Revisiting the separate products issue.(United States v. Microsoft)

United States v. Microsoft Corp., 147 F.3d 935 (D.C. Cir. 1998)

The Justice Department's tying claim against Microsoft Corporation boils down to one question: Is Internet Explorer a separate product from Windows 95? In United States v. Microsoft Corp.,(1) the government contended that Microsoft violated a consent decree by illegally tying(2) the sale of its Internet browser, Internet Explorer, to the sale of its operating software, Windows 95. The U.S. Court of Appeals for the District of Columbia, however, held that no tying arrangement existed because the two programs were part of a single, "integrated" software product.(3)

The difference between a single product and two tied products is known as the "separate products issue" in tying law. This Case Note argues that the discussion of the separate products issue in Microsoft was flawed on two levels. On a doctrinal level, the court's separate products test was not "consistent with tying law" as claimed.(4) It failed to adhere to the leading U.S. Supreme Court case on the issue, which requires courts to examine the nature of demand for two products rather than their functional relatedness when resolving the separate products question. On an analytical level, the court's test conflated the definitional question of a tie-in's existence with the question of whether a tie-in should be permissible on economic grounds. It thereby unwisely transformed the separate products test into a role-of-reason test, which directs courts to examine the efficiencies and other benefits of a challenged arrangement.(5) The Microsoft court may have manipulated the separate products test in this manner because of a desire to move judicial analysis away from per se condemnation of tying arrangements.

I

The separate products issue in Microsoft first arose when the government commenced a proceeding to hold Microsoft in civil contempt of a 1994 antitrust consent decree.(6) The decree prohibited Microsoft from entering "any License Agreement in which terms of that agreement are expressly or impliedly conditioned upon ... the licensing of any other [Microsoft product] ... (provided, however, that this provision in and of itself shall not be construed to prohibit Microsoft from developing integrated products)...."(7) The government contended that Microsoft had violated this decree by tying the sale of Internet Explorer to the sale of Windows 95. The U.S. District Court for the District of Columbia rejected this argument, however, because it found the consent decree to be vague.(8) Nonetheless, it believed that there was a high probability that Microsoft had violated the Sherman Act by conditioning its license of Windows 95 on its license of Internet Explorer. In response, the court issued a preliminary injunction to prohibit this practice.(9)

A three-judge panel of the D.C. Circuit reversed. It unanimously found that the district court had issued the injunction without providing adequate notice.(10) Two of the judges on the panel further argued that the injunction was invalid because it was based on an erroneous reading of the consent decree. The decree forbade Microsoft from conditioning the sale of one product on the sale of another unless the two products added up to an "integrated product."(11) The two-judge majority argued that Internet Explorer and Windows 95 constituted an integrated product rather than a tie of two separate ones since their combination had "facially plausible benefits to its integrated design."(12) It also contended that this finding was "consistent with tying law," citing Areeda's influential treatise, which explains that "new products integrating functionalities in a useful way should be considered single products regardless of market structure."(13) The majority concluded that the government failed to show a reasonable probability of success in its antitrust suit to warrant a preliminary injunction against Microsoft, because a tying arrangement must at a minimum involve two products.

II

After the decision regarding the consent decree, the Justice Department commenced a Sherman Act proceeding against Microsoft, claiming among other things that the company had illegally tied Windows 95 and Internet Explorer in an attempt to monopolize the Internet browser market.(14) The separate products analysis in Microsoft suggests that the D.C. Circuit may reject this claim if and when it hears the claim on appeal. Even though the consent decree did not embody "the entirety of the Sherman Act or even all `tying' law under the Act,"(15) the Microsoft court acknowledged that the decree "emerged from antitrust claims,"(16) and that its understanding of "integrated products" was "consistent with tying law."(17) This suggests that the D.C. Circuit may apply a similar separate products analysis and hold that Windows 95 and Internet Explorer are two parts of an integrated product rather than two tied products.(18) The possibility that the Microsoft analysis will be used in the Sherman Act proceeding makes it important to consider its substantive weaknesses.

III

A. The Doctrinal Flaw of Microsoft

In Jefferson Parish Hospital District No. 2 v. Hyde,(19) the U.S. Supreme Court held that "the answer to the question whether one or two products are involved turns not on the functional relation between them but rather on the character of demand for the two items."(20) Two products exist when the two product markets are "distinguishable in the eyes of buyers,"(21) and when there is sufficient demand for the tied product separate from the tying product so that it is "efficient" to offer them separately.(22) The functional relatedness of two items is not in itself sufficient to determine whether they constitute one or two products. The Court noted that prior tying cases involved products that were functionally linked, and that this linkage was never sufficient to show that only one product was involved.(23)

The discussion of separate products in Microsoft was at odds with this test. Claiming that its view was "consistent with tying law," the two-judge majority explained that an integrated product is "a product that combines functionalities ... in a way that offers advantages unavailable if the functionalities are bought separately and combined by the purchaser."(24) The majority's separate products test was thus two-pronged. First, an "integrated" product confers benefits that are unavailable if the component parts are bought separately. Second, an "integrated" product provides benefits that cannot be achieved easily by consumers who purchase the components separately. The test ignored Jefferson Parish's directive to examine consumer perception and not functionalities when making the separate products inquiry.

In a separate opinion, Circuit Judge Wald suggested as an alternative that courts balance two factors when resolving the separate products issue: (1) the benefits consumers receive from integration; and (2) the size of consumer demand for the products provided separately.(25) If there is great evidence of distinct markets, the defendant must show great consumer gain to justify integrating the two products.(26) Judge Wald's test more closely resembled that of Jefferson Parish because it examined the separate demand for two products. Nonetheless, even this test was doctrinally flawed. The Supreme Court in Jefferson Parish neither examined integrative efficiencies nor engaged in any type of balancing when it made the separate products determination. Its discussion consisted of little more than finding that the products in question could be offered separately and would be purchased if offered as such. The claimed benefits of an "integrated" relationship were irrelevant to whether the arrangement involved two distinct products.

Both the two-judge majority and Judge Wald seemed to believe that an examination of functionalities and efficiencies was necessary to resolve the consent decree's disparate treatment of MS-DOS/Windows 3.11 and Windows 95. The decree prohibited Microsoft from tying the operating software of MS-DOS with the graphical interface of Windows 3.11. However, it permitted the company to produce Windows 95, which combined functions of MS-DOS and Windows 3.11.(27) The court felt that an examination of integrative efficiencies was necessary to resolve the dissimilar treatment. It reasoned that the consent decree treated MS-DOS and Windows 3.11 as separate products because their combination did not provide significant consumer benefits. By contrast, the decree treated Windows 95 as an "integrated" product because the combination of the component parts conferred substantial efficiencies unavailable if consumers had purchased the parts separately.

The Jefferson Parish test, however, can adequately explain why MSDOS and Windows 3.11 were treated as two products while Windows 95 was treated as one. At the time of the consent decree, a significant number of consumers differentiated between MS-DOS and the graphical interface of Windows 3.11.(28) The presence of competing operating system software to be used with Windows 3.11 also indicated that it was efficient for Microsoft to provide DOS and Windows separately.(29) Under Jefferson Parish, therefore, MS-DOS and Windows 3.11 would have been separate products. By contrast, an insignificant number of consumers distinguished between the operating system and graphical components of Windows 95.(30) This is demonstrated by the fact that no separate operating system software or graphical interface product existed to be used with either component. The Jefferson Parish test would therefore have classified Windows 95 as one product. Because this test can adequately explain why the consent decree treated Windows 95 differently from MS-DOS/Windows 3.11, the Microsoft court should have given it greater deference when the court interpreted the decree.

If the court had correctly applied the Jefferson Parish test, it would have classified Internet Explorer and Windows 95 as separate products. There is no question that buyers differentiate between Windows 95 and Internet Explorer.(31) Furthermore, the fact that a competing Internet browser, Netscape Navigator, has a market share that exceeds forty percent(32) indicates that sufficient demand exists to make it efficient to offer Internet browsers separately from Windows.(33) Finally, Microsoft can easily offer the two products separately because "[s]oftware code by its nature is susceptible to division and combination."(34) In sum, the separate demand for Windows and Internet browsers, the existence of competitors that offer the Internet browser alone, and the ease with which Windows and Internet Explorer can be offered separately suggest that the programs are tied products rather than two parts of a single, integrated product.

B. The Analytical Flaw of Microsoft

It is unclear why all three judges examined the efficiencies and benefits of combining two products together when resolving the separate products issue. On an analytical level, this examination conflated the separate products test with the rule-of-reason test. It transformed a simple definitional inquiry into a question of whether a tying arrangement should be permissible on economic grounds.

The Microsoft court may have engaged in such analysis because the Supreme Court currently applies per se treatment to tying arrangements.(35) By incorporating elements of the rule of reason into the separate products test, the Microsoft court effectively avoided the effects of per se treatment without explicitly rejecting that treatment. The court may have reasoned that the manipulation Of a multi-faceted and somewhat ambiguous separate products test would less likely be challenged than the rejection of a straightforward and clear per se rule.

Many commentators have criticized the application of the per se rule to tying arrangements,(36) and it is unclear whether the Supreme Court will continue to use the rule if the issue comes before the Court again.(37) In the interim, however, the Microsoft court should have recognized that the Jefferson Parish approach to the separate products issue remains governing law. If the court had wanted to advocate applying the rule of reason to tying arrangements, it should have done so explicitly and not have manipulated doctrine. The separate products test should not have become a false tool for examining the benefits and efficiencies of tying arrangements.

IV

Several judges, including Supreme Court Justices, have suggested that they might analyze the separate products issue in a manner similar to that of the Microsoft court. The most notable suggestion is Justice O'Connor's concurrence in Jefferson Parish, where she explained that there is no sound economic reason to treat products separately if there is a substantial economic advantage to selling them as a unit.(38) The Microsoft decision demonstrates the problem with this approach: An examination of efficiencies and advantages conflates the separate products test with the rule-of-reason test.

Even if the Supreme Court eventually decides to apply the rule of reason to tying arrangements, the question of whether one or two products are involved should be kept separate from the question of whether a tie-in is permissible on economic grounds. Collapsing the two would make the rule of reason superfluous: Courts would have to examine the efficiencies of an arrangement at both the separate products determination stage and the rule-of-reason analysis stage. Regardless of the type of analysis applied to tying arrangements, therefore, courts should continue to follow the Jefferson Parish approach to the separate products issue. Jefferson Parish not only provides a clear standard for judges to follow, but it also separates two analytically distinct issues that are often entangled in tying law.

(1.) 147 F.3d 935 (D.C. Cir. 1998). This case preceded the Sherman Act case brought by the Justice Department against Microsoft. See United States v. Microsoft Corp., No. CIV. A. 98-1232, 1998 WL 614485 (D.D.C. Sept. 14, 1998); see also infra Part II (providing a description of how the two cases are related). Unless otherwise indicated, references to the Microsoft case will pertain to the first case.

(2.) In a tying arrangement, a buyer must purchase one product (the tied product) in order to get another (the tying product). See Northern Pac. Ry. v. United States, 356 U.S. 1, 5 (1958).

(3.) Tying arrangements must involve at least two products. See Jefferson Parish Hosp. Dist. No. 2 v. Hyde, 466 U.S. 2, 21 (1984); Fortner Enters. v. United States Steel, 394 U.S. 495, 507 (1969); Times-Picayune Publ'g v. United States, 345 U.S. 594, 613-14 (1953).

(4.) Microsoft, 147 F.3d at 950.

(5.) Depending on the kind of arrangement challenged, courts use the per se role or the role of reason to determine whether an arrangement unreasonably restrains trade. An arrangement analyzed under the per se rule is presumed to be illegal because of a longstanding judgment that it has a substantial negative impact on competition. See FTC v. Superior Court Trial Lawyers Ass'n, 493 U.S. 411, 433 (1990). By contrast, an arrangement analyzed under the role of reason is condemned only after some analysis of its purpose and effect on competition. A key component of this analysis is an examination of efficiencies. If an arrangement is necessary to achieve significant efficiencies in production, distribution, or development, a court undergoing a role-of-reason analysis will balance these procompetitive effects against the arrangement's anticompetitive effects when deciding the legality of an arrangement. See ABA SECTION OF ANTITRUST LAW, ANTITRUST LAW DEVELOPMENTS 69 (4th ed. 1997).

(6.) See United States v. Microsoft Corp., No. CIV. A. 94-1564, 1995 WL 505998 (D.D.C. Aug. 21, 1995).

(7.) Id. at *3.

(8.) See United States v. Microsoft Corp., 980 F. Supp. 537, 541 (D.D.C. 1997).

(9.) See id. at 543-44.

(10.) See Microsoft, 147 F.3d at 943. The Microsoft court felt that the district court had violated Federal Rule of Civil Procedure 65(a)(1), which provides that "[n]o preliminary injunction shall be issued without notice to the adverse party." FED. R. CIV. P. 65(a)(1). It contended that Microsoft had inadequate notice of the preliminary injunction because the government had requested the district court to grant a contempt citation and permanent injunction against Microsoft. See Microsoft, 147 F.3d at 943. The three-judge panel decided that neither of these requests raised the equitable factors of interim irreparable injury that a court must find before granting a preliminary injunction. See id.

(11.) Microsoft, 147 F.3d at 939.

(12.) Id. at 950.

(13.) Id. (citing 10 PHILLIP AREEDA ET AL., ANTITRUST LAW [paragraph] 1746b, at 227-28 (1996)).

(14.) See Trial Transcript, United States v. Microsoft Corp., No. CIV. A. 98-1232, 1998 WL 735825 (D.D.C. Oct. 19, 1998).

(15.) Microsoft, 147 F.3d at 946.

(16.) Id.

(17.) Id. at 950.

(18.) Lower courts are not required to give deference to the Microsoft separate products discussion for the purposes of antitrust analysis because: (1) the discussion was dicta and unnecessary to the decision about the preliminary injunction; and (2) the Microsoft court did not claim that its efficiency test was the proper test for antitrust law generally. Nonetheless, the court did not suggest that its test was an improper test for antitrust law. The court, in fact, believed that its test "was consistent with tying law." Id.

(19.) 466 U.S. 2 (1984).

(20.) Id. at 19.

(21.) Id. (citing Times-Picayune Publ'g v. United States, 345 U.S. 594, 613-14 (1953)).

(22.) Id. at 22. The Court's use of the word "efficient" relates to the size of demand for the products provided separately, not the consumer gain from integrating two functionalities. See infra note 33.

(23.) See Jefferson Parish, 466 U.S. at 19 n.30.

(24.) Microsoft, 147 F.3d at 948, 950.

(25.) See id. at 958-59 (Wald, J., concurring in part and dissenting in part).

(26.) See id. at 959 (Wald, J., concurring in part and dissenting in part). Judge Wald believed that the balancing was a factual issue to be explored by the district court on remand. See id. (Wald, J., concurring in part and dissenting in part).

(27.) See id. at 945-46.

(28.) See, e.g., Rule: Integrating Windows Is Not Predatory, NAT'L L.J., Mar. 2, 1998, at A16 (explaining that consumers probably distinguished between an operating system and a graphical user interface in 1990); see also Howard Marks, Win 95: The Great Migration, LAN MAG., Feb. 1996, at 56, 57 (noting that Windows 3.x is in reality a separate application running under DOS); Stan Veit, What Ever Happened to ... DOS? Product Information, COMPUTER SHOPPER, Dec. 1995, at 628 (noting that Windows 3.11 sits on top of DOS). Microsoft Corporation even marketed DOS and Windows 3.11 as separate products. See Microsoft Raids Firms Misusing Licenses to Pirated Software, SOFTWARE INDUSTRY REP., July 19, 1993, at 2 (noting that Microsoft separately licenses MS-DOS and Windows to computer manufacturers).

(29.) See Microsoft, 147 F.3d at 961 (Wald, J., concurring in part and dissenting in part) (noting that the presence of Novell's operating system product to be used with Windows 3.11 suggests the existence of a separate operating system market from Windows at the time of the consent decree); see also Ron White, Will the Best DOS Please Stand up: Comparison of Microsoft's MS-DOS 6.2, IBM's PC DOS 6.1 and Novell's DOS 7 Operating Systems, PC/COMPUTING, Feb. 1994, at 183 (noting that three competing operating system products existed to be used with Windows 3.11); cf. Eastman Kodak v. Image Technical Servs., 504 U.S. 451, 462 (1992) ("Evidence ... that service and parts have been sold separately in the past and still are sold separately ... is evidence of the efficiency of a separate market for service.").

(30.) Over time, fewer consumers have distinguished between operating software and graphical software. See Rule: Integrating Windows Is Not Predatory, supra note 28, at A16 (explaining that consumers probably differentiated between the graphical interface and operating system in 1990 but no longer do so today); see also Veit, supra note 28 (noting that DOS will gradually fade away because of the growing popularity of Windows 95 and newer Windows systems). The trend suggests that the two items may no longer be separate products for antitrust purposes. Product definitions under Jefferson Parish are time-sensitive because consumer perceptions may change over time. This temporal sensitivity may have the perverse effect of encouraging manufacturers to tie separate products, with the intention that the tie will change consumer perception and transform the tie into a single product. Despite this potential incentive, the Jefferson Parish approach is superior to an examination of integrative efficiencies because it permits consumers, rather than courts, to determine the benefits of tying two products through their purchasing power.

(31.) In general, people distinguish between operating system software, which manages the internal functions of a computer such as storing and retrieving data, and application software, which performs familiar functions such as word processing and spreadsheet operations. See, e.g., Mark A. Lemley & David McGowan, Could Java Change Everything? The Competitive Propriety of a Proprietary Standard, in SECOND ANNUAL INTERNET LAW INSTITUTE 453, 459 (PLI Pat., Copyrights, Trademarks, & Literary Prop. Course Handbook Series No. G0-0001, 1998) (dividing binary code into operating system code and application code); see also Jeffrey M. Shohet, When Microsoft Is Split up, the Public Will Gain, NAT'L L.J., June 29, 1998, at A25 (noting that Internet browsers have been offered as separate products for years and that consumers have "no expectation that browsers would light up the first screen when a user boots up the computer").

(32.) See Netscape Lost Market Share to Microsoft in 1st Half of '98, WALL ST. J., Sept. 29, 1998, at B12 (reporting on a survey that placed Netscape's market share at 42%). Netscape's market share was even greater at the beginning of the Microsoft case. See Kara Swisher, Online: After a Life at Warp Speed, Netscape Logs off, WALL ST. J., NOV. 25, 1998, at B1 (placing Netscape's browser market share at 57.6% in the third quarter of 1997).

(33.) In Kodak, the Supreme Court found that evidence that two products are sold separately in the past and continue to be offered separately is sufficient to establish that it is "efficient" to offer them separately. See 504 U.S. at 462 (noting that "the development of the entire high-technology service industry is evidence of the efficiency of a separate market for service").

(34.) Microsoft, 147 F.3d at 951.

(35.) See Jefferson Parish Hospital Dist. No. 2 v. Hyde, 466 U.S. 2, 9 (1984) ("It is far too late in the history of our antitrust jurisprudence to question that proposition that certain tying arrangements pose an unacceptable risk of stifling competition and therefore are unreasonable `per se.'"). However, the test courts use to determine whether the per se rule should be applied to a particular arrangement increasingly resembles a rule-of-reason inquiry. See NCAA v. Board of Regents, 468 U.S. 85, 104 n.26 (1984) (noting that "while the Court has spoken of a `per se' rule against tying arrangements, it has also recognized that tying may have procompetitive justifications that make it inappropriate to condemn without considerable market analysis").

(36.) See, e.g., Ward S. Bowman, Jr., Tying Arrangements and the Leverage Problem, 67 YALE L.J. 19 (1957) (arguing that the per se treatment of tying arrangements is based on a false notion of leverage); Donald F. Tumor, The Validity of Tying Arrangements Under the Antitrust Laws, 72 HARV. L. REV. 50, 75 (1958) (arguing that cost justifications for tying arrangements indicates that per se treatment of such arrangements is sometimes too harsh).

(37.) See, e.g., Kodak, 504 U.S. at 486-504 (Scalia, J., dissenting, joined by O'Connor and Thomas, JJ.) (arguing that the Court should abandon per se treatment of tying arrangements); Jefferson Parish, 466 U.S. at 35-42 (O'Connor, J., concurring, joined by Burger, C.J., Powell, and Rehnquist, JJ.) (arguing that the Court should apply the rule of reason to tying arrangements).

Only one of the five Justices joining the majority of Jefferson Parish remains on the Court. On the other hand, both Justice O'Connor, who wrote the concurrence, and Chief Justice Rehnquist, who joined it, remain on the Court. Furthermore, Justices Scalia and Thomas, who are on the Court, joined the dissent in Kodak, which suggests that they would follow Justice O'Connor's view of the separate products problem.

(38.) Jefferson Parish, 466 U.S. at 39-40 (O'Connor, J., concurring); see also Jack Walters & Sons Corp. v. Morton Bldg., 737 F.2d 698, 703 (7th Cir. 1984) (noting that "the practice has been to classify a product as a single product if there are rather obvious economies of joint provision"). Justice O'Connor, joined by Chief Justice Burger and Justices Powell and Rehnquist, argued that anesthesia and other hospital services should not be seen as separate products since the advantages of combining the two--advantages such as 24-hour anesthesiology coverage, standardization of procedures, flexible scheduling, and more efficient monitoring--made it reasonable to treat the packaged services as a single product. Justice O'Connor made several valid observations regarding market power and efficiencies. Her analysis, however, is inappropriate for the separate products test in Microsoft for two reasons. First, she assumed that tying law should be subject to the rule of reason. Her analysis is therefore improper for per se analysis, which the Supreme Court currently uses to examine tying arrangements. Second, she collapsed the separate products issue and rule-of-reason analysis into one. Her separate products analysis therefore amounted to an examination of whether a tie-in should be permissible on efficiency grounds.