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
Plaintiffs will offer Professor Preston R. McAfee as a testifying expert on the proposed transaction's likely competitive effects. Oracle has filed a "Daubert" motion to exclude the merger simulation set forth in Professor McAfee's expert reports. The merger simulation supports and confirms the primary conclusions that Professor McAfee draws from an extensive series of case studies, statistical regressions, and other analyses; specifically, that Oracle's takeover of PeopleSoft will result in higher prices to customers of high-function FMS and HRM. The simulation is based directly on the manner in which Oracle competes, and it is drawn directly from Professor McAfee's detailed study and analysis of numerous instances in which Oracle has engaged in head-to-head competition with PeopleSoft.
This fact-based analysis of the Oracle-PeopleSoft head-to-head competition - replete in Professor McAfee's expert reports - is a primary basis for his expert opinions. Having analyzed this competition, through a study of Oracle's internal records and other evidence from industry participants, Professor McAfee undertook a statistical regression analysis. The regression analysis used standard economic methods and practice, and it further confirmed and supported his expert opinions relating to competition between Oracle and PeopleSoft. Additionally, having already undertaken the statistical regression analysis, Professor McAfee turned to auction theory, as the realities of the markets make this the appropriate theory to use, to further confirm and support his expert opinions and conclusions.
That Oracle has nonetheless moved to exclude any testimony by Professor McAfee on this merger simulation can be best explained by the fact that: (1) Oracle has misconstrued the basis for Professor McAfee's opinion, although it was fully disclosed in his expert reports and testimony, (2) Oracle has misconstrued the applicability of auction theory in explaining firms' competitive bidding behavior, and (3) Oracle has misconstrued Professor McAfee's objectives and his methods in employing a model to fit the facts of the industry in order to further confirm his expert opinions in this case.
All of Professor McAfee's expert work in this case is based on sufficient facts and data, is the product of reliable principles and methods, and is a result of the application of those principles and methods reliably to the facts of the case. Oracle's motion should be denied.
I. Legal Standard
Expert testimony that is both relevant and reliable is admissible evidence under Federal Rule of Evidence 702. See Kumho Tire Co. v. Carmichael 526 U.S. 137, 147 (1999) (expanding Dauberts liberal application to non-scientific expert testimony); Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579. This requires that the testimony be (1) based on the special knowledge of the expert and (2) helpful to the trier of fact. 1 If it satisfies these requirements, then the trier of fact must decide what weight to accord the expert's testimony. Kennedy v. Collagen Corp., 161 F.3d 1226, 1230-31 (9th Cir. 1998) (reversing exclusion of scientific medical testimony when court improperly ignored expert's reliance on scientific journals and when expert's reasoning was based on reasoning and methodology "of the kind traditionally used by rheumatologists."). "The focus, of course, must be solely on principles and methodology, not on conclusions that they generate." See Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 594-95 (1993); Elsayed Mukhtar. California State Univ., 299 F.3d 1053, 1063 (9th Cir. 2002).
As Oracle points out, courts have, in some instances, excluded expert testimony based on economic models, yet they have done so only where the model's assumptions deviate dramatically from critical real-world facts. See Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039 (8 th Cir. 2000); Heary Bros. v. Lightning Protection Inst., 287 F. Supp. 2d 1038, 1066,1068 (D. Ariz. 2003) (finding that the economic expert's model was "useless to assist the jury" because of a gross divergence between the assumptions in the model and the market facts). However, where the expert testimony and model are derived from the facts and data of the case, and based on reliable principles and methods, there is no basis for exclusion under Daubert.
II. An Auction Theory Merger Model Is a Reliable Method By Which To Analyze the Competitive Effects of the Proposed Acquisition
Auction theory models are relied on by economists to explain the behavior of buyers or sellers in a wide variety of markets. An auction, as the term is commonly used in industrial organization economics, is "a mechanism to allocate resources among a group of bidders. An auction model includes three major parts: a description of the potential bidders (and sometimes the seller or sellers), the set of possible resource allocations (describing the number of goods of each type, whether the goods are divisible, and whether there are legal or other restrictions on how the goods may be allocated), and the values of the various resource allocations to each participant." Milgrom, Paul (2004), "Putting Auction Theory to Work," Cambridge University Press, New York , p. 37. Auctions are not limited to instances in which an an auctioneer calls out for bids, as Oracle tries to suggest. See Oracle Motion at 4:14 -16 and 7. Rather, auction theory models are used by economists to explain the buying and selling behavior. McAfee Dep. Tr. 210:l-14(Exh. A).
Oracle has attempted to criticize Professor McAfee for, inter alia, using auction theory because such simulations predict that prices will rise when mergers reduce the number of competitors in a given market. See Oracle Motion at 3, 7. But merger simulation models are standard in the economic literature and are routinely used by economists, including Oracle's own economic expert. 2 Further, one of Oracle's own experts (Frederick Warren-Boulton) explicitly supported auction theory as the way to analyze this transaction during Plaintiff's investigation of this merger. See McAfee Dep. Tr. 188:11-23 (Exh. A); Frederick Warren-Boulton, Appendix II: An Economic Analysis of oracle's Proposed Acquisition of PeopleSoft, Jan. 30, 2004 , at 12-14 (Exh. B). By applying merger- simulation methodology widely received by professional economists in peer-reviewed publications, Professor McAfee's expert opinion more than meets the threshold requirement established by Daubert and its progeny. As explained in Kumho Tire Co., 526 U.S. 137, 152 (1999), the critical objective of Daubert is "to ensure the reliability and relevancy of expert testimony. It is to make certain that an expert, whether basing testimony upon professional studies or personal experience, employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field.")
III. Auction-Theory Based Merger Simulation Is Based on Sufficient Facts and Data and "Fits" the Realities of This Industry
Professor McAfee's testimony is based on an extraordinary set of business records and data that provide an "unusually clear picture of the nature of competition in the relevant markets." McAfee Rebuttal Report at 9 (Exh. C). After having studied the nature of that competition through case studies, summary statistics, and regressions, Professor McAfee determined that auction theory would most accurately reflect the nature of that competition. Therefore, he used an auction-based merger simulation model to estimate how the proposed acquisition likely would affect prices paid by buyers. McAfee Initial Report at 55 (Exh. D). Professor McAfee concluded from this model that prices would indeed rise post-acquisition. 3
A. Professor McAfee's Testimony Is Based on Sufficient Facts and Data
Professor McAfee chose to apply auction theory after a thorough investigation of competition in the high-function HRM and FMS software markets. See McAfee Dep. Tr. 25:25 to 26:5 (Exh. A). This investigation consisted of, among other things, conducting case studies of individual bidding opportunities pursued by Oracle, derived primarily from the discount request forms this Court ordered Oracle to produce. 4 The discount request forms are replete with examples of (1) buyers winnowing the list of potential bidders to those offering the highest values and (2) Oracle making iterative offers to outbid the competition based on substantial information regarding how buyers value alternative high-function HRM and FMS software products and the identities of competing vendors. These features define the bidding opportunities as auctions.
Professor McAfee found numerous examples of multiple rounds of bidding in Oracle's internal records. 5 One example of the multiple rounds of bidding that characterize the sales process involves the competition for GAF Materials. On May 12, 2003 , Oracle executives approved a 65% discount for GAF Materials. On May 29, 2003 , the Oracle sales representatives asked to increase the discount to 77% because they were "in a head to head battle with PeopleSoft." See ORLITE0070189 (Exh. G). The Oracle sales representatives explained that the "Oracle Account team and Psft account team spent the entire day in (sic) customer's office on 5/28 bidding and counterbidding against each other." ORLITE0070188 (Exh. G). Yet, the next day, the Oracle sales representatives asked that the discount be increased to 85% because "PeopleSoft came back over night with a call from [CEO Craig] Conway with an offer of 1.2 million for licenses and support at 15% to 17%," thus "psft is at 3.240 million and Oracle is at 3.658 million (at 83% discount)." 6
Professor McAfee identifies several examples of buyers winnowing the list of potential bidders to those offering the highest values. In general, by the time vendors make their demonstrations to particular customers, the number of possible vendors is down to between three and five; this number normally drops to two or three by the time price is negotiated. See McAfee Initial Report at 8 (Exh. D) (citing Paul Ciandrini's (Oracle's then Senior Vice President of North American Sales) "Procurement Process" presentation to the Department of Justice (Exh. H)). To take just one example, during Greyhound's recent procurement for HRM, [REDACTED]. See Declaration of [REDACTED], February 6, 2004, at 5 (Exh. I). [REDACTED]. See id. (Exh. I) After these bidders were eliminated, Oracle entered into a fierce price competition with PeopleSoft, as reflected in Oracle's internal business records: "We are under extreme pricing pressure from Ppsft as they were losing this deal and put a screaming deal on the table - with nothing to lose." 7
Professor McAfee also concluded that vendors receive detailed information about customers through the sales process. In particular, when a customer submits a Request for Proposal (RFP) they provide detailed information on their specific requirements. See McAfee Initial Report at 9 (Exh. D) (citing Ciandrini's "Procurement Process" presentation (Exh. H) and Deposition of Keith Block, p. 175 (Exh. K)). Professor McAfee found that throughout the sales process vendors can acquire detailed information, including "the identity of rival suppliers, the bids offered by those suppliers, and specific, non-public characteristics of buyers such as their budgets and installed software." McAfee Initial Report at 32 (Exh. D). The Greyhound procurement is also a good example to demonstrate the information that becomes available to vendors during the sales process. There, Oracle knew that (1) its competition was only against PeopleSoft, (2) Greyhound valued the offer made by Oracle more than that of PeopleSoft, and (3) what PeopleSoft counterbid. All of these realities demonstrate the applicability of auction theory to these markets; and furthermore, support Professor McAfee's complete information assumption.
B. Professor McAfee's Simulation Model Is Reliable and "Fits" the Realities of This Industry
Apparently, Oracle has either failed or refused to understand what Professor McAfee's merger simulation model is intended to do, how it is applied, and the economic theory behind it. Oracle has a "bootstrap" argument: because Oracle inaccurately describes the model and the theory behind the model, Oracle continuously places it in an improper context. Professor McAfee's testimony and analysis falls well within the parameters set forth in Daubert.
Oracle's primary criticism is of Professor McAfee's complete information assumption in the merger simulation model. 8 What Oracle neglects to acknowledge is that complete information is the standard assumption regarding information held by buyers and sellers in merger simulation models. See Gregory J. Werden, "Simulating Unilateral Competitive Effects from Differentiated Products Mergers, " ANTITRUST, Spring 1997; Gregory J. Werden, "Simulating the Effects of Mergers in Differentiated Products Industries: A Practical Alternative to Structural Merger Policy, " 5 geo. mason L rev. 363 (1997); Jerry A. Hausman and Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 GEO. MASON L. REV. 321 (1997); Aviv Nevo, "Mergers With Differentiated Products: The Case of Ready-to-Eat Cereal Industry, " 31 rand J. econ. 395, 416 (2000). Such an assumption is an essential facet for generating a useful economic model. In an article written by the present and a former chief economist at the Federal Trade Commission and the current Senior Economic Counsel at the Antitrust Division, which was cited by Oracle, the authors point out that a "modeler" should not even "aspire" to achieve "a perfect fit between the model and the facts." Gregory J. Werden, Luke M. Froeb, and David T. Scheffman, A Daubert Discipline for Merger Simulation, Antitrust, at 9 (forthcoming Summer 2004). They explained that "[i]f structural models become too complex, through elaborate attempts to fit every detail of an industry, the models are apt to lose their value in merger analysis; they likely impose unreasonable informational demands and may yield no clear predictions." Id. Professor McAfee also noted in his intial report that "[t]his model, like all such models, makes certain simplifying assumptions. The model is designed to capture the essential characteristics of the relevant markets without being overly complex or intractable through an attempt to incorporate every feature of those markets." Expert Report of Preston McAfee at 58, fn. 125 (Exh. D).
Oracle argues, "Professor McAfee is apparently prepared to testify only that his general 'sense' is that the actual effects of this merger will be 'approximately' similar to the results of his simulation model but that he has no actual economic methodology that would allow him to even tell you which direction it goes." Oracle Motion at 14:24 . Oracle has misrepresented Professor McAfee's testimony. Professor McAfee's simulation model based on an assumption of complete information predicts prices will increase, post merger, in a range between 5% and 30%. See McAfee Second Supplemental Report at 3 (Exh. E). Professor McAfee very clearly explains that his simulation allows him to approximate the effects of the merger. 9 See McAfee Dep. Tr. 206: 11-19 (Exh A).
When asked how his prediction would change under the alternative assumption of incomplete information, Professor McAfee explains on two separate occasions in his deposition that the countervailing effects of moving from the complete information to the incomplete information assumption "balance in the overall impact of the merger," McAfee Dep. Tr. 133:5 to 134:8 (Exh. A), and that "the overall effect of the incomplete information alone is somewhat neutral," McAfee Dep. Tr. 189:24 to 190:9 (Exh. A). In other words, Professor McAfee has testified that his approximation of the effects of the merger with complete information is essentially unchanged under the assumption of incomplete information. Apparently, Oracle wants to hold Professor McAfee to a standard of precision under the assumption of incomplete information that was never established or questioned under the assumption of complete information.
Oracle's Motion suggests that the recent economic literature supplies the test for evaluating economic models; nevertheless Professor McAfee's model easily passes this test. According to Oracle's Motion, the decisive test for evaluating Professor McAfee's merger simulation model is whether it accurately explains the past and therefore can be relied on to predict the future. Oracle Motion at 8:1-4. Oracle's Motion again cites the article from Messrs' Werden, Froeb and Scheffman, which defines an economic test for evaluating economic models as whether the model "explain[s] past outcomes of the competitive process reasonably well." Werden et al., supra, at 3. Assuming, arguendo, that this is the decisive test for evaluating the reliability of Professor McAfee's merger simulation model, Oracle's Motion should be denied for the simple, yet wholly decisive, reason that Professor McAfee's merger simulation model explains past outcomes reasonably well.
Professor McAfee's regression results, for example, show that, all other factors being the same, PeopleSoft's presence in past sales opportunities leads to an increase in Oracle's discounts of approximately 8% to 14%. McAfee Initial Report at 53 (Exh. D) and Suppl. Report at 19 (Exh. L). This range of predicted price effects falls well within the range of the predicted price effects from the merger simulation model, which predicts post-merger price increases of 5% to 30%. Thus, Professor McAfee's merger simulation does "explain past outcomes of the competitive process reasonably well." Article at 3. Moreover, Professor McAfee's case studies of Oracle's discount request forms also show numerous instances where Oracle increases its discounts by amounts within the range predicted by the merger simulation model when PeopleSoft is present in the competition. 10 It is clear that Professor McAfee's predictions from the merger model explain the past outcomes in the relevant markets reasonably well and thus passes the test that Oracle deems decisive.
Oracle argues that "[s]ince customers are entirely in control of the procurement process, voluntarily limiting the number of bidders would make no sense." Oracle Motion at 8: 13-14. This, Oracle argues, is irrational firm behavior. 11 Id. at 8:17 . Oracle simply asserts that buyers limit the number of competitors because "they believe they have sufficient negotiating leverage to extract competitive pricing from the vendors even if they negotiate with only one or two vendors at a time." Oracle Motion at 8:19 . Oracle's observations that buyers voluntarily limit the number of bidders, and do so because they believe they have sufficient negotiation power, is not correct.
Here, the evidence demonstrates that bidders are eliminated during the procurement process because either the vendor or the buyer realizes that the vendor cannot meet the buyer's needs, not just for the sake of eliminating them. It is rational firm behavior to eliminate, either through self elimination or buyer elimination, vendors not likely to be selected. Professor McAfee testified that the procurement cycle leads to vendors identifying the viable alternatives. McAfee Dep. Tr. 242:2-9 (Exh. A). Even Mr. Ellison of Oracle has testified than in competing for new business Oracle's salesforce will attempt to determine whether a competitor is a genuine threat. Lawrence Ellison ( January 20, 2004 ) Deposition at 43 (Exh. M).
As Professor McAfee has testified, he has not ignored the presence of buyer negotiating leverage. Indeed, he has directly incorporated it into his analysis. See McAfee Rebuttal Report at 25 (Exh. C). Professor McAfee's model even goes beyond buyer power to incorporate the competitive significance of rival vendors. Once again, Professor McAfee has done this precisely to capture the most basic facts of the market.
For the foregoing reasons, Oracle's motion should be denied.
1. Daubert, 509 U.S. at 589-91: Expert testimony must be (1) based on sufficient facts and data, (2) the product of reliable principles and methods, and (3) the result of the application of those principles and methods reliably to the facts of the case. See FRE 702; Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993); Kumho Tire Co. v. Carmichael, 526 U.S. 137, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999).
2. See infra III.B.I; see also Hausman and Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 Geo. Mason L. Rev. 321, 322, 329-31 (1997) (predicting higher prices when analyzing a merger of two competitors in a market with a differentiated products). All such simulation models predict that horizontal mergers lead to higher prices. McAfee Dep. Tr. 191:16-23 (Exh. A). In principle, merger-specific efficiencies can offset such price increases, although that is not the case here.
4. These studies were further supported by the many declarations submitted by customers in oppostion to Oracle's takeover of PeopleSoft. In addition, by looking at data provided by both PeopleSoft and Oracle, Professor McAfee was able to support the case studies as not just ideosyncratic anecdotes, but meaningful examples of a more widespread phenomenon. .
5. See McAfee Initial Report at 4 (Exh. D) (stating among Professor McAfee's primary conclusions that "[t]he price negotiation stage of the procurement cycle is characterized by multiple rounds of bidding"); see also id. at 11 (Exh. D) (citing Deposition of Paul Ciandrini at 204-205 (Exh. F)). Oracle argues, "Professor McAfee made it clear that his decision to model these procurements as English auctions was based on an assumption that they proceed as 'multiple rounds of bidding.'" Oracle Motion at 10:5-7. This is precisely what Professor McAfee found in his analysis of Oracle's internal business records.
7. ORCL-EDOC-01186388 (emphasis added) (Exh J). In addition to demonstrating that buyers place different values on different vendors (here the highest value being Oracle, second being PeopleSoft), this internal business record also demonstrates the depth of information vendors have in these procurements.
9. Professor McAfee also explained in his testimony why this type of range, spreading across a range of values and a range of HRM and FMS deals: "I think they're going - so there are a variety of customers. We have talked about mid-market. We have talked about the large enterprise customers. I think there are customers who will experience 5 percent, 10 percent, some maybe 20 percent price increases as a consequence of this merger." McAfee Dep. Tr. 254:21 to 255:3 (Exh. A).
10. See, e.g., McAfee First Supplemental Report at 6 (Exh. L) (concluding that an increase in discount from 65% to 85% for GAF Materials can be attributed to competitive pressure from PeopleSoft). In response to Oracle's suggestion that Professor McAfee failed to complete the natural experiment of looking at PeopleSoft's acquisition of JD Edwards, it should be noted that Professor Hausman's calculated price increase for that merger (i.e. 15-17%), see Hausman Report // 12, is a result of a significant mistake in his empirical analysis. Professor Hausman just divided PeopleSoft's market share into share for PeopleSoft and JD Edwards to estimate JD Edwards' share; yet the PeopleSoft share that Hausman divided does not include any sales of JD Edwards. McAfee Rebuttal Report at 25 (Exh. C).
11. Indeed, "Professor McAfee's auction-based model predicts ... [that] .... [m]ore bidders produce lower prices." Oracle Motion at 8:4-6. Actually, more viable bidders produce lower prices, as demonstrated by the discount forms.