Models on trial: antitrust experts face Daubert challenges

ABSTRACTEconomists are often called upon as expert witnesses by the parties involved in antitrust litigation. One challenge they may face in US federal courts is compliance with the Daubert standard of admissibility of expert testimony. The interplay between model applicability and the Daubert standard is analyzed, suggesting the importance of distinguishing between weak applicability claims, those that state that a model’s critical assumptions are shared by the target, and strong applicability claims, those that connect empirical models and quantitative market features. Recent antitrust cases in which expert testimonies based on economic models have been assessed following the Daubert standard are examined using this framework. Some normative implications are drawn concerning how to improve judges’ assessment of model-based arguments.KEYWORDS: Modelsantitrust lawDaubertmodel applicabilityexpert testimonyJEL CLASSIFICATION: B41K21L41L44 AcknowledgementsI wish to thank Nicola Giocoli, Emrah N. Aydinonat, Uskali Mäki, Steven G. Medema, Kevin D. Hoover, as well as two anonymous reviewers, for their helpful comments and suggestions. This work has been presented in seminars at the TINT Centre for Philosophy of Social Science, University of Helsinki, and Center for the History of Political Economy, Duke University. I thank the organizers and all participants for lively discussions and useful remarks.Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 In addition to antitrust law, other legal areas where economic reasoning finds an eminent place are securities regulation, employment discrimination, and patent infringement (see Chassonnery-Zaïgouche, Citation2020; Maas & Svorenčík, Citation2017).2 The name originates from Daubert v. Merrell Dow Pharm. Inc., 509 US 579 (1993). See below, Section 2.3 Rule 702 – Testimony by Expert Witness. Retrieved from: https://www.law.cornell.edu/rules/fre/rule_702.4 These findings should however be taken with a certain degree of caution for it is extremely hard to build a dataset containing all and only Daubert-related challenges (see Langenfeld & Alexander, Citation2011, Appendix).5 The Daubert Tracker website reports that economics ranks fourth among the most challenged discipline under Daubert/Rule 702 after medicine, engineering, and psychology. See www.dauberttracker.com (accessed May 15, 2023).6 Merger simulation has been introduced as a methodology for antitrust enforcement in the mid-1990s and it is now considered a standard tool by antitrust authorities. For a survey on merger simulation, see Budzinski and Ruhmer (Citation2010).7 Stanford economist Robert Hall saw his testimony rejected in Concord Boat Corp v. Brunswick (207 F.3d 1039, 8th Cir., 2000) because the Cournot model proposed to describe the relevant market did not fit economic realities (see Werden et al., Citation2004, p. 89). Three years later, a federal judge excluded the plaintiff’s expert following a Daubert challenge on similar grounds in Heary Bros. Lightning Prot. Co., Inc. v. Lightning Prot. Inst., 287 F. Supp. 2d 1038 (D. Ariz. 2003). See infra, Section 6.8 The classical reference on the topic of the realism of assumptions in economics is Friedman (Citation1953) and the vast debate following it (see Mäki, Citation2009).9 Economists usually refer to the procedure of building empirical models of such kind as structural modelling approach (e.g. Reiss & Wolak, Citation2007).10 In the field of industrial economics, the distinction between game-theoretic models that represent market competition in the abstract and their empirical counterparts, which result from parametrizing such models using real-world data, is relatively clear. However, when we consider economics in a broader context, delineating a precise line between theoretical and empirical models becomes a challenging task and it remains an open question in the philosophy of economics.11 Section 5 infra will discuss a concrete example of an empirical model designed by the economic expert witness in an antitrust dispute.12 The three kinds of evidence are not mutually exclusive. For a concrete example of an expert witness using both indirect evidence and sensitivity analysis, see Section 5 below.13 It is important to note that the weak applicability of a theoretical model does not guarantee the strong applicability of an empirical model derived from it. So, that the critical assumptions of the Bertrand model are shared by the Texas fruit market does not imply that the predictions of the empirical model will be accurate (for example, if demand elasticities are poorly estimated). Therefore, once federal judges are satisfied with the weak applicability claim, they proceed to examine the strong applicability claim (see infra Sections 5 and 6).14 Castro v. Sanofi Pasteur Inc. 134 F. Supp. 3d 820 (D.N.J. 2015). I accessed the court’s opinion, legal briefs, motions, and other unpublished material through the legal database LexisNexis.15 Castro, Exhibit 51B. Retrieved from: https://advance.lexis.com/api/document?collection=briefs-pleadings-motions&id=urn:contentItem:5KS1-D4J1-JNCK-21KY-00000-00&context=1516831.16 For a textbook treatment of the model, see Tirole (Citation1988).17 To be precise, he identified such portion of the market in the Federal Supply Schedule (FSS), viz. long-term contracts with companies that provide access to commercial products and services at regulated prices to the government. As explained by Elhauge (Exhibit 51B), he used data on FSS purchasers because ‘FSS customers were not subject to the Bundle and thus their decisions could not be distorted by the Bundle’ (p. 190).18 Castro, Reply Brief. Retrieved from: https://advance.lexis.com/api/document?collection=briefs-pleadings-motions&id=urn:contentItem:5YX8-9GF1-FBV7-B32M-00000-00&context=1516831.19 District judge Colleen McMahon’s recent comment on a Daubert motion provides an example. ‘This is a classic example’, she wrote, ‘of what this Court calls, “That expert’s testimony hurts our case, so let’s try to disqualify the expert” use of Daubert. A Daubert inquiry is designed to weed out unreliable methodologies – “junk science” – not to be a substitute for cross-examination about the validity of an analysis that uses established methodologies. […] Economic bargaining models have been upheld as reliable in other antitrust cases. […] Defendants are free at trial to explore Dr. Vogt’s application (or misapplication) of the model’ (In re Namenda Indirect Purchaser Antitrust Litig., 338 F.R.D. 527 (S.D.N.Y. 2021), p. 546).20 In a section called ‘Mr. Guth misapplied the Cournot model’, judge Roslyn Silver wrote that ‘if firms compete on price, the Cournot model does not apply. It is undisputed that LPS firms compete on the price, not quantity, because they compete by price bidding. […] thus the Cournot model does not fit the economic reality’ (Heary Bros. Lightning Prot. Co. v. Lighning Prot. Inst., 287 F. Supp. 2D 1038 (D. Ariz. 2003), p. 1060). Likewise, in Food Lion, LLC v. Dean Foods (In re Se. Milk Antitrust Litig.), the district court excluded the testimony of the plaintiff’s economic expert. The court claimed that the expert’s Cournot model failed to consider crucial commercial realities of the relevant market (2012 U.S. Dist. LEXIS 37650, 2012-1 Trade Cas. (CCH) P77,948, 2012 WL 947106, U.S. Dist. Ct. E.D. Tenn., Greenville Div. March 20, 2012, Filed).21 In re Google Play Store Antitrust Litig., 2022 U.S. Dist. LEXIS 213670, 2022 WL 17252587 (United States District Court for the Northern District of California November 28, 2022, Filed).22 The pass-through rate, which is determined by estimating the market demand, reveals the extent to which App developers pass on the supracompetitive costs imposed by Google to end consumers.23 ‘Dr. Singer posits, without objection by Google, that the Android App Distribution Market is a two-sided market. […] Dr. Singer said that he used for the two-sided Android App Distribution Market the Rochet–Tirole model. […] the “pass-through” rate is a critical element of Dr. Singer’s overcharge analysis, and is the main point of contention in the Daubert dispute. […] Overall, Google has not demonstrated that unreliability or invalidity warrant exclusion of Dr. Singer’s opinions’ (In re Google Play Store Antitrust Litig., p. 8 ff.).24 Roughly speaking, a Daubert hearing is efficient when the court successfully manages three often conflicting needs: excluding unreliable scientific evidence, upholding the jury’s fact-finding role, and minimizing the time and expenses incurred by all parties involved.25 Additionally, my suggestion aligns with the views of legal scholars such as Bartholomew (Citation2014) and Haw Allensworth (Citation2012), who have argued that scrutinizing the specifics of empirical models exceeds the original intent of the Daubert ruling.Additional informationNotes on contributorsEdoardo PeruzziEdoardo Peruzzi is a PhD candidate in Economics in the programme of the Tuscan Universities (Florence, Pisa, and Siena). He works at the intersection between philosophy of economics, antitrust economics, and history of economics. His current project concerns the use of economic theory in antitrust enforcement. Personal website: https://sites.google.com/view/edoardoperuzzi/homepage.