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	LOYALTY DISCOUNTS SESSION
10	WEDNESDAY, NOVEMBER 29, 2006
11
12
13
14
15	HELD AT:
16	UNITED STATES FEDERAL TRADE COMMISSION
17	601 NEW JERSEY AVENUE, N.W.
18	WASHINGTON, D.C.
19	9:30 A.M. TO 4:00 P.M.
20
21
22
23
24	Reported and transcribed by:
25	Brenda Smonskey

2

1	MODERATORS:
2	PATRICK DEGRABA
3	Economist
4	Federal Trade Commission
5	and
6	DAVID MEYER
7	Deputy Assistant Attorney General
8	U.S. Department of Justice
9
10	PANELISTS:
11
12	Morning Session:
13	Joseph Kattan
14	Thomas Lambert
15	Barry Nalebuff
16	David Sibley
17
18	Afternoon Session:
19	Daniel A. Crane
20	Timothy J. Muris
21	Janusz Ordover
22	Willard K. Tom
23
24
25

3

1	C O N T E N T S
2
3	MORNING SESSION:
4	Introduction
5	Presentations:
6	Joseph Kattan
7	Thomas Lambert
8	Barry Nalebuff
9	David Sibley
10	Moderated Discussion
11	Lunch Recess
12
13	AFTERNOON SESSION:
14	Introduction
15	Presentations:
16	Daniel A. Crane
17	Timothy J. Muris
18	Janusz Ordover
19	Willard K. Tom
20	Moderated Discussion
21	Conclusion
22
23
24
25

4

1	P R O C E E D I N G S
2	- - - - -
3	MR. DEGRABA: Good morning, and welcome to our
4	first panel of the day on loyalty discounts which is
5	part of an ongoing series of public hearings on
6	single-firm conduct jointly sponsored by the Department
7	of Justice Antitrust Division and the Federal Trade
8	Commission.
9	This series is designed to help advance the
10	development of the law concerning treatment of
11	unilateral conduct under the antitrust laws.
12	My name is Patrick DeGraba. I'm an economist
13	here at the Federal Trade Commission Bureau of
14	Economics, and I'm one of the moderators for this
15	morning's session.
16	My co-moderator is David Meyer, Deputy Assistant
17	Attorney General of the U.S. Department of Justice.
18	Before we start, I need to do a few housekeeping
19	matters.
20	As a courtesy to the speakers, please turn off
21	your cell phones, Blackberries and all other devices
22	that will beep during the proceedings. Mine's off.
23	Second, the restrooms are across the hall to the
24	left of the guard desk where you came in. So ask a
25	guard because that description won't help you get there.

5

1	The third is in the unlikely event that the
2	building's alarm goes off, please proceed calmly and
3	quickly as instructed. If we must leave the building,
4	exit through the main entrance. After leaving the
5	building, please follow the stream of FTC people that
6	are going to the staging area. They have practiced a
7	number of times and some of them know where they are
8	going.
9	Also, we request that you not make comments or
10	ask questions during the session. It is a moderated
11	hearing. For the speakers, I'm going to ask you to
12	please speak into the microphones. The sessions are
13	being transcribed and videotaped and the microphones are
14	the means by which the sound is captured.
15	The transcripts and other materials from the
16	session will be available on the DOJ and the FTC Web
17	sites. And finally, our next hearing will be next
18	Wednesday, December 6th, on misleading and deceptive
19	conduct.
20	Today's session, loyalty discounts include a
21	host of related contracting practices. The simplest,
22	often referred to as single-product loyalty discounts,
23	involve the seller providing a discount on all units of
24	a good sold to a buyer once that buyer has reached some
25	purchasing threshold.

6

1	More complicated practices, often called
2	bundling loyalty discounts, involve the seller offering
3	discounts or rebates when a buyer has reached a
4	purchasing threshold on several possibly unrelated
5	goods.
6	Such practices have raised antitrust concerns
7	recently, and the appropriate antitrust treatment of
8	such practices is clearly in a state of flux. We are
9	honored to have this morning a distinguished panel of
10	academists, economists, and private practitioners who
11	will discuss the current thinking regarding the
12	treatment of these loyalty discounts.
13	Our panelists this morning will include Barry
14	Nalebuff, a professor of economics and management at the
15	Yale University; Tom Lambert, an associate professor at
16	the University of Missouri Columbia School of Law; David
17	Sibley, a professor of economics at the University of
18	Texas at Austin; and Joe Kattan, a partner in Gibson,
19	Dunn & Crutcher, LLP in Washington, D.C.
20	The organization of the panel is as follows.
21	The four panelists will give presentations of
22	approximately 15 to 20 minutes. It will be timed by our
23	staff here in the front row.
24	We will then take a short break. And after we
25	reconvene, the panelists will have a couple minutes to

7

1	respond to each other's presentations, and then there
2	will be a moderated discussion. We will end about noon.
3	David, do you have any comments?
4	MR. MEYER: Not at this point.
5	MR. DEGRABA: All right. Let's get on with it.
6	Our first speaker today is Barry Nalebuff, who
7	is the Milton Steinbach professor of economics and
8	management at the Yale School of Management.
9	Professor Nalebuff has written extensively on
10	applications of game theory to business strategy and has
11	coauthored the first popular book on game theory, which
12	is used in colleges and business schools throughout the
13	world.
14	His current academic research focuses on
15	bundling and tying. He has provided expert testimony
16	and seminars on antitrust matters to federal
17	administrative agencies and courts in Australia and
18	Europe and has extensive experience consulting with
19	multinational firms.
20	Barry.
21	PROFESSOR NALEBUFF: Thanks. I'm going to be up
22	there and control it?
23	Greetings, good morning. What I'm going to try
24	and do is give you my overall perspective in terms of
25	the way I think about loyalty discounts and bundling.

8

1	And I'm of the view to start with that unlike
2	physics, where one is searching for a brand unification
3	theory, you won't find that here.
4	I still believe there is nothing so practical as
5	a good theory. In this case it will be multiple
6	theories. The reason is that it is different horses for
7	different courses.
8	What matters is the nature of the competition.
9	You care about whether the products in the bundle are
10	substitutes with each other, as would be the case of
11	branded and generic tape; where they are complements,
12	such as aircraft engines and avionics; where they are
13	used in some fixed proportions, like in a nail cartridge
14	and a nail; whether or not one is essential to the
15	other, such as Windows and a media player.
16	Sometimes the goods are neither complements nor
17	substitutes, in the sense of Aspen skiing. Before you
18	go to Aspen, the different mountains are complements.
19	Once you are there, they are substitutes.
20	Sometimes there is no connection, substitutes or
21	complements between them. For example, different blood
22	tests are all essential but it is not that you use them
23	together.
24	The goods that are in the bundle might be
25	positively correlated, negatively correlated or not

9

1	correlated at all.
2	All of these factors end up changing the
3	motivations and the effects of bundling and you have to
4	consider that when you are trying to understand the
5	effects and what to do about it.
6	The good news is that we are not in the desert
7	here lost, that in fact in each case where when you
8	understand where you are, we have the tools to analyze
9	it.
10	In my speed attempt to do 10 propositions in 10
11	minutes, here we will go. I want you to know these are
12	not bundled. You are free to accept any one of these
13	individually. But there is a discount if you take more
14	than three.
15	The first point is that often bundled discounts
16	or loyalty discounts lead to negative prices. The
17	reason for that is the discount often goes back to the
18	first unit that you buy. The end result of that is very
19	peculiar prices, things that are hard to justify.
20	This issue arises both with single and
21	multiproduct rebates. Below, this is an example that is
22	an amalgam of actual prices that I have seen from
23	different cases where things have been normalized and
24	discussed.
25	But the way it works is your price for the first

10

1	31 units was 100. Your price for the 32nd unit was
2	minus 6000. Your price for the next couple units is 100
3	again. When you get to the 95th unit, your price is
4	about minus 800. And then for units 96 through 100, it
5	is 97.
6	Now, if you thought about that as sort of a
7	rational way of doing it, you would say what is going on
8	here, does that really make any sense?
9	Of course the customer should never be in a
10	position of buying fewer than 31 items because in fact
11	the first 32 are free. But then having bought 32, now
12	they are okay until they get to 85 because once you get
13	to 85, 85 through 95 is free.
14	What that means is if a rival wants to come in
15	and displace the firm entirely, it will not happen
16	because 31 units are free. Moreover, a rival will never
17	be able to sell between 85 and 95 or, in that case,
18	between 5 and 15.
19	The solution, in my view, to that is to still
20	give out discounts but to give out discounts on
21	incremental volume rather than go back to square 1.
22	And note, if that's your objective to give
23	people low prices, we have ways of doing that. I'm not
24	preventing the discounts, just trying to make them a way
25	that actually makes some sense.

11

1	The second point is that loyalty discounts can
2	actually create no cost predation. And I'm going to
3	give you a quick example of this in terms of numbers.
4	The reason is that what we do is we inflate the
5	price of A rather than really give a discount. Imagine
6	the normal monopoly price of A is 100 and you can get it
7	at the normal monopoly price if you also buy B at 20.
8	But if you don't buy the B, then I will raise
9	the price of A to 120. Hence, the effective price of B
10	is zero or certainly below cost in this case.
11	Now, the key observation is that nobody actually
12	pays the 120 because nobody is foolish enough to only
13	buy A on an a la carte basis. Therefore, since the
14	threat is credible, it doesn't have to be used and it is
15	not costly.
16	The difference between predation and this type
17	of loyalty discount is that under predation, the firm
18	actually charges below cost, and so customers benefit
19	from those low prices.
20	Here all that is happening is the firm is
21	threatening to charge a high price if you don't go
22	along. It is like the mugger who says "your money or
23	your life," and when you give him your wallet, he wants
24	credit for actually saving your life. Actually, I don't
25	think that gets to count.

12

1	Because there is no need for recoupment, it is
2	easier to implement this. Hence, there is a greater
3	danger of it. Because customers aren't necessarily
4	winning along the way, there is also more reason to be
5	suspicious.
6	To give you another disguised example of this,
7	the following is a case where an incumbent firm had a
8	market power in three goods, 1, 2 and 3, and they
9	offered prices like you see in column 1.
10	However, if you were to buy all four of their
11	products, including their competitive fourth product,
12	then you would get the discount, 16, 26, 51, so on
13	percent.
14	If you added up those discounts, what you
15	discover is that the cost of buying all of the three
16	products on an a la carte basis, which essentially you
17	had to do anyway because they were the only supplier of
18	those three products, ended up being sufficiently high
19	that you were going to save $1-1/2 million by buying the
20	bundle.
21	The end result of that was it was actually a
22	negative incremental price to go and take the
23	competitive product.
24	Once again, that is something that is very hard
25	to compete with. That leads to the following proposed

13

1	test, which is if you have a firm which has market power
2	in A and you are worried about whether or not it is
3	going to extend that to another good, B, look at the
4	price of the A-B bundle versus the price of A alone and
5	ask how much more is the firm charging for A and ask
6	could that firm itself make money selling A at that
7	incremental price or B at that incremental price.
8	So instead of asking whether or not the rival
9	can make money selling B at that price, is the firm
10	itself apparently making incremental profits or not. If
11	it isn't, then what we have is a case of exclusion, and
12	that exclusion can be achieved without cost.
13	One of the things that is nice about this test
14	is that we actually don't have to look at actual rivals
15	or hypothetical rivals, we can look at the incumbent
16	firm's own cost structure. The incumbent firm which
17	knows its own cost structure.
18	Therefore, it is well equipped to discover
19	whether it is passing this test or not. It knows
20	whether it is in the safe harbor or it isn't.
21	There is an extra element to this test that
22	David Sibley and his co-authors have emphasized, which
23	is did the price of A go up or did the price of A-B, the
24	bundle, go down.
25	We should be more worried about the case when A

14

1	alone goes up than when the A-B bundle goes down
2	because, of course, when it is a threat, there is no
3	benefit. Whereas, if the bundle has been discounted, at
4	least customers are getting some value along the way.
5	One point that I think the courts have really
6	missed about loyalty discounts is that some of the ways
7	that these rebates are paid end up being significantly
8	less competitive than a straight price cut.
9	So, again, think of a case where the incumbent
10	has market power in A, and B is a substitute. And the
11	two examples I will take you through are Scotch tape and
12	generic tape or Keflin and Kefzol, two cephalosporins,
13	where Keflin was the big money maker and Kefzol was the
14	new product which is the competitive one.
15	In the cephalosporin market, we had Lily with
16	its monopoly and Keflin, Keflex, Loradine, Kaphacen and
17	facing competition with SmithKline Ancef, which was the
18	exact same compound as Kefzol.
19	The first thing they tried doing was just
20	discounting Kefzol to match the prices on Ancef. The
21	problem with that was that Kefzol ended up being a
22	substitute for Keflin.
23	So not only did they have trouble capturing the
24	market against Ancef, as prices starting getting lower,
25	it started eating in on the demand to Keflin.

15

1	Then they got wise and said okay, we will give
2	you a rebate on Keflin and the other products if you buy
3	enough of our goods.
4	Now, note what happens here. The price of
5	Kefzol ends up being high. I'm getting a million
6	dollars back or some fixed amount of money back, but I
7	don't end up discounting Kefzol.
8	In essence, I'm bribing you to say if you buy
9	all of my goods, I will give you this fixed amount of
10	money. But because Kefzol keeps its price high, that
11	reduces the competition between Kefzol and Keflin, and,
12	hence, customers don't get that benefit.
13	We also see that by its equivalent it is almost
14	as if Lily says to the customer we will give you 100
15	units of Kefzol for free on the condition that that's
16	all you use, which of course is something again that
17	rivals would have a hard time matching.
18	We have the same issue in LePage's. If you are
19	3M, you don't want to get into a price war with LePage's
20	over generic tape, because the cheaper generic tape
21	gets, the more that will eat into Scotch tape prices.
22	What you want to do is how can I beat LePage's
23	without discounting my generic tape. Well, if I give
24	them a bribe, a million dollars just to take my goods,
25	even if they are high priced and you can spread out that

16

1	million dollars over their expected sales, then you can
2	say the overall deal is better for me, Staples, than it
3	is for taking LePage's.
4	But note the incremental cost of another roll of
5	tape is high. What that means is the price to consumers
6	for that tape is going to be high and there will be less
7	substitution of generic for branded product.
8	So in that sense, these rebates don't get passed
9	on to consumers and don't threaten the incumbent
10	monopoly.
11	That's an aspect of these loyalty rebates that I
12	don't think has been appreciated and I think is
13	problematic.
14	Another area is that loyalty rebates make
15	pricing incredibly hard to understand. If somebody
16	offers 2.93, I know that is cheaper than 2.97. But if
17	somebody says you get 3 percent off A and B if you buy
18	B, is that a good deal or not? Well, it depends on how
19	much A I'm going to buy. And sometimes I know the
20	answer to that and sometimes I don't.
21	Moreover, if rivals are trying to compete and
22	think about how much they have to undercut to get the
23	business, that means the B rival has to forecast my
24	demand for A, and, generally speaking, they are not very
25	well equipped to do that.

17

1	So we have seen cases where people misforecast
2	these demands, end up buying the wrong product or don't
3	get discounts as large as they think.
4	I have also found that actually analyzing these
5	price things can often take an MBA. And it is not an
6	understatement to say it costs $10,000 to actually
7	figure out what price is the cheapest, and many times
8	that is not worth it for the individual customer to do.
9	An issue that bothers me about loyalty discounts
10	is that the price a firm charges to a customer shouldn't
11	depend on who else the customer buys from. I have less
12	a problem if the price says if you buy many units,
13	here's the charges. If you buy this many more units
14	this year compared to last year, here's the charge.
15	I think it is very funny to say to the customer,
16	"oh, and if you buy 10 units from Fred, I'm going to
17	charge you more money" or "if you buy 3 percent of your
18	products from Fred, I'm going to charge you more money."
19	The price that I charge you should ultimately
20	depend only on what it is that you buy from me, not what
21	it is that you buy from other people.
22	Now, I realize that the effect may be the same
23	through some volume discounts. But that still leaves
24	many more options in an uncertain environment for a
25	rival to come in than when you literally price based on

18

1	what you are doing with your rivals.
2	You will hear many what I will flat out call
3	bogus justifications for bundled discounts. For
4	example, it is often said that customers like bundles
5	and, hence, that's a justification for doing bundling.
6	Yes, that's true, but it is not a justification
7	for a bundle discount. Because a customer likes it, in
8	theory you could charge more for it. You don't have to
9	offer it as a discount if you are providing something
10	customers like better.
11	We do the discount for price discrimination.
12	Well, there is no room for price discrimination if A and
13	B are consumed in fixed proportions.
14	Moreover, the arguments for price discrimination
15	generally rely on having a negative correlation between
16	the two products or no correlation in valuation between
17	the two products.
18	For example, opera tickets and wrestling tickets
19	you think of as having negative correlation. However,
20	if you look at what's bundled out there, I think you
21	will find that they generally have a positive
22	correlation in value and, hence, don't fit the normal
23	framework that we would expect price discrimination to
24	fall under.
25	Yes, Virginia, bundling can leverage and protect

19

1	market power. Here is an example of how that works.
2	If we have a monopolist whose demand is
3	represented by 10 minus P and the cost is zero, the
4	monopoly price would be 5. Profits would be 5 times 5.
5	Price is 5, quantity is 5.
6	I'm having the B product be competitive with a
7	cost of one. So the price is one. Demand I'm making
8	just to be one unit.
9	Chicago School says don't sell A and B together
10	at 6. I do better just to sell A alone at 5, because
11	there are some people who may not want B, even at the
12	competitive price.
13	What I say is consider the following contract.
14	If you buy my B, I will lower the price of A to 4. But
15	if you don't buy my B, I will raise the price of A to 6.
16	Well, if you think about the cost of that threat
17	and promise, the customer is going to save at least $2
18	on A by buying the B product since they are going to be
19	buying at least four units of A.
20	That means that it is a net savings to them of
21	at least 8, which means they are willing to pay up to 9
22	in order to get that discount. They will pay 9 on B to
23	get that discount.
24	Well, the discount doesn't cost the firm very
25	much. And the reason is that discounts my price from 5

20

1	to 4 only lowers my profits from 25 to 24. Raising my
2	price from 5 to 6 also only lowers my profits from 25 to
3	24.
4	So at a cost to me of only a dollar here, I can
5	do something that will either reward or punish the
6	customer to the tune of 8. And the reason for this is
7	the monopoly is inefficient.
8	So in essence, what I'm saying to the customer
9	is I'm willing to be a less inefficient monopolist if
10	you play ball with me and do what I'm asking on good B.
11	It doesn't make sense to take out all of your monopoly
12	rents on the monopoly product because that's what leads
13	to dead weight losses.
14	What I would like to do is some type of lump sum
15	payment and incremental pricing and charge the customer
16	for the right to buy my goods at a reasonable price.
17	Oftentimes the way we see that happen is the way
18	I charge them for being less of a monopolist is I say
19	you have to buy my other goods at B at inflated prices.
20	It is also the case that the bundle allows firms
21	with multiple market powers to protect themselves. So
22	if I have market power in A and B and charge 10 for A
23	and 10 for B but only 16 for the two together, there is
24	a $4 discount that any single-firm rival would have to
25	meet in order to undercut me.

21

1	Note that my average price is 8. In essence, I
2	get to use that same $4 discount on multiple fronts. So
3	the customer isn't benefitting $4. The customer is only
4	benefitting 2 on each.
5	Rivals would actually have to go 4 below. That
6	is a special sauce in multigood bundling that makes the
7	incumbent have an advantage over rivals. It is sort of
8	why it works.
9	It also explains to me why the right test should
10	not be whether or not the overall bundle is above or
11	below cost but whether or not the individual components
12	at the appropriate incremental price is above or below
13	cost.
14	So the Chicago School story is correct in its
15	limited environment, but it misses most of the
16	interesting cases that we look at when it comes to
17	bundling.
18	Even where there is one monopoly profit, that
19	monopoly profit can be of different sizes. In
20	particular, bundling can allow price discrimination,
21	such as through metering and some of the examples you
22	have seen, which, therefore, leads to greater profits to
23	the monopolist but less surplus to the consumer.
24	It is also the case that many of the motivations
25	for bundling are dynamic, that by preventing somebody

22

1	from getting into the B market, that may be their
2	subsequent entry into the A market which is where I
3	still have market power.
4	It is also the case that bundling and tying
5	provide potential for no cost for closure, which has the
6	same effect as predatory pricing but at no cost.
7	I recognize that bundles versus bundles is
8	generally more competitive than individual items versus
9	each other. So what I would like to be able to do is
10	take the advantage of that competition without the harm.
11	And the way that I do that is the following. I
12	actually take the example from Johnson & Johnson who
13	said, look, U.S. Surgical, you have a full line, we have
14	a full line, Coke and Pepsi, you each have full lines,
15	you can compete against me bundle for bundle.
16	But if I don't have a full line, I will not
17	count your sales in my 80 percent number or 90 percent
18	number. Whatever target I make, it is only a target for
19	other full-line competitors.
20	We have come our way through the deserts often
21	through intuition. There are now some tests that I hope
22	you will believe offer more formal approaches.
23	And I believe -- maybe this is a temptation here
24	-- that the theories of bundling loyalty discounts are
25	now ready for prime time. So I hope you will be able to

23

1	use them.
2	Thank you.
3	(Applause.)
4	MR. DEGRABA: Thank you.
5	Our next speaker is Tom Lambert, who is an
6	associate professor at the University of Missouri
7	Columbia School of Law, where he has achieved the
8	university's Gold Chalk Award for excellence in graduate
9	teaching.
10	Professor Lambert's scholarship focuses on
11	regulatory theory, including antitrust policy and
12	business law. His 2005 Minnesota Law Review article
13	provided one of the first scholarly treatments of the
14	law of bundling discounts.
15	Tom is a member of the eSapience Center for
16	Competition Policy and is a regular contributor to Truth
17	on the Market, a Weblog devoted to academic commentary
18	on law, business, economics and more.
19	Tom.
20	PROFESSOR LAMBERT: Thank you.
21	It is an honor to be here on such a
22	distinguished panel. I will talk today about bundled
23	discounts entirely. I will not focus on single product
24	loyalty discounts.
25	A word about the scope of my remarks. I'm a

24

1	lawyer, not an economist. I'm very concerned with
2	structuring rules in a way that they can be administered
3	by judges and juries and used by antitrust counselors to
4	advice their clients.
5	My focus is on the law, how we would structure
6	the rules.
7	I have a three-pronged agenda that's very
8	ambitious for 20 minutes.
9	Why are bundled discounts troubling, and I will
10	give you the straightforward view the courts have
11	adopted and most of you are familiar with this.
12	Summarizing and critiquing of the leading
13	evaluative approaches offers an alternative proposal
14	that I think is very administrable.
15	The problem with bundled discounts the courts
16	have recognized is they may lead to the exclusion of an
17	equally efficient but less diversified rival even if
18	they are above cost.
19	The classic example of this came in the Ortho
20	Diagnostic case. It is I think a little bit
21	unrealistic, but this is what the court wrote in its
22	opinion and it illustrates the problem, I think.
23	You can have two manufacturers who sell the same
24	product, manufacturer A and manufacturer B. They both
25	make shampoo. Manufacturer B is the more efficient

25

1	producer. It can produce shampoo at $1.25 a bottle.
2	Manufacturer A, it costs $1.50 to produce the shampoo.
3	Manufacturer A, though, is a more diversified
4	rival. It sells conditioner as well as shampoo.
5	So by bundling its shampoo and conditioner and
6	by offering an above-cost bundled discount -- and what I
7	mean there is that the price, the discounted price of
8	the bundle is in excess of manufacturer A's cost of
9	producing the bundle -- manufacturer A can effectively
10	exclude manufacturer B from the market.
11	If the separate price of shampoo and conditioner
12	for A is $2 and $4, so that buying them separately you
13	would have to pay $6, and manufacturer A charges a
14	package price of $5, that is still a dollar in excess of
15	its average variable cost of four dollars. Manufacturer
16	B can't compete with that.
17	In order to sell its shampoo -- and any buyer
18	that buys both shampoo and conditioner will have to pay
19	$4 for the conditioner and will not be willing to pay
20	any more than $1 for the shampoo. Manufacturer B is
21	excluded despite the fact that it is the more efficient
22	producer.
23	So the fundamental problem the courts have
24	identified is that bundled discounts can lead to the
25	sort of exclusion of equally efficient but less

26

1	diversified rivals, and that's the case even if the
2	discount is above cost.
3	All right. I have identified six approaches in
4	the case law and commentary for evaluating the legality
5	of bundled discounts. I want to march through them
6	quickly and explain why I think each is a little bit
7	troubling.
8	The first and the most sort of laissez-faire is
9	a rule of per se legality. This is the rule that's been
10	advocated most recently by Professor Hovenkamp in his
11	new book, "The Antitrust Enterprise," and also the rule
12	advocated by Demicci in the LePage's case.
13	It basically says a bundled discount should be
14	per se legal if the discounted price of the bundle
15	exceeds the aggregate cost of the products within the
16	bundle.
17	The reason for this rule is not that we don't
18	believe that above-cost bundled discounts can ever be
19	anticompetitive. The Ortho Diagnostic example showed
20	how they could lead to the exclusion of a more efficient
21	rival.
22	Administrability concerns motivate this rule.
23	The idea is that it is simply too difficult to separate
24	the pro-competitive wheat from the anticompetitive chaff
25	and will end up chilling pro-competitive bundled

27

1	discounting if we don't have the sort of safe harbor,
2	and so the best approach is to have a per se legality
3	rule for above-cost bundled discounts, very much along
4	the lines of the Brook Group rule.
5	My criticism is -- well, I'm not all that
6	critical. In the long run, this may be the best
7	approach to take. However, I'm not willing to concede
8	that at this point.
9	I think the search for anticompetitive bundled
10	discounts may be worth the cost, including the cost of
11	deterring some pro-competitive bundled discounts.
12	It is very easy to imagine instances of
13	anticompetitive exclusion. Professor Nalebuff and
14	Professor Sibley have modeled cases where this could
15	occur. The Ortho Diagnostic example is a good example.
16	I think there is a fairly easily administrable
17	weeding device that can help us separate pro-competitive
18	from anticompetitive bundled discounts. I will get to
19	that in just a minute.
20	The second approach is at the other end of the
21	spectrum -- and this is an approach from the raising
22	rivals costs literature. I'm thinking in particular of
23	Will Tom, who will speak this afternoon, and Einer
24	Elhauge, who has discussed this in testimony on hospital
25	group purchasing organizations and also in his Stanford

28

1	Law Review article defining better monopolization
2	standards.
3	This approach says that bundled discounts are
4	discounts are illegal if they unjustifiably usurp so
5	much business from their rivals that their rival's costs
6	are erased.
7	Now, the $64,000 question here is how do you
8	determine what is unjustifiable. Every discount tends
9	to usurp some business from rivals. And obviously we
10	don't want to ban discounts.
11	The concern here is that so much business will
12	be usurped from rivals that it will deny rivals
13	economies of scale, make it harder for them to raise
14	capital.
15	A couple of approaches have been advocated for
16	identifying what are unjustifiable instances of raising
17	rival's cost.
18	Will Tom suggests in his article on the
19	Antitrust Law Journal that we adopt a case-by-case test
20	where the courts look to see is this an exclusionary
21	usurpation of the business or a pro-competitive
22	usurpation of the business.
23	That is difficult because that leaves a lot open
24	to the whims of juries and judges and will likely have a
25	chilling effect on pro-competitive bundled discounts.

29

1	Professor Elhauge has suggested an approach
2	where a business-usurping discount is justified only if
3	the discounter's business stealing, business usurpation
4	occurs because the bundling has made the discounter more
5	efficient.
6	If you are stealing business because your
7	bundling is making you more efficient, then that's okay.
8	But if you are stealing business for any other reason,
9	then that's illegal.
10	I think this is a troubling approach for several
11	reasons. First, it would prevent price cutting by a
12	monopolist who has reached minimum efficient scale and
13	can't achieve any additional distribution efficiencies
14	by bundling.
15	That person is not getting any efficiency
16	benefits from the bundling and then would be precluded
17	from cutting prices, which seems bad for consumers.
18	Secondly, this approach is very difficult to
19	administer. A court would have to figure out what is
20	minimum efficient scale, very difficult for judges and
21	juries to do.
22	In addition, it has to figure out what discount,
23	what amount of discount is necessary to get the
24	discounter to the point of minimum efficient scale. Any
25	discount beyond that would be excessive discount and

30

1	under Professor Elhauge's test would be exclusionary.
2	That is extremely difficult for judges and
3	juries to administer. For that reason, this approach is
4	likely to have a major chilling effect. Discounters
5	discount at their own peril.
6	The third approach is the approach we sort of
7	see in LePage's. Everyone in this room knows it is very
8	difficult to articulate a rule of law from the LePage's
9	case.
10	There were some key facts that were very
11	important in the court's analysis there. LePage's was
12	not required to prove that it couldn't match the 3M
13	discount. It was not required to prove it was as
14	efficient a manufacturer as 3M was.
15	Instead, it just had to show that it was being
16	excluded. And once it showed that, the burden shifted
17	to 3M to justify its behavior.
18	So if you want to take away a rule from that --
19	and lots of smart antitrust counselors are trying to do
20	so and advise their clients accordingly -- it would seem
21	to be the following. A bundled discount is
22	presumptively exclusionary if the discounter is bundling
23	products not sold by rivals and is winning business from
24	those rivals.
25	Now, the discounter may rebut that presumption

31

1	if it proves a business reasons justification. There is
2	a suggestion in the LePage's case that that
3	justification must show that the bundling saves costs
4	approaching the amount of the discount, very similar to
5	Professor Elhauge's suggestion in the Stanford Law
6	Review.
7	This I believe is a very troubling rule. First
8	of all, since the plaintiff need not establish its
9	equivalent efficiency, this approach essentially creates
10	a price umbrella for less efficient rivals.
11	And there is a suggestion in LePage's that's
12	exactly what happened. LePage's expert economist
13	conceded that LePage's was a less efficient manufacturer
14	of tape than 3M and yet LePage's won.
15	Moreover, since the focus is on product line
16	breadth and not whether an efficient rival is being
17	excluded, this approach will tend to chill bundling,
18	which has a number of pro-competitive benefits which we
19	will talk about in the roundtable discussion. I assume
20	that some of my co-panelists will discuss that issue.
21	The third approach here -- fourth approach, I
22	guess -- the approach we see in the Ortho Diagnostic
23	decision, in that case, the court reasoned that a
24	bundled discount is illegal if the plaintiff shows
25	either that the bundle is priced below average variable

32

1	cost, straightforward predatory pricing, or that the
2	plaintiff is at least as efficient a producer of the
3	competitive product but cannot match the discount
4	without pricing below cost on that product.
5	In other words, you have to show you are an
6	equally efficient rival, and after you show you are an
7	equally efficient rival, you show if you attribute the
8	full amount of the discount to the competitive product,
9	that will result in below-cost pricing by the
10	discounter. You couldn't match that discount.
11	My criticism of this rule, it is a great rule in
12	theory, but this is a very difficult rule to administer.
13	The plaintiff, in order to prevail, has to show
14	that it is an equally efficient rival. To do that, it
15	has to establish its own cost and the discounter's cost.
16	In addition, there are going to be joint costs
17	in here because this is a bundling case. In figuring
18	out the discounter's cost on its competitive product, it
19	has to figure out what percentage of the joint cost it
20	should attribute to that competitive product.
21	That is an incredibly difficult rule to
22	administer. For that reason, I believe this rule, the
23	rule of law in Ortho Diagnostic, may be underdeterrent,
24	because plaintiffs are going to have a hard time winning
25	these cases.

33

1	The next approach is what I'm calling the
2	original antitrust law approach. This is the approach
3	that was advocated in the Areta/Hovenkamp treatise. It
4	was updated this summer. I had to update my
5	presentation.
6	The original approach advocated by the treatise
7	was focused on trying to fix the administrability
8	problems with the Ortho Diagnostic test.
9	Rather than asking if the plaintiff itself was
10	an equally efficient rival, the original antitrust law
11	approach said let's ask if a hypothetical equally
12	efficient single-product rival would be excluded by this
13	discount and without adequate business justification.
14	So essentially we take the Ortho Diagnostic
15	test, we lop off the part where the plaintiff has to
16	show that it is actually an equally efficient rival, and
17	we say if you attributed the entire amount of the
18	bundled discount to the competitive product, would a
19	hypothetical single product be excluded by this
20	discount.
21	This is definitely an easier to administer test
22	because plaintiffs don't have to prove the defendant's
23	costs where there are joint costs. It is troubling,
24	though, for a couple reasons.
25	First, it prevents discount cross-subsidization.

34

1	Consider a situation where you have a seller that sells
2	products A, B and C. Its cost is $4 each. It sells
3	them separately for $5 each. But it would sell the
4	bundle for $13.50.
5	Under the antitrust law approach, this would be
6	a presumptively exclusionary discount because a single
7	product seller of A that was equally efficient at a cost
8	of $4 couldn't match this discount because it would have
9	to charge a price of $3.50, a price below its cost.
10	Now, if you think about an oligopolistic
11	market -- it is not cartelized, but there is a lot of
12	what looks to be tacit collusion -- if you assume the
13	seller that sells A, B and C is selling in that market,
14	it is great that the seller can engage in the sort of
15	complicated pricing.
16	Professor Nalebuff says it is very difficult to
17	figure out exactly what price is being charged.
18	That's a fantastic thing in an oligopolistic
19	market. This sort of pricing can disrupt, this sort of
20	bundling can disrupt oligopolistic pricing. In
21	addition, it is a discount for customers. That would
22	seem to be good in itself.
23	A second problem with the antitrust law approach
24	is there was no requirement that the foreclosed market
25	be capable of monopolization, there was no requirement

35

1	that there be entry barriers in the foreclosed market
2	that the plaintiff was being excluded from.
3	The revised antitrust law approach is definitely
4	superior to the original. But I still think it is a
5	little bit troubling.
6	What Professor Hovenkamp is now saying -- which,
7	by the way, seems to conflict with his book, "The
8	Antitrust Enterprise" -- is that we should analogize
9	bundled discounts to tying and say there is a tie-in if
10	the price is below cost when the entire discount is
11	attributed to the competitive product.
12	Very importantly, the treatise points out there
13	will not be this tie-in if there is another significant
14	rival that sells all products. In the Johnson & Johnson
15	versus Tyco case or U.S. Surgical case,
16	Johnson & Johnson engaged in this bundling, but there
17	was another significant rival that had the same bundle
18	in place.
19	Professor Hovenkamp would say that does not
20	constitute a tie. But absent such a significant rival,
21	there would be a tie-in if there was a below-cost price
22	after the discount was attributed to the competitive
23	product.
24	The treatise then says that after you find time,
25	you should apply a basic rule of reasoned approach, ask

36

1	whether the foreclosed market is capable of
2	monopolization, ask if a collaborative bundle is
3	probable, ask if there are pro-competitive
4	justifications for the bundling.
5	This is a definite improvement on the original
6	version. My criticism is why involve tying at all. It
7	seems to me that the reason that we are concerned about
8	tying in cases like this is that it leads to
9	foreclosure.
10	Why should we focus on the tie rather than
11	focusing directly on the foreclosure issue?
12	Here is my alternative proposal. The goals of
13	the proposal is we want to condemn bundled discounts
14	that could eliminate competitive rivals and result in
15	price increases. We don't want to condemn other bundled
16	discounts. And we want the rule to be easy to
17	administer.
18	What I want to structure my rule to show is that
19	the complaining rival has exhausted its competitive
20	options. You are not a competitive rival unless you
21	have done everything you can to stay in business.
22	The complaining rival must have the ability to
23	match the bundled discounter's efficiency. You are not
24	a competitive rival if you are not as good as the
25	bundler.

37

1	We have to show the foreclosed market is capable
2	of monopolization. We don't want to ban discounts in
3	markets that can't be monopolized because there are very
4	low barriers to entry.
5	Here is a proposed rule. I would have a rule
6	that says that the above-cost discount, and that means
7	that if you add up the cost of all the items in the
8	bundle, they are exceeded by the price of the bundle.
9	So the above-cost discount is per se legal
10	unless the plaintiff could not match without pricing
11	below cost and, number one, barriers to entry exist in,
12	A, the product market in which the plaintiff doesn't
13	participate, and, B, the market for the competitive
14	product, a collaborative bundle is impracticable, a
15	good-faith supply offer was rejected. That means that
16	the foreclosed firm goes to the bundled discounter and
17	says, hey, let me supply my products to you, you buy my
18	product and bundle it.
19	And if those are established, then the bundle is
20	considered presumptively exclusionary, but the defendant
21	gets a rebuttal opportunity to show that it rejected
22	this good-faith supply offer because it wasn't
23	attractive, either the price being offered was too high
24	or the quality was insufficient.
25	Let me explain how this meets all of my goals of

38

1	protecting competitive rivals. We want to protect
2	competitive rivals and only competitive rivals, and we
3	want to ensure that the market that is being foreclosed
4	is capable of monopolization.
5	The above-cost discount is per se legal unless
6	the plaintiff could not match without pricing below
7	cost. That requires a complaining plaintiff to lower
8	its price to the level of its marginal cost.
9	That's what we expect will happen in perfect
10	competition. We should demand that of a complaining
11	rival.
12	Next, it has to show that barriers to entry
13	exist in a product market in which the plaintiff doesn't
14	participate. An option for a plaintiff that's
15	confronting a bundled discount is to enter the other
16	markets in which it doesn't participate. It needs to
17	show there are some entry barriers that prevent it from
18	being able to do so.
19	In addition, it has to show barriers to entry
20	into the market for the competitive product. That's
21	required to show the market is in fact capable of
22	monopolization.
23	Supercompetitive prices could be charged in that
24	market without inviting so much entry that it is
25	impossible to charge those prices.

39

1	Next, the plaintiff would have to show that a
2	collaborative bundle is impracticable. It cannot
3	compete with the bundle by entering into agreements with
4	sellers of other products to craft a competing bundle.
5	These sort of cross-seller bundles are
6	incredibly common. I sent my research assistant to
7	Target, and he found an Olympus digital voice recorded
8	bundled with batteries, Suave body wash bundled with a
9	Schick razor, Colgate White-Plus teeth whitening cream
10	bundled with a camera. Americans are vain.
11	The prima facie case here is intended to show
12	that the plaintiff has exhausted its competitive options
13	and that the market being foreclosed is capable of
14	monopolization.
15	Then we have a rebuttal opportunity. The
16	defendant may rebut by showing that the supply offer was
17	not attractive.
18	The defendant has to show that when the
19	plaintiff came and made the supply offer to me, I didn't
20	accept it because the price it was charging me was
21	higher than my cost. That shows that the plaintiff is
22	in fact a less efficient rival.
23	If the plaintiff can show its prima facie case
24	and the defendant can't rebut, then we have an exclusion
25	of a competitive rival in a market that is capable of

40

1	foreclosure or capable of monopolization, and it would
2	seem to me that liability is appropriate.
3	Otherwise, I would have a rule that these sorts
4	of discounts which are discounts, good to customers, are
5	legal.
6	Thanks.
7	(Applause.)
8	MR. DEGRABA: Our next speaker is David Sibley,
9	who is the John Michael Stuart Centennial professor of
10	economics at the University of Texas at Austin.
11	Professor Sibley was previously the head of the
12	economics research group at Bell Communications Research
13	and served as a member of the technical staff in
14	economics at Bell Labs.
15	In 2003 and 2004, David served as a Deputy
16	Assistant Attorney General for economic analysis in the
17	Antitrust Division.
18	Professor Sibley has carried out extensive
19	research in the area of industrial organization,
20	microeconomic theory and regulation, and his
21	publications have appeared in numerous leading economics
22	journals. He has consulted extensively for various
23	firms and agencies, both in the United States and
24	abroad, on antitrust and regulatory matters.
25	David.

41

1	PROFESSOR SIBLEY: Thank you.
2	The title of my talk, what have we learned since
3	LePage's about bundled discounts, I guess is sort of
4	inspired by the feeling of knowing not what to say at
5	the Antitrust Division when the parties representing
6	both sides of LePage's came to convince us either to
7	support a take cert brief or not.
8	There was not a whole lot the economists had to
9	say. Greg Warden was on the right track when he said
10	"what do the prices do?"
11	It turned out you couldn't tell from the
12	evidence in the record. I take LePage's as kind of a
13	baseline as sort of very useful knowledge.
14	What have we done since then? Well, there has
15	been some progress. We understand now I think better
16	the effects of bundled discounts on both foreclosure and
17	customer welfare.
18	I mentioned foreclosure and customer welfare
19	separately here because, as we will see, it is possible
20	to have a bundled discount which increases customer
21	welfare and yet excludes equally efficient rivals.
22	I expect that to be the case from the way
23	bundled discounts can be structured. I will also talk
24	about tests to determine if a bundled discount is
25	anticompetitive.

42

1	This would be in the spirit of the Ortho test
2	with its explication of extension by Barry Nalebuff or
3	tests whether customer welfare rises or falls. The
4	reference here would be what I was aware of without
5	having to go to any trouble to look up more, Wrightman,
6	Sibley and Roy Nalebuff.
7	The tests here I guess were designed originally,
8	both to try to see whether we could figure out whether
9	bundled discounts are good or bad but also with a view
10	toward the same type of goal that Tom Lambert had,
11	administrability here.
12	We wanted simple tests that didn't require you
13	to calculate complicated things or use data that you are
14	not likely to be able to get in practice. The result is
15	we have tests that will work sometimes but not all the
16	time.
17	To start, I will take a very simple set-up which
18	is actually I think probably the set-up behind some of
19	the slides here.
20	A is a monopoly market served by a firm we will
21	call Firm 1. B is a competitive market. It might not
22	be perfectly competitive. I think for the next five
23	minutes or so I will assume it is perfectly competitive.
24	But it doesn't have to be.
25	Firm 1 is a seller in the B market too. I think

43

1	for the purpose of the rest of the slide, I want to
2	assume a couple things, one, that the B market is
3	perfectly competitive and some B customers will buy B
4	only and some will buy A only.
5	A couple of preliminary observations. Starting
6	from independent pricing, a bundled discount or a BD can
7	raise both profits and customer welfare. That doesn't
8	mean that it will actually happen by a profit-maximizing
9	profit discounter, but it is capable of happening.
10	We should keep that in mind. The logic is
11	really as follows. Let's suppose we have a preexisting
12	time that we can observe where Firm 1 is engaged in
13	independent pricing and it is a monopolist in market A.
14	We will assume that if it hasn't anticipated the
15	onset of the regulatory rule I will be talking about,
16	that the price it charged to the A market was probably a
17	monopoly price.
18	As Barry was saying, if the monopoly price is --
19	if the price charged by firm one in the A market really
20	was the profit-maximizing monopoly price, then it is
21	always possible to have a slight discount on the price
22	of A, which will have an insignificant effect on profits
23	that Firm 1 generates in market A.
24	In Barry's example, it was a $1 increase in
25	profits. From a customer welfare standpoint, that is

44

1	not an insignificant increase in customer welfare.
2	That allows the firm, Firm 1, to bundle that
3	slightly lower price of A with a price of B that's above
4	marginal costs and still get A and B consumers to select
5	the bundle in preference to buying any A at all or
6	paying a bundled price for it and getting B at a
7	marginal cost.
8	In that situation, the A/B consumers are better
9	off. They can all select the bundle that will make them
10	better off.
11	B-only consumers are nowhere better and no worse
12	off than before. They are getting B at marginal cost
13	from all the other perfect competitors out there.
14	If the bundled discount in doing this has an
15	out-of-bundled price no higher than the previous
16	monopoly price under independent pricing, then we know
17	that consumers' options within the bundle are no worse
18	than before.
19	In fact, we have designed the bundle to attract
20	them away from independent pricing of A and marginal
21	cost of B. So they are better too.
22	So starting from independent pricing, which
23	would be the marginal cost of B for everyone, including
24	Firm 1, and the profit-maximizing monopoly price of A by
25	Firm 1, we can always construct a bundled discount which

45

1	raises customer welfare and also raises profits for Firm
2	1.
3	Now, this has an interesting effect here. There
4	is implicit in this the foreclosure result.
5	Since the A and B customers are better off
6	taking the bundle, even at a price of B that is slightly
7	above marginal cost, this means that a B-only seller,
8	one of those perfect competitors, can't appeal to these
9	folks without charging below-market costs. Equally
10	efficient providers are foreclosed and, yet, consumer
11	welfare has gone up.
12	Clearly I have contrived this example to make a
13	point. But it is a point that I suspect in practice
14	comes up often enough to make it interesting.
15	It at least points out when we are talking about
16	bundled discounts, we should not equate foreclosing
17	equally efficient firms with lowering consumer welfare.
18	In my example, consumer welfare is higher. The
19	single-line producers of B would just sell to B-only
20	consumers.
21	Okay. Another point that is implicit to what
22	Barry said which I should have mentioned a moment ago,
23	we are going to assume here that under independent
24	pricing, the Firm 1, the monopolist in the market for A,
25	has not been able to extract all consumer surplus in the

46

1	market for A.
2	In principle, the firm might do this by a
3	perfect two-part tariff, for example. In practice, I
4	think neither Barry nor I think this is a big deal. If
5	it were, we would see lots more two-part tariffs with a
6	lot fewer loyalty discounts than we do.
7	If consumers' demands have some uncertainty and
8	consumers know more about what their demands are, then
9	you will not have a two-part tariff anyway, and you
10	would find that would still be of some use.
11	So far we are talking again about a monopolist
12	in the market for A and everyone inside is a perfect
13	competitor in B.
14	Had I taken more time on this particular slide,
15	I would have had a third bullet point which contrasts
16	what you might expect Firm 1 to actually do with the
17	possibility of raising both consumer welfare and
18	profits.
19	In practice, you wouldn't expect the firm to be
20	interested in raising consumer welfare. So profit
21	maximizing in a very simple setting where B is perfectly
22	competitive and products are not differentiated and the
23	only thing consumers care about is price, in that
24	setting profit-maximizing behavior by Firm 1 is to raise
25	the out-of-bundled price of A a great deal.

47

1	The only point of the out-of-bundled price is to
2	essentially stampede consumers into buying product. In
3	fact, you would give them very bad out-of-bundle
4	alternatives, $10 trillion an ounce or whatever it might
5	be.
6	Of course, they could buy B at marginal cost
7	from competitors. This puts consumers of A and B in a
8	much worse position.
9	So in that setting, the effect of
10	profit-maximizing bundling would not be to raise
11	consumer welfare. It would be to increase profits and
12	lower consumer welfare.
13	Let's not lose sight of the fact that if the
14	out-of-bundled price of A is no higher than the
15	preexisting monopoly price of A under independent
16	pricing, we have the result which has the interesting
17	effect, as I said a moment ago, of excluding sellers in
18	B market from selling to consumers that buy A and B.
19	What I will do next is to change the story and
20	the market for B a little bit. What I talked about so
21	far I suspect people in the audience have heard before
22	from me, from what I have heard. It is on the paper on
23	SSRN for a while. My coauthors and I have labored to
24	extend the results and have had some progress.
25	The story I will tell next, suppose that the

48

1	market for B is not perfectly competitive. It has two
2	firms, one of which is Firm 1. They produce
3	differentiated products.
4	So yes, there are substitutes but not perfect
5	substitutes. Consumers have tastes which are some will
6	prefirm firm 2's version of B and some prefer Firm 1's
7	version. You have a distribution of tastes in the
8	market for B.
9	Some consumers want only B, but there is also a
10	population of A and B consumers. If you look at those
11	folks, the ones who want A, we will assume the same
12	distribution of taste as regard to B. So there are some
13	A/B consumers who really like Firm 1's flavor of B but
14	some who really like Firm 2's.
15	In this setting, the world changes a fair
16	amount. Let me talk you through things before I go to
17	the bullet point here.
18	Firm 1 now has a much more interesting role for
19	the out-of-bundled price of A than it had a moment ago
20	when I assumed that the B market was perfectly
21	competitive and all sellers in B produced a homogeneous
22	product.
23	In this case, Firm 1 realizes there are folks
24	out there wanting to buy my monopoly product which
25	really want to buy B from the other guy.

49

1	The tools at my disposal if I'm Firm 1 are I
2	will have out-of-bundled prices for A and B and bundled
3	prices for A and B.
4	I also know there are some A consumers who also
5	prefer my version of B. How hard do I want to try to
6	retain consumers that want to buy A but really want to
7	buy firm 2's version of B?
8	If I am going to keep those folks, I might have
9	to really discount the price of the bundle a lot. If I
10	do that, then I'm passing up profits that I could make
11	on A and B consumers that like my version of B.
12	So maybe I won't do it. Maybe it is better not
13	to try so hard. I will simply concede A/B consumers
14	that prefer firm 2's version of B to Firm 2.
15	Now, I still would like to make some money off
16	them. I would like them to continue to buy A from me.
17	So my out-of-bundled price for A in this
18	setting, although it is a high price, is no longer set
19	at some infinite level that is designed solely to
20	stampede people into buying the bundle. It is low
21	enough so that A/B consumers that like Firm 2's version
22	of B are still going to buy some A product.
23	So the stand-alone price of A, the out-of-bundle
24	price of A in this setting has a price discrimination
25	goal as well as incentive to buy the bundle. It is a

50

1	more complicated world.
2	Now, look at the first bundle here. Compared to
3	independent pricing, consumer welfare can go up or down
4	assuming that Firm 2 does not exit the B market. In
5	this bullet, when I say consumer welfare can go up or
6	down, I mean in the aggregate. I don't necessarily mean
7	every single consumer.
8	Now, Firm 1 -- why would that work? Firm 1 --
9	there's sort of an interesting effect here. Firm 2 has
10	a tougher job with bundling than under independent
11	pricing because it has to convince consumers to buy B
12	from it at the expense of them having to pay a higher
13	price for A.
14	Under independent pricing, it didn't have this
15	problem. In this set-up here, Firm 2 lowers its price
16	of B because it is now competing, trying to pull people
17	out of the bundle from Firm 1, which it didn't have to
18	do under independent pricing.
19	Firm 1's best response to that is to set an a la
20	carte price for B which is lower as well so B-only
21	consumers are better off in this setting here.
22	If you look at the people buying the bundle, it
23	is not clear whether they are individually better off or
24	not. Usually some are worse off.
25	In that setting, B-only consumers are always

51

1	better off. A and B consumers may be, may not be.
2	Aggregate consumer welfare can go up or down.
3	There is an interesting permutation of this for
4	either entry deterrents, if that's how you want to think
5	about this, or driving firms to another market.
6	Since Firm 2 always sets a lower price of B
7	because it has to work harder to capture consumers
8	because they will be tempted to buy B to get a lower
9	price of A, it always charges a lower price, its cash
10	flow is lower. Depending on the costs it may have, it
11	may in fact exit the market.
12	Look at this from another way. Imagine that
13	Firm 2 has not yet entered the market but it is thinking
14	about doing that and asking itself what would happen if
15	I did enter the market.
16	Well, the story I have gone through here depends
17	on a result which is in the paper that Firm 1's best
18	response to entry by Firm 2 is always to bundle.
19	Firm 2, if it hasn't entered yet, knows if it
20	does, Firm 1 will respond by bundling. Therefore, if
21	there is some cost of entry specific to the active entry
22	that Firm 2 had to incur, they may be deterred from
23	entering, somewhat like the one in the tying literature,
24	the paper by Mike Winston.
25	But there is a difference. You recall that

52

1	Mike's entry deterrence result depends on the equivalent
2	of Firm 1 giving a precommitment to a time, meaning if
3	an entry were to occur, it has to precommit to the tie.
4	There is no precommitment requirement because
5	bundling is what Firm 1 will want to do anyway, the best
6	response. So it is possible to induce Firm 2 to exit
7	even without the precommitment assumption of Winston and
8	others in the tying literature.
9	For a long time in this more complicated set-up,
10	I didn't think we were going to get any sort of fact
11	pattern that would tell us we had a safe harbor here the
12	way we did in the previous story that I just told.
13	My coauthor, David Wrightman, actually came up
14	with one. A sufficient condition in this set-up for
15	consumer welfare to be higher under bundling than under
16	independent pricing, assuming Firm 2 does not exit, is
17	the following.
18	If the a la carte price of A or the
19	out-of-bundled price of A is no higher than it would be
20	under independent pricing and if Firm 2's price for B
21	falls, then whatever happens with the bundle, we can
22	infer consumer welfare has to have gone up, even though
23	the price of B in the bundle may be a little higher.
24	So if we have this fact pattern, we can conclude
25	not only that overall consumer welfare is higher but in

53

1	fact every single consumer is better off.
2	A couple of remarks here. I talked about two
3	kinds of safe harbor tests here. The previous result or
4	model were the B markets perfectly competitive, and we
5	compare the a la carte price of A under bundling to the
6	monopoly price of A and we have a result.
7	And in this case we do the same thing. We can
8	only do that if there is a preexisting independent
9	pricing regime followed by an onset of bundling.
10	And in practice you may not find such a clean
11	set-up. Perhaps bundling began in 1932 or something
12	like that. However, in a litigation setting, the
13	chances are reasonably good that you will run up against
14	this set-up.
15	Typically what happens is firms compete, and one
16	of them will start bundling, and then there is an
17	antitrust complaint. Typically there is a before and
18	after if things make it to the litigation stage.
19	Let me contrast this with the doability of
20	Barry's test. Barry's test does not have the problem of
21	needing to find a before and after situation.
22	It basically lists as attributes the discounts
23	to the competitive line and asks if an equally efficient
24	competitor could undercut that. We could use that in
25	principal using data from the firms if we didn't have

54

1	any reason to think that was a strange point in time to
2	consider.
3	The advantage of -- Barry has a safe harbor, and
4	it's really oriented towards saying when do we exclude
5	competitors. That doesn't necessarily mean consumer
6	welfare is lower if in fact this test has failed.
7	Okay. To sum up, then, in the right
8	circumstances, at least, it seems possible that simply
9	by looking at pricing patterns in order to prepare
10