Claude F. Scott, Esq.
Attorneys for Plaintiff the United States of America
Also filed on behalf of 10 Plaintiff States (See signature block)
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION
Oracle has moved to exclude the testimony of Plaintiffs' economic expert, Professor Kenneth Elzinga, under the authority of Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993), and its progeny. At bottom, Oracle simply disagrees with Professor Elzinga's conclusions, which is not the proper focus of a Daubert inquiry, see 509 U.S. at 595, and its motion should consequently be denied.
Professor Elzinga is one of the nation's leading antitrust authorities. He has served on the faculty of the University of Virginia since 1967, and most of his academic career has been devoted to teaching and research in antitrust economics. He has served in the Antitrust Division of the Department of Justice (DOJ) at the policy level, and has been (and currently is) a consultant to the Federal Trade Commission. Professor Elzinga has lectured to federal judges on antitrust economics, and has recently served as a special consultant to Judge Lewis A. Kaplan in the Christie's- Sotheby's Auction Houses Antitrust Litigation. Professor Elzinga has testified in many antitrust cases and has written numerous scholarly papers on antitrust economics. There is no question here of Professor Elzinga's qualifications.
There is also no question that his opinions are relevant to the disputed issues in the case and probative on those issues. Using the record available, he provides economic analysis and opinions regarding market definition, entry conditions, and the effect the merger would have on competitive conditions in the markets. There is nothing methodologically exotic about this: it is what economists do in every antitrust case, and it is very similar (if not identical) to what Oracle's own economists do here.
Oracle's summary of the Daubert standard is correct as far as it goes, but omits some important qualifications.
First, the Supreme Court decided Daubert in the context of toxic tort litigation, where arguably the most pressing policy concern was that juries would be misled by experts offering novel, untested theories, or even "junk science." See Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 595-97 (1993) (new standard will prevent "befuddled juries"); Elsayed Mukhtar v. Cal State Univ., Hayward, 299 F.3d 1053, 1063-64 (9th Cir. 2002) (Daubert guards against "junk science" and "is particularly important considering the aura of authority experts often exude, which can lead juries to give more weight to their testimony"); Daubert v. Merrell Dow Pharmaceuticals, Inc., 43 F.3d 1311, 1321 n.17 (9th Cir. 1995) (scientific evidence must "not mislead the jury"). In contrast, Dr. Elzinga's opinions in this case rest on economic methodologies over which there is no serious dispute.
Although Daubert also applies in bench trials, the concern that the fact-finder will be misled there carries little weight. See Volk v. United States, 57 F. Supp. 2d 888, 896 n.5 (N.D. Cal. 1999) ("[I]t bears noting that the Daubert gatekeeping obligation is less pressing in connection with a bench trial."); In re Bay Area Material Handling, Inc., 1995 WL 729300, *6 (N.D. Cal. Dec. 4, 1995) (Walker, J.) ("Given the flexible nature of FRE 702 ... and given the fact that the trier of fact in this case was a judge ... there thus was little risk that the expert testimony would be given undue weight"). Accordingly, in a bench trial, probing the nitty gritty of experts' methodologies under Daubert to avoid misleading the court may not be an efficient use of judicial or party resources, because the court can simply hear the testimony and give it the weight it deserves.
Second, to the extent that Oracle's motion is based on objections to the factual foundation of Professor Elzinga's opinions, including Oracle's argument that his data set is flawed, nothing in Daubert or its progeny changed the fundamental rule that the factual basis of an expert opinion "goes to the credibility of the testimony, not the admissibility, and it is up to the opposing party to examine the factual basis for the opinion in cross-examination." Hose v. Chicago Northwestern Transp. Co., 70 F.3d 968, 974 (8th Cir. 1995); accord Southland Sod Farms v. Stover Seed Co., 108 F.3d 1134, 1143 (9th Cir. 1997) ("Technical unreliability goes to the weight accorded a survey, not its admissibility") (citations omitted); Kannankeril v. Terminix Int'l, Inc., 128 F.3d 802, 807, 809 (3d Cir. 1997) (reversing exclusion of expert based on "insufficient factual foundation" and cautioning that the "trial judge must be careful not to mistake credibility questions for admissibility questions").
Understood in light of all of the relevant law relating to exclusion of expert testimony under Daubert, Oracle's motion does not pass muster.
Daubert is about the admissibility of an expert's testimony, not the weight that testimony should carry with the trier of fact. Oracle dresses up its upcoming cross examination of Professor Elzinga as a Daubert motion in an attempt to have the court exclude testimony that would otherwise be helpful to the Court in its analysis of the relevant markets in this case. The economic principles and methodologies applied by Professor Elzinga are the ones cited by Oracle in its court papers, and, under the law should be admitted.
Consistently with the Merger Guidelines and the case law, Professor Elzinga identified the relevant geographic area in which a hypothetical monopolist could profitably raise prices. He explains that foreign vendors that do not have a presence in the U.S. would not be able to intervene if a hypothetical monopolist or cartel were operative in this country. "The geographic market extends to the 'area of effective competition' ... where buyers can turn for alternative sources of supply," Oltz v. St. Peter's Cmty. Hosp., 861 F.2d 1440, 1446 (9th Cir. 1988). In addition, he chose the narrowest workable market that would fit this test, again consistent with the law cited by Oracle (Oracle Mem., at 2-3).
Relevant markets are defined in order to aid a court in determining whether a merger may lessen competition "in any line of commerce or in any activity affecting commerce in any section of the country." 15 U.S.C. §18. The key question in geographic market definition is whether a hypothetical monopolist (or cartel) could increase price within the defined market, or whether customers within the market would likely be protected by turning to additional sellers who are not currently selling within the defined market. Thus, in this case, the relevance of the geographic market question is to ask if customers in the United States would likely protect themselves from a hypothetical monopolist consisting of Oracle, PeopleSoft, and SAP by turning to other firms who currently do not sell within the United States. Professor Elzinga's analysis of the geographic market addresses this question in a mainstream manner, and he correctly observes that no such other firms can be found (and, indeed, Oracle has not identified any).
Moreover, the purpose of calculating market shares is to help assess the competitive conditions within a relevant market. Measuring market shares of sales within the United States is much more likely to reflect customers' preferences in the United States than would shares of sales outside the United States, where customers may have somewhat different preferences.1 If consumers in Europe demand SAP more frequently than Oracle or PeopleSoft, that does not mean that customers in the United States - those protected by U.S. antitrust law - could be protected against a price increase in the United States by shifting to a less-desirable option. Thus, the calculation of market shares for U.S. sales is more likely to reflect the transaction's probable effect on U.S. customers than the world shares that Oracle suggests (although the transaction would likewise be a three-to-two transaction in a world market).
Against Professor Elzinga's mainstream approach, Oracle argues that a worldwide market definition is proper. In high-function software markets, the geographic dimension of competition is described by where the producer has the competitive presence to sell its product, not necessarily where its corporate headquarters is located, or where the programmers show up to work. For this reason, the German firm SAP is located within the U.S. for the purpose of the market definition analysis. Sales of SAP software to U.S. customers are made through [REDACTED] .3 Some SAP software is tailored to the needs of U.S. customers, and it is telling that this software is developed within the U.S. According to SAP, SAP Labs North America, based in Palo Alto, California, is "primarily focused on developing applications for use by SAP's United States customers"4 and, "[w]ith a local presence in Palo Alto,... is much better placed to work with U.S. customers and study local legal publications."5 In short, SAP's substantial presence in the United States is important to SAP's ability to compete for customers across the United States.
Unlike antitrust markets like supermarkets and hospitals, where the markets are highly localized and the cases usually turn on the geographic market definition, here the key determinant of pricing is the competitive bidding among close (but not perfect) functional substitutes. SAP has a presence in this country in the economically meaningful sense of a U.S. sales and support function and product development function, and its success in New Zealand or New Delhi therefore has no bearing on the market definition question. Likewise, SAP's relatively strong competitive presence in Europe will not protect consumers in the United States. As Professor Elzinga has opined:
5/18/04 Rebuttal Report of Kenneth G. Elzinga, at 3 (Exhibit D). Professor Elzinga's geographic market definition is standard practice, and his principles and methodologies are sound.
Oracle's reliance on United States v. Eastman Kodak, 63 F.3d 95 (2d Cir. 1995), is misplaced. That case does not state "the correct analysis" (Oracle Mem., at 4), but merely concluded that the district court did not abuse its discretion in finding a worldwide geographic market on the particular facts presented. See id. at 109. The court even cautioned that "another fact-finder might have weighed the evidence differently." In addition, it apparently was undisputed in Kodak that all of the film brands were essentially indistinguishable in quality. See id. at 99. Professor Elzinga found, supported by an extensive factual record, that the nature of the market here is very different from the market the court examined in Kodak. E.g., 4/26/04 Report of Kenneth G. Elzinga ("4/26 Elzinga Report"), at 44 (Exhibit E) (even "the output of Oracle, PeopleSoft, and SAP is not homogeneous").
Oracle seems to imply that, as a matter of law, the geographic market must be worldwide. That position is erroneous: "the definition of the relevant market is a factual inquiry for the [fact-finder]," not an issue of law. Rebel Oil, Inc., v. Atlantic Richfield Co., 51 F.3d 1421, 1435 (9th Cir. 1995). Kodak does not determine this case or restrict Professor Elzinga's analysis of a different factual record.
Next, Oracle argues that Professor Elzinga's geographic market analysis is inconsistent with the fact that global functionality is one of the features of high-function software. This is not a proper Daubert argument. There is no allegation that his economic principles or methodologies are flawed, but rather that global functionality requires a worldwide market. This argument fails to recognize the competitive significance of software vendors' presence in the United States and ultimately distills to a factual issue to be weighed by the finder of fact.
Next, Oracle makes light of Professor Elzinga's authorship of the "Elzinga-Hogarty" test for geographic market definition. That test has been used by the courts when a firm's ability to respond competitively depends, in part, on the location of production assets and shipment (or patient) flow data. It is not a useful test for the products at issue in this case, because the software is not "shipped" in any economically meaningful sense of the word. Professor Elzinga explained all of this in his deposition, but Oracle chooses to ignore it.6 Again, the "location" of market participants is defined by their ability to undercut the hypothetical monopolist's attempt to increase prices in the United States.
Oracle complains that the market-share data calculated by Professor Elzinga do not match perfectly the alleged relevant markets. Professor Elzinga calculates estimated market shares for the high-function enterprise market by examining new license sales data from a large number of software firms. He excluded from the calculations license revenues of less than $500,000 per customer per year in an attempt to screen out mid-market transactions. The data set available did not indicate whether the license was for a mid-market product or a high-function product, so some proxy was necessary. The $500,000 filter was used as a proxy for high-function software.
It is unremarkable that Professor Elzinga used available data and necessary approximations to calculate market shares. This is done in virtually every antitrust case because data are not necessarily collected in any industry in the form that an experienced antitrust economist would choose. Reasonable approximations are the rule, not the exception. Of course, cross-examination on the reasonableness of the approximations is appropriate, as would be competing market-share calculations from an opposing party's economist.
Indeed, Professor Elzinga's use of the $500,000 filter provides market shares that are only an approximation of the relevant economic parameters in the industry. That is unremarkable; market shares are intended to be only a rough indicator of market power, see, e.g., Ball Memorial Hosp. v. MutualHosp. Ins., Inc., 784 F.2d 1325 (7th Cir. 1986), which is more than sufficient for Daubert purposes. Again, the admissibility of the evidence is at issue here, not the weight that the finder of fact should assign to the evidence.7
Moreover, the proxy used is, if anything, likely biased in favor of Oracle. For example, Microsoft's license sale to [REDACTED] is included in the share data, but as other testimony will confirm, that was not a high-function product sale. Likewise, certain [REDACTED] sales data are included, even though [REDACTED] has only a narrow product niche and cannot undercut a hypothetical monopolist's exercise of market power. Thus, it is unsurprising that Oracle seeks to have the market shares excluded in their entirety--even after making assumptions favorable to the transaction, the shares still demonstrate that the transaction will substantially increase concentration levels in already highly concentrated markets.8
Oracle also complains that Professor Elzinga used a poor-quality data set in calculating market shares, and cites SMS Sys. Maint. Servs., Inc. v. DEC, 188 F.3d 11 (1st Cir. 1999), as support for the proposition that the testimony should therefore be excluded. Once again, Oracle confuses admissibility of the evidence with the weight the evidence might be given by the fact finder. The expert opinion in the SMS case was not "excluded" because of data flaws, but rather was admitted but given no weight because the expert offered "only a bare conclusion" and did not explain why the data was proper to use for the conclusions he drew. See id. at 25. Professor Elzinga, by contrast, explained why he used the data he had, 4/26 Elzinga Report at 46 (Exhibit G) (because "[p]recise data on sales in the relevant markets are not publicly available"); explained what was included in the data (see id.); provided in an exhibit (Ex. 7) (Exhibit G) giving more detailed information on the data and how it was used; and then confirmed the statistical evidence with declaration testimony (see id. at 49-50) (Exhibit H).
Finally, Oracle complains that Elzinga used assumptions in reconciling the sales data from different firms into one useable data set. On cross examination, Oracle is free to try and show how any of the flaws it alleges would bias the results in favor of the Plaintiffs. See, e.g., Southland Sod Farms v. Stover Seed Co., 108 F.3d 1134, 1143 (9th Cir. 1997) (in response to criticisms about questionable assumptions regarding an economist's damages model, the Ninth Circuit said "These asserted defects, however, go to the weight, and not the admissibility, of Wagner's testimony."). These concerns, however, do not justify Oracle's efforts to avoid that cross examination by twisting its arguments into a Daubert motion in an attempt to deny the Court the benefit of Dr. Elzinga's expert testimony.
Oracle's claim that the Herfindahl-Hirschman Index ("HHI") of concentration is solely "designed to indicate the propensity of a market to coordinated effects" (Oracle Mem. at 12; emphasis in the original) is wrong as a matter of economic theory. Oracle quotes from deposition testimony in which Professor Elzinga correctly opines that there is no mechanically precise or "pristine" relationship between the HHI and predicted merger effects in models of differentiated products (5/28/04 Elzinga Dep. 254:10-19, as quoted in Oracle Mem. at 12). Oracle grossly mischaracterizes Professor Elzinga's testimony as being that the HHI is uninformative for unilateral effects analysis. This erroneous claim is Oracle's, not Professor Elzinga's.
Oracle's claim that the HHI is uninformative for unilateral effects analysis is contrary both to the published views of its own expert, Professor Hausman, and to the use of HHIs made by the courts. In support of his conclusion that "AT&T has market power in the interLATA long-distance market," Professor Hausman cites the high HHI in residential long distance. Professor Hausman describes the HHI as a "standard" concentration measure, as indeed it is. (Jerry A. Hausman, "Competition in Long-distance and Telecommunications Equipment Markets: Effects of the MFJ," Managerial and Decision Economics, 1995, vol.16, pp.365-383.) Courts routinely cite HHIs as a basis for liability in merger cases, without any detailed explanation of the particular theory of oligopoly that motivates the judgment. See, e.g., FTC v. Staples, Inc., 970 F.Supp. 1066 (D.D.C. 1997) (HHIs cited in support of a unilateral effects "head to head competition" theory).
Although the HHI does not offer a mechanically precise measure of unilateral effects with differentiated products, or of coordinated effects, the HHI is informative in assessing the likelihood of both kinds of effect. This is reflected in the Merger Guidelines as well as in the way courts have used HHIs in merger analysis. Professor Elzinga's use of HHIs is entirely consistent with this history. Indeed, Professor Elzinga states clearly that he adopted the analytic paradigm of the Horizontal Merger Guidelines ("Guidelines") "to consider the issue of how closely positioned SAP, PeopleSoft and Oracle are relative to other choices in the marketplace," and that, among other things, the data most useful for doing that was "the output of the big three and others purportedly in the market by Oracle." Elzinga Dep. at 254:24 - 255:17 (Exhibit I).
Oracle mischaracterizes Professor Elzinga's expert report on the likelihood of coordinated effects resulting from the acquisition. In his expert report, Professor Elzinga clearly states, "Market concentration is one of the key variables that bears on the likelihood a group of rivals could coordinate their behavior successfully." 4/26 Elzinga Expert Report at 57 (Exhibit J). This is a completely standard economic opinion as Oracle acknowledges. (See Motion, p. 11 at 24-25, "Professor Elzinga's only mention of these standard oligopoly and collusion theories..."). Professor Elzinga goes on to say, "As I showed earlier, the relevant markets are highly concentrated. Further, for many customers, this merger is a 3 to 2 merger (or 4 to 3)." 4/26 Elzinga Expert Report at 57 (Exhibit J). It is hard to understand why Oracle characterizes these opinions as "... very carefully qualified." (Motion, p. 11 at 25). In deposition, Professor Elzinga conceded he would not be able to say that anticompetitive coordinated effects are likely here.
As Oracle notes, the Plaintiffs' Trial Brief does not discuss conventional coordinated effects theory. It is quite another thing, however, for Oracle to suggest that the Daubert line of cases precludes Professor Elzinga from expressing his general concern on this point in his report, or that this expressed concern there somehow discredits his other opinions.
Oracle's Motion to exclude Professor Elzinga's testimony should be denied.
1. "The market concentration measures provide a measure of [possible anticompetitive] effect if each product's market share is reflective of not only its relative appeal as a first choice to consumers of the merging firms' products but also its relative appeal as a second choice, and hence as a competitive constraint to the first choice." Horizontal Merger Guidelines, §2.211
7. Oracle's citation to City of Tuscaloosa v. Harcros Chemicals, Inc., 158 F.3d 548 (1 lth Cir. 1999) is misleading. The expert there was not an economist but a statistician, and his error was to include sales data from Florida when the suit alleged collusive behavior only in Alabama. See id. at 566-67. By contrast, the sales that Oracle criticizes Dr. Elzinga for including in his analysis (Oracle Mem., at 9) are all within the United States, see 4/26 Elzinga Report, at 46, and therefore are within Professor's Elzinga's geographic market and the geographic market alleged by the United States. City of Tuscaloosa does not support Oracle's real objection to Professor Elzinga's analysis, which is that Oracle believes that Lawson, Microsoft, and other sellers should be included in the product market.
8. Indeed, it is telling that Oracle's motion does not contend that the estimated shares are biased in any manner against Oracle. Rather, it simply contends that the estimates do not measure share perfectly, but perfection is not required.