1

1	UNITED STATES FEDERAL TRADE COMMISSION
2	and
3	UNITED STATES DEPARTMENT OF JUSTICE
4
5
6
7	SHERMAN ACT SECTION 2 JOINT HEARING
8	UNDERSTANDING SINGLE-FIRM BEHAVIOR:
9	EMPIRICAL PERSPECTIVES SESSION
10	TUESDAY, SEPTEMBER 26, 2006
11
12
13
14
15	HELD AT:
16	UNITED STATES FEDERAL TRADE COMMISSION
17	SATELLITE BUILDING, CONFERENCE ROOM C
18	601 NEW JERSEY AVENUE, N.W.
19	WASHINGTON, D.C.
20	9:00 A.M. TO 12:30 P.M.
21
22
23
24	Reported and transcribed by:
25	Susanne Bergling, RMR-CLR

2

1	MODERATORS:
2	WILLIAM A. KOVACIC
3	Commissioner
4	Federal Trade Commission
5	and
6	KENNETH HEYER
7	Acting Deputy Assistant Attorney General
8	for Economic Analysis
9	Antitrust Division, U.S. Department of Justice
10
11	PANELISTS:
12
13	Jonathan B. Baker
14	Luke M. Froeb
15	Robert C. Marshall
16	Wallace Mullin
17	David Reitman
18	F. Michael Scherer
19	Clifford Winston
20
21
22
23
24
25

3

1	C O N T E N T S
2
3	Introduction
4
5	Presentations:
6	Jonathan B. Baker
7	Luke M. Froeb
8	Robert C. Marshall
9	Wallace Mullin
10	David Reitman
11	F. Michael Scherer
12	Clifford Winston
13
14	Moderated Discussion
15
16	Conclusion
17
18
19
20
21
22
23
24
25

4

1	P R O C E E D I N G S
2	- - - - -
3	DR. HEYER: Okay, first, it's a pleasure to be
4	here, and since you're probably less interested in what
5	I have to say than what these people have to say, I am
6	going to be brief before turning things over to Bill.
7	I wanted primarily to thank some people, not
8	only the panelists for giving us their time and soon
9	sharing their insights with us, but I wanted to thank
10	particular people at the Antitrust Division who have
11	helped prepare this and helped prepare me.
12	We have some people from the Legal Policy
13	Section in the Antitrust Division, Deputy Chief Gail
14	Kursh, who in an earlier life helped manage the Dentsply
15	case, which you will hear more about from Dr. Reitman
16	over there. One of the attorneys in her section, Joe
17	Matelis, crackerjack paralegal Brandon Greenland, and
18	most importantly, June Lee, one of the economists in the
19	Division, who, in addition to putting up with all the
20	administrative stuff, has actually contributed
21	substantively.
22	So, with nothing further, I am going to turn it
23	over to my distinguished colleague and co-moderator,
24	Bill Kovacic.
25	COMMISSIONER KOVACIC: Welcome to the New Jersey

5

1	Avenue Conference Facility on September 26th, the 92nd
2	Anniversary of the adoption of the Federal Trade
3	Commission Act. We're delighted to have you all here
4	today and to focus on what I think is one important
5	dimension of the assessment of what standards for
6	unilateral firm behavior ought to be. Many of the
7	presumptions that run throughout discussions of doctrine
8	and policy involving the enforcement of competition law
9	against dominant firms derive from empirical judgments
10	about the state of the world. To read judicial opinions
11	and see how often the opinions say "we know, it is
12	believed, it is thought, the world is," and then to look
13	futilely in the footnotes for what editors in journals
14	would note and say "Add cite," is a striking phenomenon.
15	More than that, when you take a look at the
16	papers of some of the Justices of the Supreme Court,
17	papers that have become available, you see how
18	frequently in their deliberations they're relying upon
19	hunches, judgments or assessments about the state of the
20	world and the way in which business behavior has been
21	used in the past, and about the significance of that
22	behavior. It's impossible, in short, in looking at the
23	full range of history and enforcement policy and
24	judicial decision-making, to escape the significant role
25	that assumptions about the state of the world play in

6

1	the formulation of doctrine.
2	Our aim today is to address three questions and
3	to try to link empirical work that's been done or might
4	be done in the future to the development of standards.
5	Three questions really animate our session today.
6	The first is to consider what past empirical
7	work tells us about how firms become and remain
8	dominant, to look back and, at least selectively, to
9	take a look at what work has been done by empirical
10	researchers, whether in the form of quantitative work,
11	whether in the form of case studies, whether simply in
12	the examination of the way in which judicial decisions
13	or enforcement decisions have affected the way firms
14	behave.
15	Second, and more forward-looking, is to ask what
16	we would like to learn if we could, what additional
17	facts would we like to have if we could get them in
18	principle.
19	And last, based upon what we offer as an answer
20	to the second question, how might we go about doing it?
21	What combination of effort within public enforcement
22	agencies, among think tanks, academic research centers
23	or other bodies, might provide the means by which
24	important empirical questions could be answered?
25	Later today, as Ken has, I will acknowledge the

7

1	many contributions of our professional staff that have
2	made the event possible. For now, to begin, I just want
3	to remind you of a couple of housekeeping details about
4	the session.
5	The first is to respect our speakers by turning
6	off all of your communication devices. I was at a
7	hearing a couple of years ago in the federal courthouse
8	where the bailiff stood up and said, "If your
9	Blackberries or cell phones go off, you will be
10	removed." We won't remove you, but please do honor this
11	convention.
12	Second, those of you who want to make your way
13	to the restrooms, they are through the lobby -- the
14	signs are marked -- between the elevators and off to the
15	right. Now and then, there are planned or unplanned
16	fire drills and alarms. If one goes off, we and our
17	staff will lead you out to the street, to the right,
18	back through the lobby, and we will simply gather out in
19	front of the building until it is possible to return.
20	To begin today, we have divided our session into
21	two parts. We are going to have a series of
22	presentations before we take a break, and then we will
23	have a larger discussion joined by two of our panelists
24	who have agreed to discuss what they have heard and then
25	to add comments of their own about the proceedings.

8

1	To get us started is Mike Scherer. Mike is as
2	renowned and significant a figure in the modern
3	development of economic research and analysis at the
4	Federal Trade Commission as there is. Going back to his
5	time as Bureau Director in this institution and through
6	his recurring assistance, research and analysis, I think
7	it is fair to say that, in the illustrious collection of
8	those who have served as Bureau Director of the Federal
9	Trade Commission, none has been more distinguished in
10	that very hall-of-fame like collection of individuals.
11	Mike is also well known for the extent to which
12	not simply has he done theory, but one of the reasons we
13	asked Mike to come here is Mike's particular affinity
14	and interest in empirical work and the extent to which
15	empirical work, as well as history and an examination of
16	the past, has figured into his own scholarship.
17	Mike, please, thank you.
18	(Applause.)
19	DR. SCHERER: Thank you for those kind words,
20	Bill.
21	Let me just briefly address the third of Bill's
22	questions, how to learn. In many ways, I have been a
23	disciple of Joseph Schumpeter, not the stuff he wrote
24	about monopoly and technological progress, but what he
25	wrote about how economics advances. Schumpeter argued

9

1	that economic analysis was all about three things. It
2	was about theory, it was about statistics, and it was
3	about history. To do economic analysis right, you need
4	all three, and I have tried hard to do all three of
5	those things. I think in the profession now there is a
6	bit of an imbalance; in particular, we do too little
7	history.
8	I am not sure whether it was distributed or
9	whether it is on the web or whatever, but I do have a
10	background paper for the meetings entitled
11	"Technological Innovation and Monopolization." It is a
12	case history of seven great high-tech monopolization
13	cases in the 20th Century, and the thrust of my remarks
14	will be based upon that paper.
15	Now, first of all, how do you monopolize? Well,
16	it is pretty well known. Mergers, here we have very
17	strong precedent, so I won't dwell longer. Natural
18	advantages, such as economies of scale, the control of
19	natural resources, network externalities and the like,
20	these are fairly rare except in the traditional
21	regulated industries or in those cases where you define
22	the market very narrowly, as in certain pharmaceutical
23	deals.
24	The most interesting one is surely superior
25	efficiency and especially technical innovation. These

10

1	pose the hardest cases for antitrust. When a firm
2	achieves a monopoly position through superior efficiency
3	or innovation, one faces very difficult trade-offs. We
4	should clearly, clearly be encouraging technological
5	superiority, but where is the line crossed? That is the
6	really tough question.
7	A subset of this is patent accumulations. In at
8	least two of the seven cases I analyzed, that is the key
9	to how firms monopolized, specifically, General Electric
10	in the lamp case and AT&T in the telephone case. We did
11	not do anything about it early in the century, and
12	therefore, we had a raft of problems to deal with
13	beginning in the 1940s and later.
14	There are some puzzles here. There is one that
15	I really think the FTC or someone ought to study very
16	carefully, and that's Cisco. Cisco reached its dominant
17	position in the network switch business on the strength
18	of about 100 acquisitions and a lot of patent
19	acquisitions. Was that necessary? Would we have had
20	the best market structure for the switch industry if
21	antitrust had intervened against these mergers?
22	I remember one time being at a cocktail party in
23	Cambridge and meeting a gentleman who told -- you know
24	what you do at these cocktail parties, "What do you do?
25	What do I do?" He said, "Well, what I have done, I have

11

1	developed a switch that is a thousand times faster than
2	anything Cisco has." He ran a high-tech startup,
3	needless to say. I said, "What are you going to do with
4	it?" "Oh, we are going to exploit it. We are going to
5	market it." The next thing I know, he is bought by
6	Cisco for a couple of billion dollars.
7	Now, what would have happened if this guy had
8	been encouraged to develop the switch technology on his
9	own? These are interesting counterfactual questions
10	that ought to be explored carefully.
11	I pass on very briefly to the pricing
12	consequences of monopoly. It has to be brief, because
13	the theory and the evidence are extraordinarily complex.
14	It depends critically on entry barriers, broadly
15	defined, or cost structures. In particular, if entry
16	barriers are low, you have the paradox of explaining how
17	a firm achieved dominance despite having low entry
18	barriers.
19	The United States Steel case, decided by the
20	Supreme Court in 1920, bears careful examination. The
21	evidence is very clear. The Bureau of Corporations did
22	a superb job studying that industry. U.S. Steel had no
23	cost advantage over its rivals after the Carnegie
24	properties had settled into normality. So, it had no
25	cost advantage. How could it preserve its dominant

12

1	position? Well, the answer is it could not, and so it
2	chose an umbrella pricing strategy. It set prices high
3	enough to provide nice profits for everybody in the
4	industry. That encouraged a flood of entry, and
5	gradually, U.S. Steel's market share declined, which the
6	Supreme Court saw as evidence of effective competition,
7	the declining market share.
8	In fact, what it was evidence of was setting
9	prices monopolistically high above the entry-deterring
10	level and behaving essentially sluggishly about entry,
11	and as a result, we have a steel industry that inherited
12	this tradition of sluggishness, of not responding to
13	price signals for 50 years until it got into big trouble
14	in the 1970s and 1980s.
15	Well, much more important than pricing is
16	technological innovation, much more important. There I
17	am clearly a "Schumpeterian." The question is, are
18	monopolists, are dominant firms, superior innovators?
19	The theory we have on this -- and we have got a lot of
20	it, and evidence, too -- the theory and evidence on this
21	say there's a duality. On the one hand there are
22	situations, situations mainly associated with
23	slow-moving technologies, where the science base is
24	changing slowly. There are situations where a
25	monopolist will, in fact, be a superior innovator, where

13

1	only a monopolist is able reasonably quickly to realize
2	sufficient quasi-rents to cover the R&D cost. Those
3	cases definitely do exist in small markets and markets
4	where the science base is moving slowly.
5	But there's an exception when the science and
6	technology base is moving rapidly, where you have
7	revolutions, the kind of revolution we have had in
8	information technology in the last few decades, where
9	that is happening, and/or when monopolists are reluctant
10	to cannibalize the rents that they are earning on the
11	products that they already have marketed. In those
12	cases, firms in dominant positions are almost surely
13	sluggish innovators. I say "almost surely" because
14	here, too, one can find exceptions.
15	The most interesting exception in recent years I
16	think has been Intel. Andy Grove's book Only the
17	Paranoid Survive is a really nice example. I
18	participated for the FTC in the case against Intel and
19	read all of Andy Grove's memoranda for several years.
20	Intel was really terribly alert to new technological
21	challenges and tried hard to stay abreast of them and
22	not be out-competed by upstart innovators. Even so, the
23	record is quite interesting. I do not have a slide
24	projector, and I did not bring a slide anyway -- I
25	forgot to bring it, it was the most important slide I

14

1	was going to bring with me, and I forgot to put it in my
2	portfolio --
3	COMMISSIONER KOVACIC: We have a sketch artist
4	in the back.
5	DR. SCHERER: No, I will wave my arms so you can
6	see. I did a graph, this was in the FTC's Intel case,
7	from public data. I had a graph on which time was the
8	horizontal axis, and on the vertical axis was the speed
9	of microprocessors, and what one sees is two things.
10	First of all, in the period when Intel had a
11	monopoly, at least in 32-bit chips, where Intel had a
12	monopoly, the trajectory introducting speed improvements
13	was like this, quite gradual, but then AMD and then
14	Cyrix caught up and got into the 32-bit technology and
15	began competing with Intel, and what you see, that slope
16	abruptly turns sharper. There was more rapid increase
17	in the key variable of competition, the speed of the
18	microprocessor, and one also found the individual new
19	product points more tightly clustered, showing that more
20	new products were being brought into the market as a
21	result of the competition from AMD and Cyrix.
22	Intel argued in the FTC's case that we are our
23	own best, sharpest competitors, because we have got all
24	this installed base out there, and we have to bring out
25	new products constantly or people will just stick with

15

1	their old microprocessors. I did a series of simulation
2	analyses, and what I found was that using reasonable
3	parameters, Intel would try to maintain a generation for
4	five or six years in the absence of competition. When
5	there was competition, however, it moved the speed of
6	the introduction process to two or three years.
7	Now, this blends into another aspect where you
8	really have serious problems for antitrust, and that is
9	the so-called fast second strategy. This is a concept
10	that was introduced in the late 1960s by Lee Baldwin
11	and -- I don't know his first name -- Childs, and there
12	has been a good deal of theoretical development on it
13	since. The basic idea is that the dominant firm holds
14	back until there is a real threat -- Andy Grove's Only
15	the Paranoid Survive -- and then when that threat
16	appears on the horizon, the dominant firm comes onto the
17	market with a new product, with all guns blazing, and
18	perhaps with a whole panoply of practices to make life
19	difficult for the new company. You can see them
20	described in the paper I submitted for the record, but
21	you clearly see this kind of conduct in Standard Oil, in
22	General Electric, in AT&T, in Xerox, in IBM, and in
23	Microsoft, you see at least delayed innovation, and for
24	IBM and Microsoft, a powerful fast second strategy.
25	How much time do I have?

16

1	MR. HEYER: You have got another ten minutes or
2	so.
3	DR. SCHERER: Oh, okay. Then I will read Judge
4	Jackson's -- I think it's the penultimate paragraph --
5	MR. HEYER: Five or ten minutes.
6	DR. SCHERER: -- in Judge Jackson's decision in
7	Microsoft.
8	"Most harmful of all is the message that
9	Microsoft's actions have conveyed to every enterprise
10	with the potential to innovate in the computer industry.
11	Through its conduct toward Netscape, IBM, Compaq, Intel
12	and others, Microsoft has demonstrated that it will use
13	its prodigious market power and immense profits to harm
14	any firm that insists on pursuing initiatives that could
15	intensify competition against one of Microsoft's core
16	products. Microsoft's past success in hurting such
17	companies and stifling innovation deters investment in
18	technologies and businesses that exhibit the potential
19	to threaten Microsoft. The ultimate result is that some
20	innovations that would truly benefit consumers never
21	occur for the sole reason that they do not coincide with
22	Microsoft's self-interest."
23	Well, Intel pursued similar policies. Actually,
24	the truth is more nuanced than what Judge Jackson said.
25	What he said was basically right, but recognizing this,

17

1	firms that had to compete with Microsoft or had to
2	compete with Intel pursued more sophisticated
3	strategies. Sometimes they simply tried to avoid areas
4	of dominant firm strategic interest, and therefore, we
5	may have missed significant innovations. We will never
6	know what we have missed.
7	But in other cases -- and I think this is the
8	larger majority of cases -- what they did was made their
9	appearance on the scene and then made it clear that they
10	really would like to be acquired by the dominant firm at
11	a very hefty price, and here we face a tough
12	counterfactual question. Would technological progress
13	be faster if they had seen their way clear to innovate
14	independently rather than having their operations taken
15	over by the dominant firm?
16	Now, my own view is that open competition is
17	clearly superior in inducing vigorous innovation as
18	compared to situations in which one has a relatively
19	secure dominant firm. The presumption of antitrust
20	should be to err on the side of maintaining competition
21	and especially, especially keeping both conduct
22	barriers, including fast second strategies, and
23	structural barriers at minimum feasible levels. This is
24	hard. There is no way to evaluate such situations
25	without a careful rule of reason analysis guided by

18

1	appropriate economic theory. But when monopoly
2	positions exist, the job can be done, and it should be
3	done.
4	At this, I will stop and will be happy to take
5	questions. Thank you.
6	(Applause.)
7	MR. HEYER: I think what we are going to do is
8	we are going to hold off on questions until we get into
9	the post-break round table discussion. We will let each
10	of the panelists go.
11	Let me say a few words about Luke, eager to get
12	up here. Luke has a very long title. He teaches at
13	Vanderbilt. He is particularly proud of his work
14	recently at the Federal Trade Commission, and I am happy
15	to say I know Luke back from when he was a staff
16	economist at the Antitrust Division. Despite his work
17	there, he became chief economist at the Federal Trade
18	Commission.
19	With no further adieu, we can --
20	DR. FROEB: Can we bring up the slides?
21	MR. HEYER: Actually, these aren't Luke's. All
22	right.
23	DR. FROEB: Thank you. It's a pleasure to be
24	here. Every time I go in and out of academia, I get
25	more discouraged about what we are doing in academia.

19

1	We work hard on problems no one cares about and publish
2	results in journals that nobody reads, and so it is a
3	delight to be back here working and thinking about
4	important problems that people care about.
5	This area is the source of the biggest policy
6	disagreement between the U.S. and the rest of the world.
7	The U.S. is relatively permissive towards single-firm
8	conduct, while the rest of the world is not. We have
9	reached agreement, by and large, on how to analyze
10	price-fixing and merger cases. And while we do have
11	differences about individual cases and evidence, we do
12	agree on the analytical framework.
13	There is no such agreement on single-firm
14	conduct, and why do we have this disagreement? What do
15	we really know about single-firm conduct? But more
16	importantly, do we know what we don't know about
17	single-firm conduct, and the message of this talk, there
18	is a lot of stuff we do not know, and I think we have
19	got to be really careful about policy in this area.
20	Before I start, I want to thank those who have
21	contributed to my thinking in this area. I thought I
22	would stop taking credit for other people's work once I
23	left the FTC, but apparently not for a couple more
24	years.
25	Okay, so why is horizontal merger analysis

20

1	easier than vertical? The biggest reason is we ignore
2	the long-run indirect and strategic effects of
3	horizontal mergers. We focus solely on the short-run
4	increases in market power, and we have relatively good
5	understanding of how that occurs. Most disagreements
6	focus on the magnitude of the effect and how to estimate
7	it. In other words, we disagree about the evidence, but
8	not on the analysis.
9	The second reason is that we have these distinct
10	mechanisms through which mergers affect consumer
11	welfare: unilateral effects, entry, product
12	repositioning, efficiencies, and coordinated effects.
13	I think we know less about coordinated effects than we
14	want to, but the other mechanisms are well understood.
15	To analyze cases, we gather evidence on each mechanism,
16	and estimate the net effect by estimating the magnitude
17	and likelihood of each individual mechanism.
18	So, why is analyzing single-firm conduct harder?
19	Well, we are concerned about long-run, indirect
20	strategic effects. We just cannot ignore them. If we
21	did, we would have a very simple analysis. And the
22	second reason is that mechanisms with opposing effects
23	usually appear in a single kind of behavior. Predation
24	is the simplest example. In the short run, firms reduce
25	price, but in the long run, we get fewer competitors.

21

1	Vertical integration has the same problem. In
2	the short run, we have the unilateral effect of vertical
3	integration where firms eliminate the double
4	marginalization. But in the long run, we might have a
5	raising-rivals'-costs or reducing-rivals'-revenue
6	mechanism.
7	Exclusive dealing, again, has two opposing
8	mechanisms. The immediate effect of exclusive dealing
9	is to reduce consumer choice, but indirectly, exclusive
10	dealing serves to align the incentives of the retailer
11	with the goals of the manufacturer. So, balancing these
12	effects is really, really difficult. They appear
13	together, and we do not really have good ways of
14	balancing them.
15	So, for these three reasons, single-firm conduct
16	is hard to analyze. There is a taxonomy that I borrowed
17	from Tim Brennan that says, let's consider the simplest
18	case where we have some kind of behavior that has only
19	two effects, two mechanisms at work. There is a
20	proximate, immediate, direct, short-run mechanism that
21	we may know something about, but the effects of the
22	distant mechanism are much less certain.
23	There are four possible outcomes, the distant
24	mechanisms and the proximate mechanisms can both be good
25	or bad. Those are the relatively easy cases. Where we

22

1	run into problems is when the mechanisms work in
2	opposing ways, where the distant mechanism can be bad or
3	good and the proximate mechanism has the opposite sign.
4	When you are doing single-firm analysis,
5	evidence determines which box you go in, and most of the
6	kind of behavior we are concerned about goes in either
7	the off-diagonal boxes. The good-bad box and the
8	bad-good box, those are the ones where we run into
9	problems. Most of the problem cases fall into the lower
10	left box where we have a distant bad and a proximate
11	good, and you can think about bundling, as an example.
12	Bundling offers consumers a better price for the
13	bundle. That is why they buy the bundle, and they are
14	better. But in the long run, the bundle may exclude
15	competitors, and that may have a negative long-run
16	effect. I have already talked about vertical
17	integration, but loyalty discounts and predation give
18	rise to the same kinds of problems.
19	So, how do we characterize the different
20	regimes? The big difference between the U.S. and the
21	rest of the world is that we disagree on the distant
22	effects of mechanisms, i.e., what is the magnitude of
23	these distant effects and how frequently do they occur?
24	The Europeans are much more concerned with the
25	long-run negative effects of things like bundling and

23

1	predation and loyalty discounts, and so they are
2	concerned with avoiding type II errors. If regulatory
3	agencies are uncertain about the effects of single-firm
4	behavior, they are going to make mistakes. They will
5	either deter behavior which is good, type I error, or
6	let bad behavior go through, type II error. And there
7	is an inevitable trade-off: The only way you can reduce
8	type I error is to increase type II error and vice
9	versa.
10	The U.S. regime is more concerned with type I
11	errors. We are more concerned with deterring good
12	behavior. So, we tend to regulate less aggressively.
13	Europeans are more concerned with type II errors, so
14	they regulate more aggressively. We cannot determine
15	who has the better regime, but we can say that relative
16	to the U.S., the Europeans commit more type I errors;
17	and relative to the Europeans, we commit more type II
18	errors.
19	The "makes no business sense" standard is really
20	about trying to find cases in that box so we do not
21	deter any good behavior. We miss more bad behavior than
22	the Europeans; but they deter more good behavior than
23	we.
24	So, the interesting question and the focus of
25	this hearing is, how do we determine the effects? Mike

24

1	correctly states that the effect question is a difficult
2	counterfactual. How do we know what would have happened
3	had a firm behaved differently?
4	This requires comparing two states of the world,
5	only one of which we observe. That is what Mike means
6	about the counterfactual. We have to figure out what
7	would have happened had the firm behaved differently.
8	There are two ways to do it. You can construct
9	a theory that describes competition, and use that theory
10	to tell me what would have happened had the firm behaved
11	differently.
12	The other way is to use what we call natural
13	experiments, and this is really a misnomer. Any
14	statistician in the audience will cringe when I use the
15	word "experiment," because there is nothing experimental
16	about economics data. We do not get to run experiments
17	with the economy, probably for good reason.
18	When I talk about natural experiments, I am
19	talking about comparing a market with the behavior to a
20	market without the behavior, and drawing inference about
21	the effect of the behavior by comparing those two
22	markets. The big questions here are how well does the
23	experiment mimic the effect of interest; and did we hold
24	everything else constant that could have accounted for
25	change. These are tough questions to answer.

25

1	We would particularly want to draw inference
2	about the distant, long-run, or strategic effects,
3	because we know less about them, and because uncertainty
4	about their effects is the source of conflict between
5	policy-makers, attorneys, and economists. I hate to be
6	so hackneyed, but we need more information; we need more
7	research. However, do we have natural experiments that
8	estimate the effects of these distant effects?
9	Here is my favorite study. It is from a paper
10	by Mike Vita of the FTC, and it estimates what happened
11	when the appeals court overturned the must-carry
12	regulations for cable TV. Local cable TV monopolists
13	must carry local over-the-air broadcast channels, and in
14	close areas like Baltimore/Washington, they must carry
15	both the Baltimore and the D.C. stations. When the
16	Court overturned those regulations, which stations did
17	the cable TV monopolist drop?
18	Would the Baltimore cable system drop the
19	Baltimore over-the-air broadcast stations which compete
20	for audience share and advertising revenue, or would
21	they drop the Washington over-the-air stations where
22	they do not compete and can get the same content? And
23	Mike found that they dropped the channels that had the
24	lower rating, and these tended to be the competitors.
25	Competitors were less likely to be dropped, and Mike

26

1	interprets this as evidence refuting the anticompetitive
2	hypothesis. He found that in the long run a firm will
3	not exclude its competitors, as long as they are
4	carrying a good product. I thought it was a very clever
5	kind of use of the decision to try to draw inference
6	about these long-run distant effects.
7	Another Whinston natural experiment is Indiana's
8	ban on exclusive territories for beer distributors.
9	After a state law banned exclusive territories, beer
10	consumption fell by 6 percent. Here again, the author
11	concludes exclusive territories were pro-competitive.
12	Other experiments show that gasoline prices are
13	3 cents higher in states where refiners are prohibited
14	from owning their own gas stations. For fast food,
15	prices at company-owned stores are 3 percent lower.
16	Another experiment which is pretty messy, and I have
17	given this talk over in the UK, and they fight me on
18	this one, on the banning of tied pubs -- so if you are a
19	beer manufacturer, you can't own your own pub to
20	exclusively promote your own -- you have to carry at
21	least two brands of beer. Small beer manufacturers
22	liked having their own pubs because they were using them
23	to promote their beer, and they thought it was an
24	effective way of competing against large brewers. And
25	once they got rid of tied pubs, price went up and

27

1	quantity went down. However, there were a lot of other
2	changes that were going on at the same time, so it is a
3	hard experiment to interpret. But more telling was that
4	the small beer manufacturers fought the change. They
5	liked being able to own their own tied pubs and to have
6	exclusives with a pub so they could promote their
7	brands, and sure enough, the small -- the small beer
8	manufacturers were hurt by the change.
9	At the same time that we were reviewing the
10	literature, Francine Lafontaine, who knows more about
11	franchise agreements than I, and Margaret Slade, who
12	used to be at the FTC and is now in the UK, were
13	reviewing the literature as well, and they used a
14	different taxonomy than we did. We were trying to
15	determine what can we learn about these distant effects,
16	but they were looking at government-imposed changes
17	versus voluntary changes, and they looked at a lot of
18	the same studies that we did. Here is their conclusion:
19	When manufacturers impose restraints, not only
20	do they make themselves better off, but they also
21	typically allow consumers to benefit from higher quality
22	products and better service provisions. In contrast,
23	when the Government prevents these kinds of contracts,
24	the effort is typically to reduce consumer welfare as
25	prices increase and service levels fall. And they

28

1	conclude that the interests of manufacturers and
2	consumer welfare are apt to be aligned, while
3	interference in the market is accomplished at the
4	expense of consumers, and, of course, manufacturers.
5	I would interpret this as evidence that these
6	kinds of arrangements are doing what we want them to do,
7	which is the U.S.'s relatively lenient attitude toward
8	single-firm behavior relative to the rest of the world.
9	I do realize there is a lot that we do not know, and I
10	think it is important to recognize that there is much we
11	do not know.
12	More importantly, how do we generalize these
13	studies to cases? I am not naive enough to think that
14	in a litigation context we are going to have a nice
15	natural experiment that we can interpret cleanly to tell
16	us what to do in a specific case. However, I am not
17	sure how frequently we have been looking for experiments
18	like these.
19	I am much less sanguine than Professor Scherer
20	that we know that much about innovation. So, you look
21	at the Intel innovation, who knows what the innovation
22	rates would have been had we had more people in there?
23	Maybe there was room for only one firm in the market?
24	It is a really tough counterfactual. I wish we knew
25	more.

29

1	And finally, how do we test for the effects of
2	antitrust intervention? Bill Kovacic has been a real
3	advocate for what he calls competition R&D. When we go
4	around the world and talk to new antitrust regimes, we
5	say, look, don't just adopt a regime and freeze it,
6	because what if you get it wrong? Instead, build in
7	some kind of feedback mechanism, and start with the kind
8	of follow-up studies that are done at the FTC and DOJ.
9	I think they are absolutely crucial to try to
10	characterize what are we doing, and to try to figure out
11	what would have happened had we done something
12	differently, in hope of improving.
13	So, characterizing what we do and determining
14	what its effects are really tough, but there are some
15	instances where we can figure out what is going on, and
16	I think we have to be on the lookout for good natural
17	experiments.
18	I guess that is all I want to say.
19	MR. HEYER: Thank you.
20	(Applause.)
21	MR. HEYER: Okay, our final panelist presenter
22	pre-break is Professor Wally Mullin. You have got his
23	bio. He is a professor at George Washington University,
24	and particularly of interest to us I think here is that
25	he has done a fair amount of empirical work on some of

30

1	the issues we are trying to grapple with. A lot of us
2	have a lot to say about theory, but he has gotten his
3	hands dirty a bit, and we look forward to his remarks.
4	DR. MULLIN: Thanks. I am delighted to have
5	this opportunity to appear in these public hearings, and
6	I thank the Department of Justice and the Federal Trade
7	Commission for jointly sponsoring these hearings and, of
8	course, in particular, the co-moderators today, Ken
9	Heyer and Bill Kovacic.
10	So, switching gears, today I want to talk about
11	what lessons we can draw from the history of antitrust
12	enforcement, okay? Now, these may very well be lessons
13	that are kind of in the DNA of current antitrust
14	enforcers, but in the interest of redundancy, I am going
15	to include some of those lessons as well.
16	So, the initial set of dominant firms arose out
17	of the trust movement in the sort of merger to monopoly
18	way. So, in saying that this should be an area of
19	contemporary interest, you know, I certainly acknowledge
20	that similar economic and legal conditions may never
21	return; however, the historical emphasis can still
22	provide a modern researcher with a relatively large
23	sample of dominant firms which faced antitrust scrutiny.
24	So, as an empirical economist, that is very attractive.
25	So, I am going to focus in the discussion today,

31

1	in part, as reflected in my own work, on an admittedly
2	non-random sample of these firms, okay, Standard Oil,
3	U.S. Steel, which Mike has already talked about a little
4	bit, and American Sugar Refining Corporation. So, this
5	choice arises out of a variety of factors. One is sort
6	of the economic importance of the firms, you know, at
7	that particular time, the legal significance of the
8	associated antitrust decisions, and to some extent the
9	similarity and differences in their business strategies.
10	In work with co-authors, I have studied two of
11	these firms. I haven't published any work on Standard
12	Oil, but other people here have, and obviously it's a
13	well-known case in terms of monopolization law.
14	So, since all three firms faced antitrust
15	prosecution, we can examine not only dominant firm
16	behavior, but also the effects of prosecution, and we
17	can also study the effects of remedy as implemented or,
18	admittedly, more speculatively, consider the effects of
19	remedies that were not ordered, because in some cases no
20	liability was found.
21	So, let's start with Standard Oil. My remarks
22	on this will be relatively brief, reflecting sort of
23	comparative advantage issues. So, Standard Oil, right,
24	if we want to have a poster child for different types of
25	dominant firms, Standard Oil was an aggressive

32

1	competitor, okay? So, while the claim that Standard Oil
2	engaged in predatory pricing has been debunked by McGee,
3	the company had other practices that still marked it as
4	an aggressive competitor. For example, Granitz and
5	Klein in 1996 published an article studying how Standard
6	Oil obtained differential rebates from the railroads on
7	petroleum transportation, and that is a source,
8	according to Granitz and Klein, of their sort of
9	supra-competitive rents, and those rebates, of course,
10	advantaged it relative to other refiners.
11	Of course, Standard Oil was found guilty and
12	dissolution was ordered, and it was kind of alluded to
13	by Mike, Bill Comanor and he have argued in a paper that
14	dissolution of Standard Oil raised long-term industry
15	performance, and also in that paper, this is
16	counterfactual, it would have been good had U.S. Steel
17	been dissolved.
18	In his academic work, Bill Kovacic has argued
19	that the effect of this dissolution rests in part on the
20	fact that the dissolution involved formerly independent
21	entities. So, one shouldn't necessarily take this as a
22	dissolution child's story in which everyone lives
23	happily ever after as an automatic indication that
24	structural remedies in all forms and in all
25	circumstances will work. You have to be sensitive to

33

1	the particular facts involved, but given the fact that
2	Standard Oil was organized as such that what was spun
3	off were things that were in some sense formerly
4	independent or had a certain amount of autonomy within
5	Standard Oil in terms of decision-making, in terms of
6	things like corporate culture, the enterprise was able
7	to grow and prosper going forward, and so my take-away
8	would be that, you know, a different remedy in another
9	industry or even with a firm with a different internal
10	organization and history might have unduly sacrificed
11	production costs, but that is merely a speculative
12	comment with a note of caution.
13	So, in terms of U.S. Steel, Mike has already
14	touched upon part of this. So, you know, John D.
15	Rockefeller and Standard Oil is the poster child for the
16	aggressive competitor. United States Steel is sort of a
17	poster child for a dominant firm that may be good for
18	competitors and bad for competition, which was something
19	that the Supreme Court didn't realize at the time.
20	So, in published work with co-author brothers,
21	and it's otherwise hard to find two other Mullins, we
22	have presented evidence that dissolution, which, of
23	course, was never ordered, would have lowered steel
24	prices in that case, in particular, and raised steel
25	output. So, in particular, the pattern of

34

1	contemporaneous stock market reactions to events from
2	the dissolution suit, okay, basically from 1911 to 1920,
3	not only judicial decisions but periods when it was
4	rumored U.S. Steel might dissolve itself to basically
5	avoid prosecution, and then a denial of that rumor the
6	next week, some subset of the events that I mentioned
7	ended up having big stock market reactions for U.S.
8	Steel, indicating that there was news sent to the
9	securities markets in those particular events, and in
10	those weeks, the stocks of customers, in particular, of
11	U.S. Steel, particularly the railroads, reacted in a way
12	that suggested that the stock market believed that
13	dissolution would have lowered steel prices.
14	So, interestingly -- and this is a bit in
15	contrast to maybe what Mike Scherer was talking about --
16	one of the things I also find of interest, and this is
17	part of the tension of monopolization law, is that there
18	are parts, going back to things that might have
19	potentially been sources of market power, that
20	contemporary scholarship would suggest maybe were, in
21	fact, efficiency-enhancing. So, in particular, U.S.
22	Steel was losing market share over time, and you might
23	think, well, wait a minute, is there some sort of scarce
24	factor upstream from steel production that they could
25	use and acquire in order to foreclose entry, you know,

35

1	or at least put a limit on that, right?
2	So, historically they were vertically integrated
3	into iron ore properties, as the Carnegie properties had
4	been, and during the period where they were undergoing
5	antitrust scrutiny at the start of the 20th Century,
6	they added to that a significant amount by long-term
7	leasing the iron ore properties of the Great Northern
8	Railway and James J. Hill. So, that is why they are
9	referred to as the Hill properties. And that was viewed
10	as anticompetitive by contemporary antitrust authorities
11	for some reason, as I will sort of talk about in the
12	next slide, but that is not only criticized by the
13	standing Congressional Committees -- the Federal Trade
14	Commission wasn't around at the time -- but the Bureau
15	of Corporation's report criticized it, and, in fact,
16	U.S. Steel ends up cancelling the lease in 1911 in part
17	to try to forestall prosecution because this was that
18	big of deal to the Department of Justice at the time.
19	Okay, so what might be some of the lessons we
20	take from there? So, as before, of course, the law
21	should protect competition, not competitors. You know,
22	it strikes me -- as I said, I recognize that this would
23	be known by the contemporary court, but it is a good
24	case to assign students, because you have them read the
25	case, and, of course, the Supreme Court is praising U.S.

36

1	Steel because its competitors had such nice things to
2	say about it at trial, and the contrast with Standard
3	Oil is pretty stark. U.S. Steel's anticompetitive
4	effect is not only due to single-firm conduct in a
5	narrow sense, but U.S. Steel's actions in organizing the
6	Gary dinners, which it later abandoned, clearly had a
7	collusive intent, and they were also bad for
8	competition, although good for competitors.
9	So, another tension of monopolization law is
10	that even a firm with market power may have
11	efficiency-enhancing innovations, right? So, the easy
12	case would be in which, you know, if you wanted to do
13	some variation of the diagram, the easy case would be,
14	oh, there are firms that have market power and there are
15	firms that have cost reductions, and they are completely
16	disjoint. I say empirically, that is not the case. In
17	fact, in terms of work that we have done, U.S. Steel was
18	a firm with both elements.
19	So, in a paper with one of my brother
20	co-authors, okay, we didn't have a falling out over the
21	difference in these papers, orthogonal to that issue,
22	the paper with Joe Mullin examines the Hill ore lease,
23	and says that, on balance, that it seems to be best
24	explained as being efficiency-enhancing rather than as
25	vertical foreclosure.

37

1	There are several reasons for this. So, if you
2	sort of back up, the underlying problem of developing an
3	iron ore mine is a problem of relationship-specific
4	investment, something that was studied later by
5	transaction cost economics, both for kind of developing
6	the mine or the investment in the mine, which, of
7	course, is not mobile once it is sunk, and also
8	development of transportation to get the ore or some
9	variation of the ore to market, and that transportation,
10	given where those mines were, was over the Great
11	Northern Railway, which otherwise would have owned the
12	mining rights.
13	So, the specific contractual terms that were in
14	the lease, which caused the Bureau of Corporations to
15	scratch its head circa 1906, has been studied by people
16	like Crocker and Masten. So, one example of this is
17	they had a take-or-pay provision which was quite large,
18	so U.S. Steel was basically committed to making these
19	large payments, and, in fact, during the initial period
20	of the execution of the lease before it fell under
21	antitrust scrutiny, they were, in fact, investing --
22	they were basically scaling up to exploit that property
23	at a very high level.
24	And it's striking, also, in the sense that you
25	might imagine some notion of vertical foreclosure or

38

1	barrier to entry would be, oh, well, they are going to
2	acquire this iron ore. They have other iron ores. They
3	don't need to exploit it to produce right now. They are
4	just going to sit on it and prevent anyone else from
5	gaining entry to it, but, in fact, they invested heavily
6	in trying to exploit the iron ore.
7	It is possible, of course, it had an
8	anticompetitive effect, so it is not so much a -- you
9	know, a complete nesting of the hypotheses, but rather,
10	sort of saying, our judgment, my judgment, the bulk of
11	the evidence would be that that particular aspect of
12	their innovation was something that was
13	efficiency-enhancing.
14	And, of course, the challenge for contemporary
15	antitrust enforcers is what sort of humility should they
16	exercise when faced with some sort of business practice
17	that they don't automatically have an obvious efficiency
18	explanation for? Now, obviously the staff and other
19	people are going to be aware of transaction cost work,
20	et cetera, right, but presumably, we will figure out 20
21	years from now other reasons why some firms might have
22	some sort of purpose. That doesn't necessarily mean
23	that the behavior is necessarily benign, but that's the
24	situation that requires the people to look at it.
25	So, finally, love of my life, American Sugar

39

1	Refining. So, David Genesove and I have written a
2	series of paper on this. This is one of those things
3	that you don't necessarily know what you're getting into
4	when you start. So, in a paper that recently appeared
5	in the Rand Journal, they profitably engaged in
6	predatory pricing, and that was one of their business
7	practices.
8	Now, these joint hearings have already included
9	a rich discussion of predatory pricing in an earlier
10	session, so I won't recapitulate that now. We might get
11	into some element of that in the discussion. David and
12	I noted in the paper that compelling evidence of
13	predation is rare. That is reflected not only in the
14	academic consensus, but obviously also in the case law,
15	but we think the evidence that we present in the paper
16	in this case is compelling.
17	So, in terms of a couple of things to point out,
18	American Sugar engaged in predation. They didn't prey
19	on all entrants. Every single entry episode didn't
20	trigger predation or didn't trigger immediate predation;
21	however, the nature of the market was such that after
22	they preyed, they acquired the entrants and other fringe
23	firms at lower buy-out prices. So, in a sense, if they
24	were making the dynamic calculation, they were sort of
25	saying, well, here's some small firm, it's entering, you

40

1	know, no big deal. As more firms enter, they are sort
2	of like, okay, well, now it's time to prey and buy
3	people out and raise up our market share.
4	In terms of trying to rationalize the
5	observations under different theories of business
6	behavior, that manipulation of rivals' beliefs played a
7	very big role as in some of the reputation models. So,
8	once again, it is not as if they sent out a clarion call
9	saying that, oh, they were going to prey and then they
10	were going to buy people out, so, in fact -- precisely
11	because there were multiple firms they were basically
12	preying on simultaneously, there are cases in which they
13	basically made an arrangement with one of the firms to
14	say, okay, well, fine, we are going to buy you out, here
15	are these terms, but let's keep this secret, and so --
16	and then continue the war, and then buy out the other
17	firms.
18	So, in some sense, part of the aspect of kind of
19	buying out firms and engaging in predation is that the
20	process is sort of the reverse of what we are calling
21	the free-rider problem when you form a trust, right? If
22	you form a trust, you are going to restrict output, and
23	so people will want to stay outside of it and just take
24	advantage of the output lowering entity.
25	Conversely, if there's predation going on, and

41

1	you know there will be a buy-out and the predatory
2	pricing is going to end, of course, people also want to
3	free-ride on that. So, the manipulation of rivals'
4	beliefs is I think part and parcel of being able to be
5	successful.
6	So, there was a monopolization suit, and it
7	stretched on over a period of time, that eventually
8	resulted in a consent decree. But there are some other
9	sort of, you know, maybe, you know, happy lessons here
10	that antitrust serves as a deterrent on a variety of
11	levels. Part of the rationale of the antitrust law is
12	to be punitive, but obviously you also want to think,
13	well, gee, you hope other firms get the message and we
14	don't have to go prosecute them, or this firm in the
15	future, once bitten, twice shy, and so will behave
16	better, and have some sort of implicit consent decree.
17	So, there are two examples of this, and one
18	deals with American Sugar and one deals with other
19	firms. So, during its monopolization case, American
20	Sugar underwent sort of partial "voluntary" dissolution,
21	so this was before the consent decree, because of the
22	government victories in the American Tobacco and
23	Standard Oil cases.
24	So, focusing on American Tobacco or Standard Oil
25	as cases, those basically had a spillover effect on the

42

1	behavior of another firm, in this case American Sugar,
2	and presumably other firms. The difficulty of the
3	non-random sample is, of course, it may be that the
4	whole universe of firms behaved differently, which is a
5	reason why people should do more work on it.
6	Later on, there is also an impact on American
7	Sugar itself. David Genesove and I also studied not a
8	single-firm conduct, but in terms of collusive conduct,
9	we studied The Sugar Institute of the twenties and
10	thirties, of which American Sugar was the largest and
11	most important member, but no longer as large as in 1911
12	or 1914.
13	So, this is noted in our AER paper, even though
14	it wasn't the focus of that paper, which was that the
15	legal representatives of American Sugar at these
16	basically collusive meetings within the industry were
17	very sensitive to things like discussion of price. That
18	was a part of the battle, in a sense, within The Sugar
19	Institute, one person complaining to his boss, oh, gee,
20	we are never allowed to do anything that's going to have
21	any real effect, and so that may just be the wise
22	counsel of American Sugar at the time, but one has to
23	think that the fact that they had had this antitrust
24	prosecution was something that empowered people within
25	the firm to say, okay, compliance is important. It is

43

1	certainly something you think that going forward would
2	be an important part of antitrust enforcement.
3	So, all I have for now.
4	(Applause.)
5	COMMISSIONER KOVACIC: Thanks, Wally.
6	I would now like to invite Jon Baker to present
7	his comments. Jon, as you know, like Mike and Luke, is
8	part of the galaxy of superb economists who have headed
9	the Bureau of Economics at the FTC. In addition to
10	Jon's affiliation with the Commission, in many ways he's
11	been what I consider to be hitting for the scholarly
12	cycle. Not only has he done excellent quantitative
13	work, both at the Commission in matters such as Staples,
14	but also, in his own published work, he has contributed
15	wonderfully to theory. In studying the deliberations
16	that took place over the Verizon-Twombly matter, I many
17	times went back and referred to Jon's paper on two
18	Sherman Act dilemmas from the early 1990s. And quite
19	apropos for this panel as well, Jon, like so many of our
20	presenters, has a good aptitude for history, reflected
21	not only in his survey paper in the JEP on competition
22	enforcement, but also in his recent paper in the
23	Antitrust Law Journal on the development of widely
24	accepted norms and standards, and his political
25	bargaining paper. We are delighted to have Jon here

44

1	today.
2	DR. BAKER: Thank you. Thank you, Bill. That
3	was a very nice introduction. It is not what I would
4	expect from a case book co-author, but I appreciate it
5	anyway.
6	COMMISSIONER KOVACIC: I should have added, he
7	is the co-author of the most astonishing and --
8	MR. HEYER: Copies on sale in the lobby.
9	COMMISSIONER KOVACIC: During the break, there
10	will be the signing process --
11	DR. BAKER: And I am always delighted to be back
12	to see all my former FTC and Justice Department
13	colleagues. I worked with Ken and Luke back in the old
14	days at the Antitrust Division.
15	Well, so let me -- I have a -- sort of several
16	comments on what we have heard this morning. They are a
17	little bit disjointed, and I will just get into them and
18	see how far we get.
19	The first is on the question of what can we
20	learn from the old monopolization cases. On the one
21	hand, there are very few of them. They are often high
22	profile, but there aren't many, and a lot of them were
23	reviewed when antitrust standards were very different
24	than they are today and when ideas about remedies were
25	different than they are today. I don't think we would

45

1	remedy the Standard Oil monopoly were that to have
2	appeared today anything like the way it was remedied
3	then. We would have tried to get the parts that were
4	broken up to engage in head-to-head competition from the
5	beginning.
6	So, there's something funny about this exercise.
7	The -- you wouldn't -- it's a little like saying, well,
8	what can we learn about merger analysis from studying
9	Pabst and Von's, you know, some poster children of
10	merger cases that are no longer thought to be good
11	precedents, although they are technically controlling
12	Supreme Court precedents, as an aside.
13	Well, what we learn from Mike Scherer and Wally
14	Mullin, I think, is something that perhaps we have
15	always known, which is the value of careful
16	case-specific analysis. This is what the judicial
17	system at its best makes possible.
18	Now, that's not to say that the courts have
19	always undertaken this -- the adversarial system has
20	always forced the same level of analysis that later
21	scholars have been able to bring to these cases. I
22	mean, it took 50 years, but the Mullin Brothers finally
23	got to the bottom of the U.S. Steel case. One would
24	like that to have happened, in the case itself. But on
25	the other hand, it shows you the power of case-specific

46

1	analysis to hear Mike and Wally go through what they
2	have learned about these cases.
3	That's not to say that their conclusions are
4	undisputable, but the kind of analysis they do, they can
5	focus in on the issues, and it really does support the
6	kind of work that we do in the enforcement agencies and
7	the courts.
8	Now, let me move on to say something about the
9	issues Luke raised. It struck me, one interesting point
10	is the short-term focus, Luke says, of our antitrust
11	thinking. He didn't quite put it this way, but I mean I
12	guess I'm a little -- I read it in the light of also
13	thinking about a paper that John Lopatka and Bill Page
14	wrote where they argued that antitrust enforcement
15	courts are more congenial to -- or the decisions, I
16	suppose you would say, the decisions are more driven by
17	the short-term benefits and costs than the long-term
18	ones.
19	If you take that perspective and think about
20	Luke's charts, it seems to me that one message is we
21	shouldn't just give a free pass to all those kind of
22	practices in the lower left box of Luke's taxonomy:
23	Price predation, bundling, vertical integration and
24	loyalty discounts. These are things where I think Luke
25	says the proximate effect is good and the distant effect

47

1	is bad.
2	Now, I suppose that my characterization of the
3	implication of those boxes is a little different from
4	Luke's, but in order to go beyond the picture Luke drew
5	to an enforcement regime that gives a free pass -- well,
6	free pass is a little strong -- but that makes it tough
7	to bring cases in the lower left-hand box, you have to
8	take another step in the logic. You have to argue, as
9	some people do, things like the Government can't do a
10	good job analyzing these practices, separating out the
11	two kinds of effects, and remedying it, and you have to
12	conclude that the costs of one type of error are greater
13	than the other. There's a whole additional apparatus
14	that we have to apply before we can reach the conclusion
15	that antitrust should be hands off on all these
16	practices.
17	In thinking about Luke's taxonomy a little more,
18	I started thinking about most favored customer clause
19	cases or most favored nation clause cases. The Justice
20	Department for a while had an enforcement program
21	involving dominant firms that instituted these kinds of
22	practices. It was a dominant health insurer that had a
23	most favored customer clause in its contracts with
24	healthcare providers, and I'm thinking of -- was it
25	Delta Dental, there's a bunch of Delta Dental cases, and

48

1	I think there's some other ones.
2	So, the idea was the provider, the doctor or the
3	dentist or whatever it was, wouldn't lower rates to
4	rival health insurers without also lowering it to the
5	dominant provider, let's call it Blue Cross, and so that
6	makes it impractical for the rivals or the entrants to
7	make procompetitive deals; that is, rivals to Blue
8	Cross. Insurers want to come in and say if you give me
9	lower rates, I'll funnel more business to you, the
10	provider, and we will both do better, and then this
11	creates competition for Blue Cross.
12	Of course, these most favored customer clause
13	provisions can also result in collusion by making
14	discounting more costly, but we are in the dominant firm
15	context here, so we will put that aside.
16	The interesting thing about these most favored
17	customer clauses as a practice is that there are
18	efficiency justifications that are often offered, but in
19	a health care setting, they are not very plausible. The
20	best efficiency justifications are either preventing
21	opportunism when futures markets are unavailable, which
22	sometimes happens in long-term contracting where you see
23	these kinds of provisions, or perhaps signaling low
24	prices where buyer search is costly, and these are the
25	kind of -- here, we're thinking there about retail

49

1	businesses selling to customers.
2	Perhaps Luke will say to me I just moved these
3	provisions in the health care context from his lower
4	left box to his upper left box, where the efficiency
5	justification isn't very good, and so there isn't a
6	problem, but I think if you accept what I have gotten to
7	so far, that these provisions can be troublesome for
8	dominant firms to contract using them in many of these
9	health care contexts, you have to ask, well, when we
10	move outside the health care context, perhaps to one
11	where the efficiency justification is potentially more
12	plausible, don't we have to analyze? Don't we have to
13	think about whether the bad guy story and the good guy
14	story -- which is more powerful as between the two? So,
15	my take from Luke's taxonomy is we ought to think hard
16	about practices in the lower left-hand box and analyze
17	them as best we can.
18	On natural experiments, Luke, I think you missed
19	an opportunity when you were talking about experiments.
20	I have a new motto for the FTC, and this really would be
21	your motto, not mine, "We fool around with the economy
22	every day." Natural experiments are fine in
23	principle -- that was just a joke -- natural experiments
24	are fine in principle, and I basically am sympathetic to
25	what Luke was trying to do with them.

50

1	Tim Bresnahan and I have a recent paper where we
2	talk about something similar. We say that a key
3	challenge for antitrust analysis and empirical
4	industrial organization economics going forward, which
5	is not recognized in antitrust to the same extent that
6	it's recognized in economics, is to exploit similarities
7	among related industries that focus an inquiry involving
8	the industry and the firms under study. We have some
9	examples different from Luke's, but I think the spirit
10	of the exercise is similar. An important question, even
11	assuming it's a good natural experiment, is what
12	generalization you can make from it.
13	Tim and I think that the right generalization is
14	the level of the industry. In other words, I would look
15	at some of the examples that Luke has about -- oh, I
16	don't know, gasoline divorcement or something like that,
17	but not -- and perhaps that would create a presumption
18	about gasoline retailing, but I wouldn't connect the
19	dots and generalize to all vertical restraints. All of
20	Luke's examples, for example, in his representative
21	studies are about manufacturer- distributor
22	relationships in consumer products. They do not tell us
23	much about most favored customer clauses, for example,
24	in health insurer contracts with providers.
25	Finally -- I am not sure how much time I have

51

1	left. Do I have time left? Okay.
2	MR. HEYER: Is it good?
3	DR. BAKER: It's not as good as what's happened
4	already, Ken. I don't get better.
5	No, I think I'll just stop right there, and I
6	will -- it's not that good, Ken.
7	MR. HEYER: Save it for the discussion, all
8	right.
9	DR. BAKER: We will save it for the discussion.
10	Thank you.
11	(Applause.)
12	MR. HEYER: The final person we are going to
13	hear from before the break is Cliff Winston, who you'll
14	see is a long-time economist at The Brookings
15	Institution and has done just an incredible amount of
16	empirical work, largely having to do with regulated
17	industries but not exclusively, and partly because he's
18	really taken on some tough challenges empirically, he
19	seems like a perfect person to invite to talk here, and
20	let's just hear from Cliff.
21	DR. WINSTON: Thanks a lot for inviting me to
22	this conference.
23	Let me, since I'm a little bit on the fringe in
24	this enterprise, sort of tell you my context and how I
25	was thinking about this and eventually how I synthesized

52

1	what we have heard.
2	When Jim Taronji called me about this, my sort
3	of immediate perception was you were planning a series
4	of conferences that were basically assessing the
5	antitrust activity at the federal level of DOJ and FTC,
6	and I naturally thought this, and it turns out that -- I
7	had just finished a book called Government Failure verse
8	Market Failure that looks at all areas where the
9	government intervenes in trying to correct market
10	failures, including but certainly not limited to market
11	power, but information problems, externalities, public
12	good, public production and the like, and figured, well,
13	this is right along the lines of what I have just
14	written up, and so I can sort of look at what you're
15	doing from this perspective.
16	But I also pointed out that I was going to be
17	away a couple of weeks before the conference and
18	literally just got back late the night before, so it
19	would be good if I got the presentations beforehand.
20	Otherwise, you know, I would have to be on the fly, but
21	I thought there obviously might be difficulties in
22	getting things to me, and I was checking my web when I
23	was in Europe, but late last night, I realized a couple
24	had come in, but unfortunately one was in WordPerfect,
25	and Brookings doesn't use WordPerfect. I assume Mike

53

1	does this as a protest against Microsoft. I like
2	WordPerfect better. So, I didn't have them, but I did
3	have a fall-back position.
4	What I was going to do was sort of outline a
5	template, in general, about how I would assess the
6	performance of a federal agency and what recommendations
7	that I might make in terms of improving performance, set
8	that up, say, okay, and I'll just plug in everything I
9	hear in these areas.
10	So, let me outline the template and then just
11	make a few comments on what we've heard. So, the first
12	thing in general that I would ask and think about for
13	any federal agency is, is there compelling evidence of a
14	problem to begin with? That is, you know, are there
15	some stylized facts, summary measures of welfare, you
16	know, that something is going on, you know, information
17	problems are costing consumers hundreds of millions of
18	dollars a year, monopoly is causing similar kinds of
19	costs?
20	Okay, the first thing, just get a big picture
21	overview, when I do these things with transportation, it
22	is very easy, because I can just point to graphs of the
23	lake, there is a problem, congestion going on, airline
24	delay, going up, there's a problem, not too much
25	controversy about that.

54

1	The second question one would ask, you know,
2	what is the scholarly evidence -- when I mean the
3	scholarly evidence, I mean quantitative, welfare type
4	calculations, and certainly counterfactuals isolating
5	the effects of other factors, on first market failure,
6	what do we know about how markets are performing or not
7	performing, since they may be the source of the problem,
8	and government failure, that is, how are governments
9	doing in all of this, and third, government success.
10	So, you know, here are the things you want to look at
11	from the bottom up, the little pieces of evidence that
12	we look at to assess the agency.
13	Then the third thing, since this really is a
14	scholarly enterprise, when I ask the big picture
15	question, where is the field going? You know, since
16	we're getting a lot of the intellectual infrastructure
17	from the scholars who work in the area, how does the
18	field look at this problem? What kind of research are
19	they doing? Where are they likely to help in the
20	future, if at all? Are there incentives the agency
21	could give to researchers to sort of get them focused on
22	problems that they are interested in, so on and so
23	forth?
24	And then finally, you know, given one, two and
25	three, where do we go from here? How do we put all this

55

1	together and say, okay, here is how I think you can
2	improve your performance and your interventions, or here
3	is what I think we, you know, we need to know before we
4	can give confident recommendations.
5	So, let me go through these now with an eye
6	toward what has been said and what has not been said
7	about them. Okay, first, the big picture question, I
8	didn't really hear exactly what I was looking for there,
9	but there's a reason. It's really hard. They are
10	trying -- and I think it is one of the big problems --
11	maybe the biggest problem with industrial organization,
12	is unlike other fields in economics, there isn't this
13	stylized fact that you're constantly facing that reminds
14	you of what's going on out there.
15	It's not like in labor economics where you hear
16	about what the unemployment rate is, okay, or the
17	percent of people below the poverty line. You hear
18	these numbers, you know, these are the kinds of things
19	that researchers get to work on in dealing with this.
20	It is not like trade where we hear what's going on with
21	the dollar, the trade balance. Recently, it just came
22	out about we now have sort of have negative net capital
23	funds, I assure you now a lot of paper is going to come
24	out about this, trying to explain it to us, what is
25	going on, so on and so forth. You can think of a whole

56

1	bunch of things, but when you talk about IO, yes, your
2	instincts are well, we want some measure of economic
3	welfare, but that's not presented by the Commerce
4	Department. It's hard to construct that kind of thing.
5	Now, that said, there was an effort to do that.
6	In the sixties, there was a lot of effort to think of
7	things in terms of concentration ratios, and that was
8	sort of our stylized fact, and there was even a
9	Commission, the Neal Commission, you know, that met and
10	made recommendations about, you know, deconcentration of
11	industries that exceeded a 70 percent level of
12	concentration, and that may not be something that people
13	take seriously today, but there was a time when that was
14	sort of an orientation towards thinking about IO and
15	even antitrust policy, okay?
16	But there really isn't that, which is a bit of a
17	concern, because you never sort of know, well, are you
18	working on a problem that's really important? And the
19	only one who talked about that was Luke in terms of
20	motivating -- while we care about this, and he said this
21	in terms of, you know, apparent disagreement or I would
22	say just different approaches toward antitrust policy
23	between the U.S. and the EU, and I just simply say,
24	well, does that signify different concerns with the same
25	problem?

57

1	To the extent the U.S. is less aggressive and
2	more permissible and allows certain things to go on,
3	does it basically feel that competition is pretty
4	intense, and maybe this is just signifying we really
5	don't have that much of a problem, whereas in Europe,
6	they might feel that there is more, but this is
7	certainly something to think about.
8	Okay, secondly, the scholarly evidence on the
9	various issues, you know, first, looking at market
10	failure -- and I agree completely with Mike, it's an
11	excellent point, a point that is not made enough, that
12	too much of economists' orientation on market failure is
13	static inefficiencies, so price distortions and the
14	like, where so much of the big gains from policy
15	improvements are the dynamic ones, because that's the
16	counterfactual that you don't see.
17	So, if you look at what we've learned about
18	deregulation in terms of what regulation we're doing,
19	the big ticket effects were suppressing innovation,
20	right? So, there you get, you know, more than first
21	order effects. You get really big effects, you know,
22	shifts of cost curves as you completely change what
23	you're doing, shifts of demand curves where you provide
24	new products, okay? So, to the extent that a dominant
25	firm is working like a constrained regulatory policy,

58

1	you know, the effects can be big.
2	Now, that said, you know, measuring these things
3	are very difficult, and, you know, it's not clear to me
4	that we really have hard evidence on this kind of thing.
5	I think the anecdotes are informative, but it would be
6	nice if there was a really strong body of literature on
7	the dynamic effects of delayed innovation, so on and so
8	forth.
9	I would also add, though, just for balance, more
10	emphasis on the self-correcting nature of markets. All
11	the time you are listening to these firms, they are all
12	dinosaurs, right? Look what's happened to them all.
13	Mike mentioned U.S. Steel. Look what happened to them,
14	right? And it was foreign competition, the mini-mills,
15	right? I mean, look at the auto companies, you know,
16	look at Ford, GM, and it's amazing. You know, go on
17	down the line. Now, this does take time, but I think,
18	you know, it's important to keep in mind the
19	self-correcting nature of markets in all of this.
20	Along with that, then, is the parallel of
21	government failure. Now, there are parallels of all the
22	policies we're talking about. Antitrust is not made in
23	a vacuum. Everything that you're talking about
24	intersects a lot of major policies. Trade protection,
25	for example, right? You know, more often than not we

59

1	hear about, well, we need more competition in the
2	airline industry. Yes, let's allow cabin -- oh, no, we
3	are not going to do that. So, here you have trade
4	policy effectively working against what antitrust policy
5	is trying to do.
6	We talk about technology policy with no mention
7	of what happened in the early 1980s with the change in
8	the patent law, right? Patents are going up now,
9	lawsuits are going up now, you know, talk about, you
10	know, impact on innovation and technical changes, look
11	what's done in technology policy. That's not antitrust
12	policy, but it's the crazy patent system that we've got
13	now with, you know, the change in the '82 Act.
14	Regulatory policy, Luke's point was fair enough
15	about cable behavior, but again, it's a regulatory
16	policy that's facilitating that, you know, the whole
17	communications regulatory policy is screwed up. Again,
18	this is not antitrust's, you know, cross to bear, but to
19	some extent, it is. So, where you have a policy that is
20	constantly at cross-purposes with other areas of what
21	the Government is trying to do, it is going to make it
22	very difficult for you to figure out to do, but I might
23	add, the first best thing to do would be to have a
24	technology policy, regulatory policy and trade policy
25	that makes some sense, okay?

60

1	Government successes, you know, I think the key
2	thing on the government successes is almost more of the
3	learning rather than the status assessments. You know,
4	Standard Oil was interpreted as a success, and let me
5	just suggest that there is some controversy about that,
6	Bob Crandall and I head our exploration on antitrust
7	policy, and you know, our look at what the
8	counterfactual evidence was that, you know, there was
9	very little that we could see from changes in prices, if
10	one wanted to use that as a measure of welfare, and it
11	is certainly not a reasonable starting point for what
12	Standard Oil did.
13	I think the more attractive thing that I would
14	point to about antitrust is the learning just how one
15	thinks about problems in terms of anticompetitive --
16	what was initially thought of as sort of knee-jerk
17	anticompetitive reaction as to whether these things were
18	really efficiency-enhancing types of behavior and also
19	just the nature of dynamics, how things are changed, and
20	I think that's where antitrust policy has gone and is
21	certainly a lot better.
22	Now, the big thing about all of this and my
23	concern about this whole area is the effectiveness of
24	this evidence accumulated, because that's what you
25	really want. In certain areas, just to go to a

61

1	completely different area, you know, one's seen study
2	after study about congestion policy in this country,
3	every one of them, huge welfare losses, the Government
4	ought to have efficient pricing, and no one is really
5	disagreeing with that. There are obviously variations
6	from here to there, but the evidence really builds
7	beautifully, and you can just sort of drop it on
8	somebody's lap and say, okay, look, deal with this, and
9	it's easy to do that.
10	Here, it is quite hard. I mean, yes, there are
11	fragments of evidence, cases here and there, and as I
12	said, what Crandall and I attempted to do was actually
13	get a base case for a starting point of saying that
14	this -- and if you disagree with that, fair enough, but
15	at least build on that, reshape it, and then start
16	adding more, and frankly, the disappointment has been,
17	at least in the reaction to that paper, is, you know, I
18	could -- is predictable either pro or critical antitrust
19	people reacting to it, but in terms of actually new
20	evidence being added to the enterprise, that just
21	doesn't seem to be what idle people care about these
22	days, which leads to my third concern, where is the
23	empirical IO field going? And there was very little
24	mention of that here, and with good reason.
25	I mean, it is not clear where it is going in

62

1	relationship to your interest in what is going on here.
2	I mean, my sense, as I would say more of an observer
3	than a participant, that empirical IO is sort of trying
4	to get "uber" dynamic model of industry behavior, you
5	know, that's what we're looking for, for the -- what's
6	the word -- the Holy Grail, I guess that's because I saw
7	The Da Vinci Code on the plane. That's what we are
8	trying to do, and to the extent there's empirical work,
9	it's pretty much demonstration papers, right?
10	I mean, a lot of them are really pretty trivial,
11	you know, you can get data on it -- and I won't go into
12	examples, but you know what I'm talking about, and you
13	know, who cares? And they don't care. They just want
14	to show, yeah, I can get something estimated with some
15	generalized method of moments estimator and add some
16	structural stuff and something is going to get there,
17	and yeah, I'll talk about an industry, about some hotel
18	off a Nebraska highway, no one cares, but you know, the
19	results actually made sense.
20	The question is, where is this research going?
21	Now, I don't want to rule this out, because this is a
22	big ticket item. If people can succeed -- and this is I
23	think really the positive spin on it -- in really
24	building, you know, a structural dynamic model of an
25	evolution -- structural dynamic model of the evolution

63

1	of industry, to hell with these case studies. You have
2	got your tool, right? You just use this, run through
3	any policy scenario, and you could figure out, you know,
4	where things are going, what you ought to be doing, and
5	that is your guidance.
6	Well, you know, we've tried that with Keynesian
7	models (ph), we have tried that with rational
8	expectations, we have tried that with real business
9	cycles, you know, in a sense it's a parallel to macro
10	that we are really going to figure out in a big picture
11	way analytically how markets behave, industries behave,
12	and that will be your guidance for policy.
13	So, you know, that's where it's going. It's not
14	intersecting I think small case studies will build up,
15	it is not doing thing in terms of big picture facts,
16	even motivating what's going on, what people view to
17	within industry seems to be more the availability of
18	data and possible consistency with the analytics they
19	want to pursue.
20	All right, so, you know, where does that leave
21	us? Well, you know, there are three ways to go, and to
22	some extent you can pursue them simultaneously, you
23	know, you can think about first looking more what the IO
24	field is doing, the general model, that kind of work, or
25	I would say more constructively try to focus that kind