Group Purchasing Organization (GPO) Purchasing Agreements and Antitrust Law

Posted by on in Uncategorized

In his comments on the DOJ-FTC hearings on group purchasing,
Professor Einer
Elhauge opines that some group purchasing organization (GPO)-facilitated purchasing
agreements involving manufacturers and GPO members continue to pose antitrust concerns,
notwithstanding that the industry has, adopted Voluntary Codes of Conduct in response to
previously stated concerns. While the practices that Professor Elhauge identifies, namely
exclusive dealing [sole source contracting] and some related discounting practices, can be
anticompetitive under narrow and well defined market conditions, these conditions do not obtain
in this particular industry. Further, the agreements in question are clearly justified insofar as
they reduce product costs and attendant marketplace risks.
The extreme case in exclusionary vertical agreements is exclusive dealing, but the
contracts negotiated by GPOs under which their members purchase medical devices are almost
always much less than exclusive dealing, even less so under policies requiring dual sourcing and
limiting rebates.
Typically, the agreements at issue offer buyers a discount of a certain
percentage in exchange for the buyers’ commitment to purchase a minimum percentage of its
needs from a designated purchaser or perhaps a small number of designated purchasers. GPO
members have substantial freedom to purchase alternative products and do so in significant
volumes, particularly where the products in question are differentiated. In such cases hospitals
or individual purchasers such as physicians are likely to have specific product preferences and
will purchase accordingly, whether or not some particular variation is offered on a GPO
By contrast, when a product is fungible, so purchasers are typically indifferent as to

1. I have been retained to give my views on these matters by the Health Industry Group
Purchasing Association (HIGPA).
2. Einer Elhauge, Comments Regarding Hearings on Health Care and Competition Law and
Policy — Statement for DOJ-FTC Hearing on GPOs — Antitrust Analysis of GPO Exclusionary
Agreements (Sept. 26, 2003) (hereinafter “Comments”).
3. More than 60 percent of Novation’s contracts are multi-source. Robert E. Bloch, Scott P.
Perlman & Jay S. Brown, An Analysis of Group Purchasing Organizations’ Contracting
Practices under the Antitrust Laws: Myth and Reality, ANTITRUST REPORT (Matthew Bender &
Co. Fall 2003).
4. Ibid.
Hovenkamp, Antitrust/GPO Agreements 3
which manufacturer they use, then price is the dominant factor in making purchasing decisions
and the member is more likely to purchase under the contract. However, competitive concerns
are virtually nonexistent in the case of fungible products, because these are typically
manufactured by numerous firms in competitively structured markets.
Even outright exclusive dealing is procompetitive and lawful in the great majority of
circumstances. Exclusive dealing enables firms to predict the quantity of sales, and to control
certain types of free riding. In the cases at issue, which do not involve exclusive dealing, an
important rationale for discounts or rebates in exchange for customer loyalty is to encourage
hospitals to give the seller under such contracts a reliable minimum volume of sales. A
vertically integrated firm does not have this concern because it controls its own subsidiaries’
consumption decisions. For example, if Standard owns its own retail stations then it can be
assured that all the gasoline sold through them will be its own, and it can predict its market
simply by having good information about the sales volume of those stations. But if the stations
are independent, as hospitals and other GPO purchasers are, then the upstream firm cannot
automatically be assured of a specific volume of sales. In that case exclusive dealing or related
incentive agreements help the supplier to assure its markets. For example, a gasoline refiner
might enter into exclusive dealing contracts with its retail gasoline stations, requiring the stations
to sell this refiner’s gasoline and no one else’s. Such a contract guarantees the refiner that it will
have the full volume of gasoline sales that pass through this station, and by knowing something
about the station’s general volume the refiner can then predict its own sales volume. See 11
HERBERT HOVENKAMP, ANTITRUST LAW ¶ 1811 (1999); Jan B. Heide, Shantanu Dutta & Mark
Bergen, Exclusive Dealing and Business Efficiency: Evidence from Industry Practice, 41 J.L. &
ECON. 387 (1998); Wesley J. Liebeler, Antitrust Law and the New Federal Trade Commission,
12 SW. UNIV. L. REV. 166, 186-96 (1981); Howard P. Marvel, Exclusive Dealing, 25 J.L. &
ECON. 1 (1982). See also Bloch, Perlman & Brown, An Analysis of Group Purchasing
Organizations’ Contracting Practices under the Antitrust Laws: Myth and Reality, supra note 3.
Exclusive dealing and similar arrangements are particularly important in the presence of
economies of scales in R&D or production, which Professor Elhauge acknowledges exist in the
medical device market. When a significant portion of the cost of a device lies in its R&D, the
expected per unit cost of producing the device can fall dramatically as the volume of expected
sales rises. R&D costs are typically incurred before a device is marketed, and they are “fixed” in
the sense that they do not vary with the number of units eventually made and sold. For example,
if R&D costs for developing a device are $1,000,000, R&D costs would be $1000 per unit if only
1000 units are sold, $100 per unit if 10,000 are sold, or only $1.00 per unit if 1,000,000 units are
sold. The impact varies with the relationship between the costs of simple production and
distribution and overall R&D costs. Assume that the device in this example cost $100 to
produce and distribute. A firm contemplating sales of 1000 units would break even at a price of
$1100. If it contemplated sales of 10,000 units the break even price would be $200. But if it
predicted an output of 1,000,000 units the break even price would be $101.
In sum, for products that have a significant fixed cost component, such as R&D costs, the
manufacturer who can be assured up front of a large production run can typically offer a much
lower price than one who lacks such assurance. Thus a manufacturer might readily determine
Hovenkamp, Antitrust/GPO Agreements 4
that a certain group of hospitals is likely to use 1,000,000 units of a certain device in a one year
period. By negotiating a contract giving the hospitals an inducement to purchase most of their
needs from that manufacturer, this manufacturer can plan its output and as a result price more
aggressively. See Dennis W. Carlton, Vertical Integration in Competitive Markets Under
Uncertainty, 27 J. INDUS. ECON. 189 (1979). This fact fully explains why contracts that reward
hospitals for purchasing a certain quantity or a certain percentage of their anticipated needs are
procompetitive and should not be forbidden in the general run of cases.
The structural requirements for the anticompetitive foreclosure which Professor Elhauge
fears are not satisfied in the medical device markets in which hospital GPOs participate. As I
have noted previously, 5
when markets in this industry are properly defined, no GPO has a market
share as high as 20%. Further, there are many GPOs, and hospitals that can and do join multiple
GPOs or switch their memberships.6
No antitrust case in several decades has condemned
exclusive dealing on such low percentages, and particularly not in the presence of a strong
defense of cost savings, as previously mentioned. Elhauge cites older decisions, such as
Standard Oil Co. of California v. United States, 337 U.S. 293 (1949), that were willing to
condemn exclusive dealing on significantly lower market share percentage. But in Jefferson
Parish Hospital District No. 2 v. Hyde, 466 U.S. 2 (1984), five Justices of the Supreme Court
refused to condemn tying (a per se offense) and four additionally refused to condemn exclusive
dealing on a market share and foreclosure of 30%. See 11 HOVENKAMP, ANTITRUST LAW ¶
Professor Elhauge also suggests that when there are multiple exclusive contracts in a
market, foreclosure must be measured by aggregating the shares covered by all such contracts.
This is true only in a few situations, however. One such situation is if all the contracts emanate
from a single firm whose conduct is challenged under Section 2. For example, if Standard is a
dominant gasoline refiner and enters exclusive dealing contracts with 10 retail stations in the
same market, the relevant foreclosure percentage in that market is measured by aggregating the
shares of the 10 stations. The other situation is when there is collusion among the upstream
firms, which could be challenged under Section 1. For example, if Standard, Texaco, and Exxon
agreed with each other to impose exclusive dealing on their stations, then under antitrust
conspiracy doctrine we ordinary aggregate the shares of the conspirators. See 13 HOVENKAMP,
But aggregating foreclosure percentages makes no economic sense when numerous
upstream firms impose exclusive dealing on numerous downstream firms and there is no
horizontal collusion at either level. The most likely explanation of market wide exclusivity
agreements in such settings is cost savings. For example, the market for branded gasoline is

5. See Herbert Hovenkamp, Competitive Effects of Group Purchasing Organizations’
(GPO) Purchasing and Product Selection Practices in the Health Care Industry, part II.
6. Ibid.
Hovenkamp, Antitrust/GPO Agreements 5
competitive but exclusive dealing contracts are common — independent Exxon stations are
required to sell Exxon’s gasoline exclusively, Texaco stations to sell Texaco’s gasoline
exclusively, and so on. Indeed, among branded gasoline stations today “split pump” stations
(one station selling two brands of gasoline) are virtually non-existent. The most likely
explanation is efficiency. In this case a dealer permitted to sell two different brands would likely
take a free ride on the investment made by the premium brand by transferring goodwill to the
inferior brand. Further, in a highly competitive situation, such as health group purchasing,
competition forces participants to adopt arrangements that reduce costs. For example, if one
GPO can get a lower price by using an arrangement that reduces manufacturers’ costs, other
manufacturers as well as GPOs will compete by offering similar arrangements.
Professor Elhauge indicates that aggregation of two different suppliers’ exclusive sales
could be appropriate if such arrangements are being used to “anticompetitively protect a
duopoly.” Elhauge, Comments, at 13. That might be true if the two firms were colluding with
each other. In the present situation, however, there are typically many more than two
manufacturers competing for hospital sales and no evidence that they are colluding. Professor
Elhauge appears to acknowledge this when he cites me for the proposition that “When exclusive
dealing is used to facilitate collusion … the relevant foreclosure becomes the aggregate
foreclosure imposed by the upstream firms in the collusive group.” (quoting 11 HOVENKAMP,
ANTITRUST LAW 160 (1998)). But once again, he cites no evidence that collusion among
manufacturers is occurring or that collusion is even a likely explanation of the contracts at issue.
He also acknowledges that if a market had “twenty firms, all of whom had exclusionary
agreements,” then there would be no anticompetitive effects. But it hardly takes twenty firms to
produce competition. In most markets a half dozen will do, and the government antitrust
enforcement agencies today routinely approve mergers, which eliminate all competition between
the merging firms, in markets with a half dozen firms. See 4 PHILLIP E. AREEDA, HERBERT
HOVENKAMP & JOHN SOLOW, ANTITRUST LAW ¶ 932 (rev. ed. 1997).
Under the Supreme Court’s decision in NYNEX Corp. v. Discon, Inc., 525 U.S. 128
(1998), a purely vertical agreement must be assessed under the rule of reason. This means that
such restraints must be assessed for competitive effects in a properly defined relevant market,
and no GPO has a sufficient market share (generally, 50 to 60 percent or more) to make such an
agreement unlawful. See 12 HOVENKAMP, ANTITRUST LAW ¶ 2204c (1999); cf. Department of
Justice and Federal Trade Commission Statements of Antitrust Enforcement Policy in Health
Care, Statement #7 (1996), available at (safe harbor for purchasing
agreements that account for less than 35 percent of a properly defined relevant market).
Professor Elhauge’s criticism of the Fourth Circuit’s holding in Dickson v. Microsoft
Corp., 309 F.3d 193, 212 (4th Cir. 2002), cert. denied, 123 S. Ct. 2605 (2003), might be well
taken (See Elhauge, Comments, p. 17). When a single seller imposes exclusive dealing on
different buyers, foreclosure in a section 2 case is properly measured by aggregating those
buyers’ shares. But that still does not justify the conclusion that aggregation is appropriate when
different sellers, acting noncollusively, impose such agreements on their customers, and
particularly not where there are strong procompetitive explanations for the practice, as there are
in the present case.
Hovenkamp, Antitrust/GPO Agreements 6
In all events, exclusivity or quasi-exclusive restrictions imposed by a seller on a buyer
never yield antitrust liability for a buyer under Section 1 of the Sherman Act or Section 3 of the
Clayton Act. While under suitable structural conditions the dominant seller of a product can
exclude rivals by entering into exclusive agreements with its customers, the customer is injured
by any anticompetitive exclusion. That is, the customer cannot profit by making its supply
market less competitive. For this reason dealers are properly regarded as victims rather than coconspirators
when tying, exclusive dealing or quasi-exclusive buying restrictions are imposed on
them. This outcome is explicit in the case of Section 3 of the Clayton Act, which makes it
unlawful for a person “to lease or make a sale or contract for sale..,” but does not apply to
For example, while under the proper structural conditions exclusive dealing by a market
dominating medical device manufacturer might be unlawful for that manufacturer, a purchaser
such as a hospital would be injured rather than benefitted by the exclusivity to the extent that it is
denied the chance to purchase elsewhere. Of course, such a buyer might agree to pay a lower
price in exchange for a promise of exclusivity, but the lower price paid by the buyer is not
antitrust injury. The injury comes from the effect of the agreement in limiting the output of
rivals, and the buyer is injured rather than benefitted by this effect.
For this reason antitrust limits the antitrust liability of buyers to situations in which it was
the buyer’s purpose as well as the seller’s to limit competition in the upstream market. See 7
HOVENKAMP, ANTITRUST LAW ¶ 1474b; accord Va. Vermiculite, Ltd. v. Historic Green Springs,
Inc., 307 F.3d 277, 281-82 (4th Cir. 2002), cert. denied, 123 S. Ct. 1900 (2003); Harold
Friedman, Inc. v. Kroger Co., 581 F.2d 1068, 1074 (3d Cir. 1978); Fuchs Sugars & Syrups, Inc.
v. Amstar Corp., 602 F.2d 1025 (2d Cir.), cert. denied, 444 U.S. 917 (1979); Golden v. Kentile
Floors, 475 F.2d 288, 290-91 (5th Cir. 1973); Sancap Abrasives Corp. v. Swiss Indus. Abrasives,
19 Fed. Appx. 181, 2001 WL 966495 (6th Cir. 2001); Deauville Corp. v. Federated Dep’t Stores,
Inc., 756 F.2d 1183, 1192 & n.5 (5th Cir. 1985).
By the same token, the courts refuse to find an unlawful conspiracy when an employer
“steals” an important employee from a rival by offering higher wages. The employer’s purpose
may in fact be to harm the rival, but the employee is agreeing only to work elsewhere for more
money and as a result does not share in the anticompetitive purpose. See, e.g., Military Servs.
Realty, Inc. v. Realty Consultants of Va., Ltd., 823 F.2d 829 (4th Cir. 1987); Universal
Analytics, Inc. v. MacNeal-Schwendler Corp., 707 F. Supp. 1170, 1181 n.14 (C.D. Cal. 1989),
aff’d per curiam, 914 F.2d 1256 (9th Cir. 1990).
Professor Elhauge’s comments note the possibility that under suitable structural
conditions exclusive contracts could operate so as to deprive rivals of similar economies that will
enable them to reduce their unit costs. In cases involving significantly higher foreclosure
percentages than those at issue in the present situation that might be true. However, antitrust
[law] categorically rejects the implication that a firm should be expected to increase its own costs
in order to permit a rival to lower its costs. Often when some firms increase their output in order
to attain scale economies rivals are deprived of sufficient business to attain optimal production
themselves. For example, when a new grocery store comes into a town, an older or less efficient
Hovenkamp, Antitrust/GPO Agreements 7
store may have inadequate trade to compete at prevailing prices. But this is hardly
anticompetitive. Indeed, it is precisely the way that competition works. A contrary rule would
force firms to keep their own costs higher by selling less, so that there would be room for other
firms to operate in the market. The result would be that consumers would pay higher prices.
Consider the previous example of the firm who could profitably sell 1000 units at a price
of $1100 per unit, 10,000 units at a price of $200 per unit, or 1,000,000 units at a price of $101.
There may be situations (although none has been evidenced) in which one supplier’s $101 bid
for a contract covering 1,000,000 units in anticipated sales would place rival suppliers in a
position where they could not make 1,000,000 sales themselves. Or to state it differently, if the
market in question has room for fewer than 2,000,000 sales, then one firm’s attainment of a
1,000,000 unit contract could force a rival to incur higher costs because it could not produce as
economically. Nevertheless, antitrust policy would never force one firm to forgo its own
economies of scale in order to make room for a rival to attain them. Doing so would force higher
costs on consumers or other purchasers.
Professor Elhauge also suggests that some medical devices, or the industries that produce
them, are subject to network effects, or positive network externalities. Elhauge, Comments, at p.
6. No such devices are identified. A network effect is an economy of scale in consumption
rather than production. See RICHARD A. POSNER, ANTITRUST LAW 247 (2d ed. 2001). That is to
say, a product becomes more valuable to consumers as the number of users (or uses) increases.
For example, the telephone system, which is a classic example of network effects, is more
valuable to users as more other users are connected to the system. As Professor Elhauge notes,
when a good experiences network effects that “seller’s product is more valuable to buyers the
more that other buyers have purchased the same good from that seller.”
I cannot identify a medical device that is subject to significant network effects, and
Professor Elhauge does not identify any. That does not mean that they do not exist. There may
be network effects in certain things such as databases or other technologies that must be shared
or transferred among different hospitals. More likely, what Professor Elhauge describes as
network effects is simple economies of scale in production, or the lower per unit costs that
accrue to a product as more units of that product are produced.
In all events, both network effects (scale economies in consumption) and scale economies
in production are good things in and of themselves, and the presence of either provides a
complete and procompetitive explanation why GPOs would solicit contracts for the purchase of
large quantities of the same good — namely, to reduce the costs of goods purchased by their
members. Indeed, network effects generally argue in favor of, rather than against, widespread
single sourcing. For example, a hospital whose technicians use a computer database that must be
shared would undoubtedly be better if all the technicians used the same database.
Assuming that network effects do exist, Professor Elhauge’s argument would be
tantamount to saying that because the shared database is subject to significant network effects it
is anticompetitive for buyers to contract to rely on a single database program for all of their
needs; rather, they should be forced to give smaller producers of rival databases a chance as well.
But that policy would be highly inefficient, forcing customers to pay the significantly higher
Hovenkamp, Antitrust/GPO Agreements 8
costs of possibly incompatible systems for the benefit of smaller vendors who want to be able to
sell their systems as well. Antitrust would never impose such a socia lly harmful obligation on
Professor Elhauge also concedes that standardization within hospitals is good because it
reduces error costs. Elhauge, Comments, at pp. 11-12. This justifies a hospital’s commitment to
purchase all of its basic needs of a certain device or group of devices from vendors using a
common technology, and in the case of product differentiation this may entail that all be
purchased from a single vendor. Elhauge adds that such standardization is unnecessary across
multiple hospitals, however. While that may be true, it is inconsistent with the proposition that
these devices are subject to significant network effects. If there were significant network effects
for some device, then users at one hospital would be better off as the number of users at other
hospitals increased as well. For example, an organ donor registry is a good that is very likely
subject to network effects in that each participant is better off as the amount of information in the
registry is greater. Thus, assuming a national market, it would be better for hospitals all to
participate in the same registry than to have separate registries covering individual states or
municipalities. Likewise, standardization across hospitals may be useful in developing a broad
database of results from using a particular device.
But the main inducement to inter-hospital standardization is scale economies. While
hospitals may not need to standardize technologies across different hospitals in order to reduce
error rates, economies of scale in production apply at the producer end, not the consumer end, of
the market. That is, the lower per unit costs that accrue to high volume production do not depend
on whether the manufacturer sells to one customer or several. Standardization needs may justify
a hospital in procuring all or most of its needs of a particular device from a single vendor (or
small group of vendors employing the same technology). Production economies of scale thus
explain why different hospitals can obtain lower costs when they use the same technology — or
more specifically, why a manufacturer is willing to offer a lower price to a group purchasing
agent who represents multiple buyers. For this reason group purchasing is an efficient activity in
a wide variety of markets. See 13 HOVENKAMP, ANTITRUST LAW ¶ 2135 (1999). See also
Northwest Wholesale Stationers, Inc. v. Pacific Stationery & Printing Co., 472 U.S. 284, 295
(1985), which noted that joint purchasing by retailers:
permits the participating retailers to achieve economies of scale in both the purchase and
warehousing of wholesale supplies, and also ensures ready access to a stock of goods that
might otherwise be unavailable on short notice. The cost savings and order-filling
guarantees enable smaller retailers to reduce prices and maintain their retail stock so as to
compete more effectively with larger retailers.
In addition, there is no reason for thinking that rebates or discounts given directly by
manufacturers to member hospitals in exchange for loyalty raise competitive concerns, even on
concentration levels far higher than exist in this market. See Elhauge, Comments, at pp. 19-20.
Such offers necessarily exclude much less than does outright exclusive dealing for the simple
reason that an equally efficient rival can match them. For example, if I sell a device at a nominal
price of $1.00, but offer a 10 cent rebate to a buyer who takes 80% of its needs, or who
Hovenkamp, Antitrust/GPO Agreements 9
purchases at least 1000 units, from me, I have basically offered a 90 cent price in exchange for a
given quantity commitment. Getting the commitment up front helps the seller ensure its market
and thus enables it to plan longer production runs, thus achieving lower costs. Significantly,
however, assuming that my 90 cent price is not predatory, an equally efficient rival can capture
the market simply by agreeing to sell at 90 cents as well (whether by using a similar rebate
program or else by simply charging 90 cents to begin with). Accord Concord Boat Corp. v.
Brunswick Corp., 207 F.3d 1039 (8th Cir.), cert. denied, 531 U.S. 979 (2000), which dismissed
such a complaint in the case of a monopolist. Thus, even where a medical device manufacturer
may be a monopolist, such discounts negotiated by a GPO would not be anticompetitive.
In making his argument that GPO purchasing contracts are anticompetitive, Professor
Elhauge relies on decisions under Section 2 of the Sherman Act involving discounting practices
by dominant firms. In Microsoft Corp. v. United States, 253 F.3d 34 (D.C. Cir.) (en banc), cert.
denied, 534 U.S. 952 (2001), the court refused to condemn exclusive dealing under Section 1 of
the Sherman Act. However, it condemned the same exclusivity agreements under Section 2 of
the Sherman Act. See id. at 69-71. Significantly, however, the Section 2 requirement requires
market power, and in that case the defendant had more than 90% of the relevant market. The
same thing is true of the other decisions that Elhauge relies on, Standard Oil Co. v. United
States, 221 U.S. 1 (1911); and Les, Inc., v. 3M., 324 F.3d 141 (3d Cir. 2003) (en banc),
petition for cert. filed, 72 U.S.L.W. 3007 (U.S. June 20, 2003)(No. 01-1865) . In the latter case
3M’s share of the market at issue, like that of Microsoft, hovered around 90% during the relevant
time period. Les, 324 F.3d at 146.
Several things must be said about these cases. First, both Microsoft and Le’s were
monopolization actions against clearly dominant sellers. By contrast, the GPO industry is
intensely competitive, with the largest GPO accounting for medical device market shares in the
range of 15%, and many smaller firms with market shares ranging downward from 12%. The
market would be considered “unconcentrated” under the standard promulgated in the Antitrust
Division and FTC’s 1992 Horizontal Merger Guidelines.7
Discounts by nondominant actors are

7. See Hovenkamp, Competitive Effects of Group Purchasing Organizations’ (GPO)
Purchasing and Product Selection Practices in the Health Care Industry, Part II, noting the
following market shares for the ten largest GPOs, based on 2001 data:
Novation 14.6%
Premier 12.5%
AmeriNet 4.6%
MedAssets 4.5%
ManagedHealth 3.3%
Consort 2.2%
HealthCare Purchasing Partners 1.1%
National Purchasing Alliance .7%
AllHealth .6%
Hovenkamp, Antitrust/GPO Agreements 10
readily explained as efficient, and anticompetitive exclusion is not possible. For example, if a
firm with a 20% market share negotiates a discount for taking even 100% of its needs from one
manufacturer. the balance of the market, 80%, remains available to that manufacturers’ rivals.
Further, if the incentive takes the form of a discount, any equally efficient rival can match the
price by matching the discount.
Second, the Le’s case involved individually targeted bundled discounts designed to
deprive the defendant’s only rival in the cellophane tape market of its trade. 3M created discount
programs that aggregated discounts across multiple products (such as tape, Post-it notes, and
paper clips), and effectively deprived the buyer of the discount unless it took the specified
amount of all three products. By contrast, the plaintiff Le’s produced only tape. While
Le’s could have matched 3M’s discounts on the tape alone, in order to capture the sale it
would also have to compensate the customer for the foregone discounts on the other two
products. See AREEDA & HOVENKAMP, ANTITRUST LAW ¶ 749 (2003 Supp.). There are no cases
I am aware of in which discount practices were strategically designed to exclude rivals who
made smaller subsets of competing products. Indeed, most discounts available to GPO member
purchasers are across the board and result from competitive bidding by manufacturers.
Third, and most importantly, even in a case where the seller is found liable under section
2 there is no warrant for finding buyers liable. While a market dominating supplier (such as 3M
in the Le’s case) might have an incentive to suppress the growth of its rivals, neither GPOs
nor their purchasing members share that incentive. They are purchasers, not sellers or
competitors of their supplier. Their welfare is maximized when the supply market is as
competitive as it can possibly be.
It must be kept in mind that discounting is in nearly all cases a socially beneficial,
procompetitive practice that should be encouraged. In the GPO situation suppliers are willing to
give lower prices in exchange for larger sales. That serves the interests of all classes of
purchasers and their agents, from GPOs, to hospitals and other purchasing institutions, to
individual patients who obtain lower prices. The exceptions to this rule are very rare and include
predatory pricing (below cost discounts intended to exclude or discipline a rival) and the
exceedingly rare case of a bundled discount such as in Le’s, which depends on an exercise
of monopoly power. And even in those rare cases it would be the monopolist, not the GPO, who
violated the antitrust laws. GPOs have no incentive to reduce the competitiveness of the markets
in which they purchase.

(… cont’d)
Innovatix .6%.
That evidence also tends to show that the GPOs behave very competitively vis-a-vis one another.
For example, there have been significant shifts in the market shares of leading GPOs. Further,
there are no structural barriers to entry in the GPO market. See ibid.
Hovenkamp, Antitrust/GPO Agreements 11
Third, the theory of the Le’s case offered in Elhauge, Comments, is inconsistent
with the theory Professor Elhauge previously offered that group purchasing is an exercise in
monopsony power by the group purchasing organizations.8
Targeted bundled discounts is a
dominant sellers’ strategy of earning monopoly returns by specifically targeted discount
packages that rivals offering a single product (or smaller subset of products) are unable to match.
In very sharp contrast, monopsony purchasing by GPOs is a buyers’ strategy of forcing suppliers
to accept infracompetitive returns through the buyer’s exercise of its monopoly buying power. A
practice cannot simultaneously be a targeted discount aimed by one monopoly seller against a
rival and yielding supracompetitive prices, and an exercise in monopsony power suppressing that
sellers’ prices to infracompetitive levels.
Professor Elhauge also suggests that the larger GPOs are firms with market power who
are bargaining with their suppliers to share the monopoly and pass higher prices down to their
customers. See Elhauge, Comments, p. 25. In particular, he suggests that
intermediate buyers have incentives to agree to preserve or enhance seller market power
(by excluding the seller’s rivals or raising their costs) in exchange for side-payments that
split the seller’s supracompetitive profits, or for special discounts that give the participant
buyers market advantages over other buyers and thus enhance the partic ipating buyers’
downstream market power.
Ibid. Such a strategy could occur only in situations where both the device manufacturer and the
GPO in question had substantial market power in their respective markets. No particular device
manufacturers are identified as being monopolists, but one could presume that there are some.
However, the device manufacturer would have no incentive to share its monopoly profits with a
GPO unless the GPO were a monopolist in the market in which it was purchasing. That is, even
monopolists are best off when they sell into downstream markets that are competitive. See 3A
AREEDA & HOVENKAMP, ANTITRUST LAW ¶ 758 (2d ed. 2002). The GPOs are neither
monopsonists or monopolists.
But even if they were, such strategies are output reducing — i.e., designed to induce
higher prices by keeping volume down. I know of no evidence that the device manufacturers
bidding for large GPO contracts have any interest other than getting their output as high as
possible through aggressive bidding. Thus Elhauge’s lengthy discussion of the 1911 Standard
Oil decision and Standard’s use of contracts requiring its railroad suppliers to charge more than
other firms seems completely inapt here, on both factual and structural grounds. (See Elhauge,
Comments, pp. 25, 28-29). Only a monopolist could force the kinds of contracts that Standard
forced, and in any event I have seen no evidence that any GPO has entered into any contracts of
such a nature.

8. See Einer Elhauge, The Exclusion of Competition for Hospital Sales Through Group
Purchasing Organizations (June 25, 2002).
Hovenkamp, Antitrust/GPO Agreements 12
The Voluntary Codes of Conduct
Professor Elhauge is distressed by the fact that the Voluntary Codes of Conduct adopted
by the industry and GPOs individual Codes are merely voluntary and that they apply only to
physician preference products and clinician preference products. As a result, he suggests, those
firms can change their minds and violate these codes whenever they please. Elhauge,
Comments, pp. 18-19.
But antitrust is not an affirmative regulatory enterprise. Rather, it depends on firms’
“voluntary” compliance with the law and intervenes only in the case of a violation. If any firm
should violate the Codes and engage in anticompetitive tying, exclusive dealing, price-fixing, or
some other unlawful practice the antitrust laws are available to provide a remedy.
Second, the fact that certain GPOs might increase administrative fees to 3% poses no
risks to competition. Every intermediary, whether retailer, broker, jobber, distributor, or other
facilitator, requires compensation. As a general proposition, no anticompetitive inference can be
drawn from the mere existence of such payments. They have to be collected in some form, and
collecting them from suppliers as a percentage of actually made sales is undoubtedly more
efficient than collecting them from the much larger group of purchasers.
Elhauge proposes that “[t]o be meaningful, these codes must be modified to prohibit any
device maker payments for restricting purchases from rivals, whether made to GPOs or hospitals,
before or after the GPO contracts, and no matter what their nominal label.” Elhauge, [delete gap
between sentences]
Comments, p. 19. In addition, he would apparently bar all contracts that give hospitals discounts
or rebates in exchange for purchases of 90-95% of their purchases from a single source,
regardless of underlying structural conditions.
Such rules would go far beyond any existing conception of competitive injury
requirements. Further, they would prevent many efficient arrangements. A discount (or rebate,
or side payment, or commission, however denominated) in exchange for exclusivity is nothing
other than exclusive dealing; and a discount in exchange for preferential access or purchases of a
minimum quantity or share are all something less than exclusive dealing. Such arrangements
should be condemned only when the structural conditions for monopoly or anticompetitive
effects are present. This would occur only where the contracts in question covered a significant
portion (typically, 60% or more) of a properly defined relevant market. A discount or rebate in
exchange for a buyer’s commitment to take a large percentage of its goods from a single seller
can be matched by any equally efficient firm, provided that the price after taking off all discounts
or rebates remains above cost. As a result, such practices should ordinarily not even be
condemned in the presence of high market shares unless the net price is also shown to be
No additional legislation or “compulsory” rule making is needed to produce competitive
results in this market. The recent decisions of the D.C. Circuit and the Third Circuit in Microsoft
and Les indicate that the antitrust courts are completely capable of dealing with the
Hovenkamp, Antitrust/GPO Agreements 13
occasional anticompetitive discounting practice when it arises. One of those suits was a
government equity proceeding and the other was a private treble damage action. Additional
action forcing firms not to enter such agreements when the requirements for competitive harm
are not met would deprive hospitals of the benefits of both scale economies and competition in
an industry where sellers are willing to price more aggressively in exchange for higher volume.
Professor Elhauge also fears that nothing in the codes prevent firms from engaging in
“future bundling of multiple products.” But such bundling is fully subject to the anti-tying
provisions of the antitrust laws. Further, there are few people today who believe that the “per se”
rule which governs tying arrangements is underdeterrent. See 9 AREEDA & HOVENKAMP,
ANTITRUST LAW ¶ 1720 (2d ed. 2004). If anticompetitive bundling should occur antitrust tying
law is sufficient to address it. Further, as noted previously, tying law applies to seller of tied-up
combinations, not to buyers, who are not benefited from a reduction of competition in their
suppliers’ markets.
While exclusive dealing, sole-source contracting, quantity or market share discounts and
similar practices are occasionally anticompetitive, the conditions for competitive harm are rarely
satisfied. These conditions are generally not met in the markets in which health care group
purchasing organizations participate. More importantly, all of these practices are cost-reducing
and output increasing most of the time, so competition policy should encourage rather than
discourage them. Finally, all such practices are fully reachable under the existing antitrust laws,
which are more than sufficient to identify and correct the occasional injury to competition that
might occur.