998

1/1/1999

3/1/1999

5/1/1999

7/1/1999

9/1/1999

11/1/1999
-5.00

-10.00

-15.00

-20.00

-25.00
Week

Chicago Houston Virginia


BIG Policy Question
What are ex-ante incentives created by ex-
post enforcement?
Enforcement vs. regulation?
Type I error (over-deterrence): dont raise
price, even if costs increase
Type II error (under-deterrence): wait 2 years
and then raise price
Will your merger be challenged?
Rule of thumb
Is there a benign or pro-competitive reason for
merger?
Are customers complaining?
Will merger lead to price increase?
FTC Merger Challenges,96-03

90

80

70
Number of Markets

60

50

40

30

20

10

2 to 1 3 to 2 4 to 3 5 to 4 6 to 5 7 to 6 8+ to 7+
Significant Com petitors
Enforced Closed
Whats Wrong w/Structural
Presumptions?
1. Market delineation draws bright lines even
when there may be none
No bright line between in vs. out
2. Market Shares may be poor proxies for
competitive positions of firms
Market shares and concentration may be poor
predictors of merger effects
HOWEVER: you still have to delineate
a market
Rookie mistake to bring a case without one
The Hypothetical Monopolist Test in the U.S.
Horizontal Merger Guidelines
group of products and a geographic area
such that a hypothetical profit-maximizing
firm likely would impose at least a small but
significant and nontransitory increase in price
Depends only on demand
Tests whether merger creates market power
Not designed to test whether a firm is already
exercising significant market power
(Dominance)
17
Quantitative Market Delineation
Critical Elasticity of Demand Analysis
Profit-Maximization Calculation
Breakeven Calculation*
Critical Sales Loss Analysis
Profit-Maximization Calculation
Breakeven Calculation*

--------
*covered today
18
Critical Elasticity of Demand Analysis
Breakeven Calculation: The maximum
elasticity of demand a monopolist could face
at pre-merger prices and still not experience a
net reduction in profits from a given price
increase, e.g., 5%
Depends on demand functional form
Linear: 1/(m+t)
Constant elasticity: [log(m+t)-log(m)]/log(1+t)
where m=margin, t=5%
19
Critical Sales Loss Analysis
Breakeven Calculation: The maximum
reduction a monopolist could experience in its
quantity sold and still not experience a net
reduction in its profits from a given price
increase, e.g., 5% [critical loss=t/(m+t)]

Pre- 10% 30% 50% 70% 90%


merger
margin
Critical 33% 14.2% 9.1% 6.7% 5.2%
sales loss 20
FTC v. Tenet Health Care Corp.
17 F. Supp. 2d 937 (E.D. Mo. 1998),
revd, 186 F.2d 1045 (8th Cir. 1999)
District court accepted FTCs contention that
the geographic scope of relevant market was a
50-mile radius around Poplar Bluff, Missouri.
On appeal, the defendant argued that its
critical loss analysis demonstrated that the
FTCs market was too narrow.
Eighth Circuit held that the FTC failed to show
that hospitals outside its alleged market were
not practical alternatives for many Poplar
Bluff consumers. 21
US v. Mercy Health Services
902 F. Supp. 968 (N.D. Iowa 1995),
vacated as moot, 107 F.3d 632 (8th Cir. 1997)

Relying on defendants breakeven critical loss


of 8%, the court found sufficient switching
would occur in the event of a 5% price rise
to make the price rise unprofitable.
Govt. predicted the total elimination of
managed care discountsa far larger price
increase, so the court also considered a larger
(albeit not large enough) price increase.
Court reckoned the critical loss at 2035%,
although it was actually about 46%. 22
FTC v. Swedish Match
131 F. Supp. 2d 151 (D.D.C. 2001)

Both experts relied on critical elasticity


analyses, which differed
Court discussed these analyses in detail, but
found neither experts evidence persuasive.
Court applied its own critical loss analysis,
finding that it cannot be unprofitable for the
hypothetical monopolist to raise price . . .
because the hypothetical monopolist would
lose only a small amount of business.
23
U.S. v. SunGard Data Sys., Inc.
172 F. Supp. 2d. 172 (D.D.C. 2001)

Court noted defendants contention that


margins > 90% so critical loss was very low.
Government said nothing about this analysis.
Court held that the government had failed to
show that the customers who would not
switch in the face of a price increase were
substantial enough that a hypothetical
monopolist would find it profitable to impose
such an increase in price.
24
FTC v. Whole Foods
Appeal from the United States District Court
for the District of Columbia, Civ. No. 07-cv-Ol021-PLF

XX% retail margins XX% critical loss


Defense expert inferred actual loss from
marketing studies
FTC expert inferred actual loss from store closing
experiments
If we [close the Wild Oats Store right across the street],
we believe approximately 50% of the volume their
store does will transfer to our store, with the other 50%
migrating to our other competitors (these estimates are
based on our past experience with similar situations).
Whole Foods website
25
Assessing Price-Cost Margins
Never simply use whatever the parties call
their margins; rather, get data from which
margins can be computed.
Get disaggregated revenue and cost data.
Find out exactly how the data were complied.
Treat the determination of margins as a
central task of the investigation and anticipate
the parties arguments.

26
Paradox of High Margins
A high pre-merger margin implies a low critical
elasticity and critical sales loss
Does this suggest a broad market?
In oligopoly models, a high margin implies low
actual demand elasticity and actual sales loss.
And large merger effects
Small differences in demand elasticities are
important
but may be difficult to measure precisely
27
Can Modify Monopoly Model
to Fit Industry Features
Adjust model to account for:
Different types of consumers;
monopolist may price discriminate;
prices may increase non proportionally on
different goods
Standard formulae presume constant marginal
cost and no avoidable fixed costs, but actual
cost functions may be quite different.
Profit maximizing monopoly price increase
may be much larger than 5% 28
Oligopoly
Models
Mergers Among
Parking Lots, J.
Econometrics
Capacity
constraints on
merging lots
attenuate price
effects by more
than constraints
on non-merging
lots amplify them
Bottom Line:
Advantage of Quantitative Analysis
More persuasive: Some number beats no
number
Models, natural experiments are complements,
not substitutes
Use models to interpret experiments; and
Use experiments to inform models
Clearer mapping from evidence to opinion
Sharpen focus: tells you what matters and how
much it matters
Calculation replaces intuition
30

You might also like