ELIZABETH DE COSTER et al., on behalf of themselves and all others similarly situated, v. AMAZON.COM, INC., a Delaware corporation,
CASE NO. 2:21-cv-00693-JHC
UNITED STATES DISTRICT COURT WESTERN DISTRICT OF WASHINGTON AT SEATTLE
July 1, 2025
Plaintiffs,
v.
AMAZON.COM, INC., a Delaware corporation,
Defendant.
I INTRODUCTION
This matter comes before the Court on Defendant Amazon.com, Inc.‘s Motion to Exclude Testimony of Parag Pathak, Ph.D. Dkt. # 230 (sealed). The Court has considered the materials filed in support of and in opposition to the motion, the rest of the file, and the governing law. The Court finds oral argument unnecessary. Being fully advised, for the reasons below, the Court DENIES the motion.
Plaintiffs sued Amazon.com, Inc., claiming that the company violated Sections One and Two of the Sherman Act. Dkt. ## 125 (sealed), 126 (redacted). They contend that Amazon denies customers the “benefits of lower prices and fees” that would arise in a competitive market; and they say Amazon does so by imposing on third-party sellers “Most Favored Nation” policies that cause customers to pay supra-competitive prices. Dkt. # 126 at 9 ¶ 15. Plaintiffs allege that Amazon‘s pricing policies prevent “third-party sellers from offering lower prices off of Amazon, and punish them for violations, which in turn insulates Amazon from competition from low cost, alternative platforms.” Id.
Plaintiffs’ economics expert Dr. Parag Pathak, Ph.D. is the Class of 1922 Professor of Economics at Massachusetts Institute of Technology. Dkt. # 262 (sealed) at 14 ¶ 1.1 He is also a Research Associate at the National Bureau of Economic Research (NBER) and is the founding Director of the NBER‘s working group on market design. Id. Market design is a branch of microeconomics that focuses on the design and performance of market clearing institutions. Id. at 14 ¶ 2.
Dr. Pathak concludes that Amazon‘s anti-discounting policies and practices collectively function as a Platform Most Favored Nation (PMFN) restraint. Dkt. ## 262 (sealed) at 18–19 ¶ 30; 307-1 at 20 ¶¶ 44–54.2 In his report, he says that Amazon‘s conduct prevents price competition with other online retailers, which in turn allows Amazon to charge “monopoly referral fees—i.e., the price of connecting merchants and consumers to each other and
Dr. Pathak determines that the higher fees charged by Amazon results in higher prices for products purchased on Amazon. Id. at 25 ¶ 45. He also applies the model to transactional data provided by Amazon to empirically assess the impact of the anti-discounting policies. Id. at 17 ¶ 42. Dr. Pathak asserts that in a counterfactual world without Amazon‘s anti-discounting policies, increased competition between companies in the Online Retail Marketplaces Market would have resulted 12-20% lower fees, depending on the product category. Id. at 24 ¶ 42. According to Dr. Pathak, Amazon‘s anti-discounting policies have thus harmed the class members. Id. at 26 ¶ 48.
Amazon moves to exclude Dr. Pathak‘s expert testimony. Dkt. # 230. The company challenges Dr. Pathak‘s methodology, arguing that (1) the model Dr. Pathak used is not generally accepted in the field of economics; (2) the model has an extraordinary error rate; (3) the model rests upon unreliable and unfounded assumptions; and (4) the model ignores heterogeneity in sellers’ business strategies. Id. at 8–16. Amazon also contends that Dr. Pathak‘s regression
III DISCUSSION
A. Legal Standards
Federal Rule of Evidence 702 governs the admissibility of expert testimony. Under Rule 702, a witness “who is qualified as an expert by knowledge, skill, experience, training, or education may testify in the form of an opinion or otherwise” provided that
- the expert‘s scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue;
- the testimony is based on sufficient facts or data;
- the testimony is the product of reliable principles and methods; and
- the expert‘s opinion reflects a reliable application of the principles and methods to the facts of the case.
Courts must ensure “that an expert‘s testimony both rests on a reliable foundation and is relevant to the task at hand.” Hyer v. City & Cnty. of Honolulu, 118 F.4th 1044, 1055 (9th Cir. 2024) (quoting Elosu v. Middlefork Ranch Inc., 26 F.4th 1017, 1024 (9th Cir. 2022)). They have “broad discretion” in making such evidentiary rulings. Id. (citing City of Pomona v. SQM N. Am. Corp., 866 F.3d 1060, 1065 (9th Cir. 2017)).
Expert testimony is relevant if it “will assist the trier of fact to understand the evidence or to determine a fact in issue.” Daubert v. Merrell Dow Pharms., Inc. (“Daubert I“), 509 U.S. 579, 589 (1993) (citing
Courts apply four factors in determining whether expert testimony is reliable. These include “1) whether a theory or technique can be tested; 2) whether it has been subjected to peer review and publication; 3) the known or potential error rate of the theory or technique; and 4)”
The proponent of the expert testimony bears the burden of establishing admissibility by a preponderance of the evidence. See Lust v. Merrell Dow Pharm., Inc., 89 F.3d 594, 598 (9th Cir. 1996); see also Qualey v. Pierce Cnty., No. 3:23-CV-05679-TMC, 2025 WL 254810, at *3 (W.D. Wash. Jan. 21, 2025). And courts liberally construe Rule 702 in favor of admissibility. See Daubert I, 509 U.S. at 588; see also Chinn v. Whidbey Pub. Hosp. Dist., No. C20-995 TSZ, 2021 WL 5200171 (W.D. Wash. Nov. 9, 2021).
B. General Acceptance of Economic Model
Amazon contends that Dr. Pathak‘s methodology, derived from a 2016 paper by Andre Boik and Kenneth S. Corts, is unreliable because the model used is not widely accepted in the field of economics. Dkt. # 230 at 8.4 The company asserts that there are no standards for
Plaintiffs respond that Dr. Pathak‘s impact and damages methodology derives from the Boik-Corts model, which methodology shows that PMFNs restrain competition from rival platforms and ultimately raise platform fees and retail prices. Dkt. ## 182-6; 308 at 7. They say that Amazon‘s own expert, Dr. Lorin Hitt, undermines the company‘s argument that the Boik-Corts model is not widely accepted. Dkt. # 308 at 7. They note that Dr. Hitt testified that he reviewed 82 papers and did not know of any that criticized the Boiks-Corts model or suggested that the model contained errors. Id. (citation omitted). As to Amazon‘s argument that the model lacks an “established error rate,” Plaintiffs respond that the Boiks-Corts model is a mathematical model (as opposed to a regression model with a statistical error rate) and that any errors would derive from the model itself. Id. at 9. They underscore that the Boiks-Corts model has been subject to peer-review and no economists have identified errors therein. Id.
In forming his opinion, Dr. Pathak examined the “competition softening and fee-inflating effects of prohibiting merchants from setting lower off-platform prices than prices on Amazon.” Dkt. # 262 at 113 ¶ 284. In his rebuttal report, Dr. Pathak mentions that he selected the Boik-Corts model, as opposed to another economic model, because “it is a robust and widely-used model that aligns closely with the real-world setting in which the conduct [at issue] took place.” Dkt. # 307-1 at 15 ¶ 22. He explains that the model “starts from certain economic axioms (assumptions) and derives results, such as PMFNs have an inflationary effect on prices.” Id. at 15 ¶ 23. Dr. Hitt testified that the 2016 Boik-Corts paper is “one of the papers that you would normally cite if you [were] studying a platform MFN.” Id. at 31 (citing Hitt Tr. 243:13–18).
Moreover, in response to Dr. Hitt‘s contention that there is “no generally accepted consensus that PMFNs necessarily result in higher prices,” Dr. Pathak says that the third-party research papers that Dr. Hitt relies on to support this argument are either support Dr. Pathak‘s finding or are “statistically inconclusive.” Id. at 15–16 ¶ 24. And he says that “all three of the theoretical papers that Dr. Hitt highlights model situations that are very different from the challenged conduct.” Id.
The reliability inquiry focuses on “whether the reasoning or methodology underlying the testimony is scientifically valid.” Daubert I, 509 U.S. at 592–93. As describe above, several factors may bear on the reliability analysis including, “whether the theory or technique enjoys general acceptance within the relevant scientific community.” Hankey, 203 F.3d at 1167 (citing Daubert I, 509 U.S. at 592–94). But as discussed above, “the Daubert factors are exemplary, not constraining.” Murray v. S. Route Mar. SA, 870 F.3d 915, 922 (9th Cir. 2017) (citing Kumho Tire Co., 526 U.S. at 150, 159) (Scalia, J., concurring) (“[T]he Daubert factors are not holy writ[.]“).
Dr. Pathak‘s application of the Boik-Corts model to Amazon‘s transactional data does not render his expert opinion unreliable. The Boik-Corts paper was first published in a peer-reviewed journal in 2016 and, in the words of Amazon‘s expert, has become “one of the papers that you would normally cite if you [were] studying a platform MFN.” Dkt. # 307-1 at 15 (citations omitted). And Amazon does not point the Court to any economic literature describing flaws or errors in the model. See Dkt. # 308; see also Dkt. # 307-1 at 15 n.29 (During his
And Amazon‘s contention that Dr. Pathak improperly extended the Boik-Corts model to the facts of this case is unavailing. Dr. Pathak says that he applied the Boik-Corts model to transactional data provided by Amazon to assess the impact of the company‘s anti-discounting policies. Dkt. # 262 at 24 ¶ 42. He analyzed about 236 million individual items sold on Amazon from May 2017 to July 2023 across 30 different categories. Id. And he explains that although the Boik-Corts model assumes that a merchant is a “monopoly seller” and controls prices across all platforms, he extended the model to consider other economic conditions, including a variant that included “perfectly competitive” assumptions—i.e., the seller faces so much competition for its products that prices are driven down to costs. Id. at 126 ¶ 316; see also App‘x at 276–79 ¶¶ 196–217. Dr. Pathak states that his conclusion was the same regardless of the competitive conditions imposed on the model—anti-discounting policies result in higher prices. Dkt. # 262 at 126 ¶ 316. In sum, Dr. Pathak took a peer-reviewed economic model and applied that model to transactional data provided by Amazon.5
As other courts have noted, “[d]isputes about the . . . faults in an expert‘s decision to use a particular methodology . . . or the lack of textual authority for an expert‘s opinion go to the weight, not the admissibility, of his testimony.” Clarke v. LR Sys., 219 F. Supp. 2d 323, 333
And the cases Amazon relies on are distinguishable. For example, in United States v. Cordoba, 194 F.3d 1053, 1060–61 (9th Cir. 1999), in reviewing the district court‘s decision to exclude polygraph evidence under Rule 702, the Ninth Circuit determined that the district court did not abuse it discretion in determining that the “relevant scientific community did not generally accept polygraph exams as being sufficiently reliable to be used as evidence in a trial.” Id. at 1061. In that case, the court noted that the district court relied on evidence, including a scholarly treatise and testimony from an FBI agent, calling into question the validity and scientific soundness of polygraph exams. Id. Here, there is no such scholarship or testimony.
And the reliability issue in Great American Alliance Insurance Co. v. Sir Columbia Knoll Associates Limited Partnership, 484 F. Supp. 3d 946, 956 (D. Or. 2020), involved a wood scientist providing expert testimony about the rate of wood decay in an apartment building following water damage. Id. The court noted that Columbia Knoll conceded that its expert witness‘s application of the model at issue “ha[d] not been tested for proof of accuracy and there
Last, Otto v. Refacciones Neumaticas La Paz, S.A., DE C.V., No. 16-cv-00451-MMD-WGC, 2020 WL 907560, at *4 (D. Nev. Feb. 25, 2020), involved a Daubert motion to exclude an expert‘s proposed safer, alternative jackleg drill design. Id. There, the plaintiff brought a strict liability claim against the defendant alleging that a design defect in a jackleg drill caused her husband‘s death. Id. at *1. The court, in determining that the expert‘s opinion on a safer, alternative jackleg drill design was unreliable, noted that the expert‘s proposed design had never been peer-reviewed or tested and lacked general acceptance within the relevant mining community. Id. at *4.
Unlike the opinions in the cases discussed above, Dr. Pathak‘s opinion derives from an application of a well-known, peer-reviewed economic model.
C. Error Rate
Amazon contends that Dr. Hitt performed validation tests that demonstrate that Dr. Pathak‘s model “gets it wrong more often than not” with error rates ranging from 60% to 100%. Dkt. # 230 at 9. The company asserts that Dr. Pathak‘s model has a 100% false positive rate because it always concludes that a PMFN is inflating all fees and prices even when analyzing data when no PMFN was in effect. Id. at 11. It also asserts that Dr. Hitt compared the model‘s predictions to real world data following Amazon‘s removal of its Price Parity Policy (PPP) in the United Kingdom and Germany. Id. Amazon says that real-world data shows that consumer
Plaintiffs respond that Amazon‘s assertion that the model has a 100% false positive rate is misleading. Dkt. # 308 at 10. They explain that contrary to Amazon‘s assertion that the model predicts 100% of the time that a PMFN is responsible for inflated fees and prices, the model is not designed to detect the existence of a PMFN. Id. Instead, the model demonstrates how a PMFN increases platform and retail fees and measures whether prices would have been lower absent the PMFN. Id. As to Amazon‘s argument about the model‘s failure to detect real-world changes, Plaintiffs counter that Amazon did not remove the PPP until late August 2013 in Germany and late November 2013 in the United Kingdom, and Dr. Hitt compared prices in September 2013. Id. at 11. Plaintiffs say that Dr. Hitt‘s price comparison is “senseless” because the removal of the PPP in the United Kingdom had not yet occurred, and German sellers had only days or weeks to respond to the changes. Id.6 In response to Amazon‘s argument about marginal costs, Plaintiffs say that because marginal costs were not directly observed in the available data, Dr. Pathak used an “inverse optimization to approximate marginal costs.” Id. at 11–12. Plaintiffs say that, at most, Amazon‘s argument presents an argument that goes to the weight of Dr. Pathak‘s testimony, not its admissibility. Id.
Dr. Hitt mischaracterizes the Boik and Corts model and criticizes the very concept of mathematical reasoning from stated premises, even though he acknowledges that a model “that was proven based on fundamental math axioms” does not “assume its conclusion.” The conclusions of the Boik and Corts model are not “assumed” -- they follow logically from principles and interactions set out at the outset.
Id. Dr. Pathak states that he reviewed the record, and the facts support his conclusion that Amazon‘s anti-discounting policies constitute a class-wide PMFN. Dkt. # 307-1 at 19 ¶ 37. In his report, Dr. Pathak explains the facts that lead him to reach this conclusion. Dkt. # 262 at 16–22 ¶¶ 20–35. And in response to Dr. Hitt‘s argument that he did not perform calibration tests, Dr. Pathak explains that he compared “Amazon‘s US fees against those in other more competitive international markets.” Dkt. # 307-1 at 17 ¶ 27.
That Dr. Pathak‘s model assumes the existence of a PMFN does not automatically render it unreliable. As noted above, Dr. Pathak reviewed the facts and explained his basis for concluding that Amazon‘s anti-discounting policies act as a PMFN. Thus, Amazon‘s argument does not show that the economic model Dr. Pathak used is unreliable. See, e.g., City of Pomona v. SQM N. Am. Corp., 750 F.3d 1036, 1048 (9th Cir. 2014) (“[O]nly a faulty methodology or theory, as opposed to imperfect execution of laboratory techniques, is a valid basis to exclude expert testimony.“). And Amazon can cross examine Dr. Pathak and present contrary evidence regarding the factual basis for his opinion. See Hangarter, 373 F.3d at 1017 n.14.
As to Amazon‘s arguments about real-world data and marginal costs, the Court is not persuaded that this contest between economic experts is best resolved here. Dr. Pathak states in his report that he compared the predicted United States fee outcomes to fee outcomes in
The Daubert inquiry is flexible, and the listed factors do not apply equally to every type of expert testimony. Here, Dr. Pathak‘s conclusions are capable of being tested. And his opinions “are supported by rational explanations which [a] reasonable [person] might accept, and none of his methods strike the court as novel or extreme.” Lappe v. Am. Honda Motor Co., Inc., 857 F. Supp. 222, 228 (N.D.N.Y. 1994). Thus, Amazon raises issues that go to the weight a factfinder should afford Dr. Pathak‘s expert opinion, not its admissibility.
D. The Model‘s Underlying Assumptions
Dr. Pathak‘s opinion assumes the existence of a PMFN. Dkt. # 262 at 15 ¶ 10. Amazon asserts that Dr. Pathak “assumes without justification” that Amazon‘s policies and practices
Plaintiffs counter that Dr. Pathak‘s report describes the facts that support his opinion that Amazon‘s anti-discounting policies act as a PMFN. Dkt. # 308 at 12. As for Amazon‘s argument that Dr. Pathak‘s model does not reflect market realities, Plaintiffs point out that economic models necessarily simplify market complexities, and Dr. Pathak has explained why simplifying these assumptions does not undermine the model‘s conclusions. Id. at 13. And Plaintiffs say that Dr. Pathak has provided reasons to reject Dr. Hitt‘s modifications to the model. Id. at 14.
In his report, Dr. Pathak evaluates (1) the Price Parity Clause, (2) the Select Competitor Featured Offer Disqualification program, (3) the Marketplace Fair Pricing Provision, (4) Amazon‘s Standard for Brands, and (5) the Seller Code of Conduct. Dkt. # 262 at 68–95 ¶¶ 159–241. He discusses these policies, describes how Amazon enforces these policies, and assesses their impact on merchant and consumer conduct. Id.
Rule 702(b) requires that expert testimony be based on “sufficient facts or data“; the rule “is not intended to authorize a trial court to exclude an expert‘s testimony on the ground that the court believes one version of the facts and not the other.” Bosley v. DePuy Synthes Sales Inc., No. C21-1683-MLP, 2023 WL 6038010, at *4 (W.D. Wash. Sept. 15, 2023). Amazon may disagree with the conclusions Dr. Pathak arrived at based on his review of the record, but such a disagreement does not render Dr. Pathak‘s opinion unreliable. See In re Valve Antitrust Litig.,
And Amazon‘s contention that Dr. Pathak‘s model is unreliable because the underlying assumptions do not reflect reality is unpersuasive. To be sure, economic “models must be tethered to theories of liability, fit the case, have a reliable basis, and avoid guesswork. But they will, as any economic model inevitably will, simplify the world.” Maldonado v. Apple, Inc, No. 3:16-CV-04067-WHO, 2021 WL 1947512, at *22 (N.D. Cal. May 14, 2021) (citing Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 563 (1931)). As other courts have noted “every model relies on assumptions and no model can account for every conceivably relevant factor.” In re Folgers Coffee, Mktg. Litig., No. 21-00828-CV-W-BP, 2024 WL 4068851, at *5 (W.D. Mo. July 31, 2024) (quoting S&H Farm Supply, Inc. v. Bad Boy, Inc., 25 F.4th 541, 552 (8th Cir. 2022) (analyzing an expert‘s model on lost wages and noting that the defendant “was free to challenge—in fact, did challenge—[the plaintiffs] assumptions during cross-examination“)). Amazon‘s critiques of the model‘s assumptions go to the weight that should be afforded to Dr. Pathak‘s opinion, not its admissibility.
Furthermore, Dr. Pathak reasonably explains why Dr. Hitt‘s adjustments to the model do not affect his conclusions regarding class-wide injury and damages. In his rebuttal report, Dr. Pathak examines each of Dr. Hitt‘s critiques and says that
these arguments miss [his] model‘s purpose: to examine how a PMFN affects competition between marketplaces. When a dominant marketplace prevents merchants from offering discounts on other marketplaces, merchants have two rational responses. First, they might set identical prices across all platforms. Second, as . . . discussed in [his] opening report, they might abandon multi-homing entirely and sell exclusively through the dominant platform, even if they would have used multiple platforms without the anti-discounting policy. Both imperfect enforcement and merchants’ decisions to use only one platform are consistent with
Dkt. # 307-2 at 61 ¶ 169. Amazon asserts that Dr. Pathak‘s analysis is unreliable, but it has not shown how, considering Dr. Pathak‘s explanations, this disagreement between Dr. Hitt and Dr. Pathak should lead to the exclusion of the latter‘s testimony. See In re Vitamin C Antitrust Litig., No. 05-CV-0453, 2012 WL 6675117, at *5 (E.D.N.Y. Dec. 21, 2012); see also Deutsch v. Novartis Pharms. Corp., 768 F. Supp. 2d 420, 456 (E.D.N.Y. 2011) (determining that the expert “satisfied his burden under Daubert by identifying the alternative causes and providing a reasonable explanation for dismissing specific alternate factors identified by Novartis. . . Novartis’ contention that [the expert] should have controlled for these factors goes to the weight that ought to be afforded to [the expert‘s] findings, not the reliability of his methodology“) (cleaned up).
Amazon appears to ask the Court to take a side in a dispute between experts about complex economic modeling. This is not the proper function of a Daubert motion. This is not a case in which an expert cannot articulate a rationale for his methodology; nor is it a case where the expert‘s rationale is obviously flawed or unreasonable. As demonstrated above, Dr. Pathak has provided explanations for his methodological decisions that are grounded in economic literature. See In re Elec. Books Antitrust Litig., No. 11 MD 2293 DLC, 2014 WL 1282293, at *25 (S.D.N.Y. Mar. 28, 2014) (“A minor flaw in an expert‘s reasoning or a slight modification of an otherwise reliable method does not itself require exclusion; exclusion is only warranted if the flaw is large enough that the expert lacks good grounds for his or her conclusions.“) (internal quotation and citation omitted); In re Vitamin C Antitrust Litig., 2012 WL 6675117, at *4 (in rejecting the plaintiff‘s Daubert challenge to the defendant‘s economics expert, the court noted
And Amazon has not shown that Dr. Pathak‘s choices are unsound or so flawed as to make his opinion unreliable. See In re Valve Antitrust Litig., 2024 WL 4893373, at *5 (“Valve, in making its analytical gap argument, takes an overly exacting view of Rule 702‘s requirements. Dr. Schwartz provides common evidence of the varied ways in which Valve establishes its PMFN expectation. Whether that evidence and Dr. Schwartz‘s conclusions deserve credence is an inquiry for a different day.“).
E. Heterogeneity in Sellers’ Business Practices
Amazon contends that Dr. Pathak‘s model ignores that “sellers price their products using different strategies and face different economic constraints.” Dkt. # 230 at 15. The company says that Dr. Pathak‘s methodology does not account for “focal point” pricing—i.e., a practice in which sellers commonly set prices ending with certain values such as $0.99. Id. Amazon explains that Dr. Pathak‘s model predicts that consumer prices for about 20% of class products change by less than 15 cents. Id. at 16. Thus, according to Amazon, if a seller prefers to set prices ending in 99 cents or $9.99, they will likely not increase their prices based on a small increase in fees. Id.
Plaintiffs counter that Dr. Pathak accounts for focal point pricing in his analysis. Dkt. # 308 at 14. They say that Dr. Pathak explains in his report why the chance of any class member being uninjured due to focal point pricing is trivial, given that most of the products are not focal point priced and most consumers bought multiple products on Amazon. Id.
Focal point pricing occurs when retailers set prices at “focal points,” such as prices ending in 99 cents or a round number. See Dkt. # 262 at 145 ¶ 383; see also Sidibe v. Sutter Health, 333 F.R.D. 463, 495 (N.D. Cal. 2019) (describing focal point pricing as “the practice of
Given an assumption of focal-point bias, it might be possible to identify particular incidents in which a sale could be plausibly argued to have had the same price in the but-for world as it did in the real world, despite the lower referral fee in the latter. But this possibility does not affect [his] conclusion that all or virtually all class members [were] harmed by the conduct because virtually all class members made enough purchases to have overpaid on at least some of them.
Dkt. # 262 at 145–46 ¶ 387. In his rebuttal report, Dr. Pathak makes clear that
The possibility of focal point pricing behavior does not affect [his] conclusion that all or virtually all class members were harmed by the conduct. This is because virtually all class members made enough purchases to have overpaid on at least one of them, even if they were not harmed on purchases of focally-priced items of merchandise. For example, under the conservative assumption that all items ending in 99 cents in the real world would have also been priced at the same 99-cent increment in the counterfactual world, less than 1% of class members under the modified class definition would have escaped injury.
Dkt. # 307-1 at 93 ¶ 266. Thus, Dr. Pathak accounts for focal point pricing and reasonably explains why focal point pricing does not impact his determinations.
And the cases Amazon relies on are distinguishable. For example, in In re Apple iPhone Antitrust Litigation, No. 11-CV-6714-YGR, 2022 WL 1284104, at *8 (N.D. Cal. Mar. 29, 2022), the court, in excluding the expert‘s opinion, observed that the expert‘s “pricing model ignore[d] Apple‘s focal-point pricing and pricing tiers in calibrating but-for pricing.” Id. at *8. As the court noted, the expert “failed to use or address the issue” and “the model d[id] not provide a reliable method for determining but-for pricing in the presence of focal pricing.” Id. And in In re Lithium Ion Batteries Antitrust Litigation, No. 13-MD-2420 YGR, 2018 WL 1156797, at *1 (N.D. Cal. Mar. 5, 2018), the court noted that the expert acknowledged that his analysis could be impacted by focal point pricing strategies “but his analysis did not explain how they would affect his analysis of pass-through or his calculation of damages.” Id. As stated above, Dr. Pathak addresses focal point pricing.
F. Reliability of Dr. Pathak‘s Regressions Analyses
Dr. Pathak also studies how Amazon‘s fees affect merchandise prices by analyzing price changes following Amazon‘s partial fee reduction in 2019 for four product categories: Baby, Health & Personal Care, Beauty, and Furniture. Dkt. # 307-1 at 99 ¶ 284. Dr. Pathak states that the results from his analyses confirmed the model‘s predictions: lower fees lead to lower prices. Id.
Amazon contends that Dr. Pathak‘s regression analyses are unreliable because they rely on a small, unrepresentative data sample. Dkt. # 230 at 16. The company says that Dr. Pathak analyzed only a small percentage of products affected by the fee reductions, amounting to only 0.0001% of the class products. Id. It also argues that Dr. Pathak‘s regressions do not show a relationship between fees and prices. Id. at 17. Amazon says that the regression is unreliable because it assumes that all 2.5 million third-party sellers on Amazon act uniformly in adjusting prices for all products subject to a fee change. Id.
Plaintiffs respond that a difference-in-difference regression analysis is a widely accepted econometric tool that courts routinely allow in antitrust cases. Dkt. # 308 at 16 (citations omitted). And as to Amazon‘s argument that the size of the data sample is too small, Plaintiffs say that Dr. Pathak analyzed all the data available across the four product categories. Id. They also assert that Amazon‘s argument about sample size goes to the weight rather than the admissibility of Dr. Pathak‘s testimony. Id. Plaintiffs also state that Dr. Pathak‘s regression analyses do not assume that all sellers decrease prices when fees decease; instead, they say, the tests confirm this prediction across different product categories. Id. at 17.
Dr. Pathak used a difference-in-difference econometric model to compare the prices of individual goods sold on Amazon to other online marketplaces like Walmart. Dkt. # 262 at 148 ¶ 396. He says that his analysis “supplements and supports the findings of the economic model.”
In his rebuttal report, Dr. Pathak emphasizes that he did not “cherry-pick subsets of the data.” Dkt. # 307-1 at 99 ¶ 285. He says that he “analyzed all available prices in every category where a fee reduction occurred.” Id. Dr. Pathak says that Dr. Hitt‘s observation that the regression model accounts for “0.001 percent of the nearly 240 million unique ASINs sold by 3P sellers on the Amazon Marketplace” is misleading because
[t]he millions of items that Dr. Hitt highlights in these comparisons are not included in [Dr. Pathak‘s] analysis for a simple reason: they did not experience any fee reduction. They are therefore uninformative about whether drops in fees were passed through to prices. [Dr. Pathak] .used all of the available data, between 109,056 and 166,656 observations, for which a fee reduction occurred. This dataset has been sufficient to establish, with high statistical confidence, that the predicted relationship between fee reduction and price decreases does exist.
Dkt. # 307-1 at 114 ¶ 332.
As one court stated in ruling on a Daubert motion to exclude, so long “as a sample is representative—that is, it was not selected in a biased manner—sample size will not skew the results of the analysis.” U.S. Info. Sys., Inc. v. Int‘l Bhd. of Elec. Workers Loc. Union No. 3, AFL-CIO, 313 F. Supp. 2d 213, 232 (S.D.N.Y. 2004); see also In re Countrywide Financial Corp. Mortgage-Backed Securities Litig., 984 F. Supp. 2d 1021, 1034 (C.D. Cal. 2013) (“[A]
Moreover, whether the results are statistically significant is testable. Dr. Pathak notes that he tested for statistical significance using measures such as the t-statistic. See Dkt. # 307-1 (App‘x A) at 138 ¶ 8.7 There is nothing to suggest that Dr. Pathak selected the data in a biased manner; instead, he appears to have analyzed all the data available to him. And Dr. Pathak performed these regression analyses on available empirical data to corroborate the conclusion of his economic modeling. See Teradata Corp., 124 F.4th at 568 (in determining that a district court abused its discretion in excluding an expert witness‘s qualitative analyses, the Ninth Circuit noted that “those analyses were merely confirmatory, any flaws they might have would not be a sufficient basis to exclude his tying-market testimony“); Obrey v. Johnson, 400 F.3d 691, 695 (9th Cir. 2005) (“[O]bjections to a study‘s completeness generally go to the weight, not the admissibility of the statistical evidence and should be addressed by rebuttal, not exclusion.“) (cleaned up).
And as to Amazon‘s second argument, Dr. Pathak explains that his regression model does not assume that all sellers decrease prices when fees decrease. Dkt. # 307-1 at 101–02 ¶ 293. He says that the
empirical model must analyze many items of merchandise together to isolate a price effect precisely because of the variation that Dr. Hitt mentions. Each individual item of merchandise is affected by idiosyncratic pricing effects unrelated to the conduct, in addition to the effect of the fee change itself. [His] model acknowledges that prices are driven by many factors unconnected to fee changes.
Dkt. # 307-1 at 102 ¶ 296. Dr. Pathak also addresses each of Dr. Hitt‘s critiques of his regression analyses and explains why these criticisms do not impact his findings. Dkt. # 307-1 at 99-117 ¶¶ 284–342. And Amazon has not shown that Dr. Pathak‘s methodological decisions are so flawed as to make his opinion unreliable. See In re Digital Music Antitrust Litig., 321 F.R.D. 64, 75 (S.D.N.Y. 2017) (“As long as an expert‘s scientific testimony rests upon ‘good grounds, based on what is known,’ it should be tested by the adversary process—competing expert testimony and active cross-examination—rather than excluded from jurors’ scrutiny for fear that they will not grasp its complexities or satisfactorily weigh its inadequacies.“) (quoting Ruiz-Troche v. Pepsi Cola of Puerto Rico Bottling Co., 161 F.3d 77, 85 (1st Cir. 1998)).
And the cases Amazon relies on are distinguishable. For example, in In re Graphics Processing Units Antitrust Litigation, 253 F.R.D. 478, 494 (N.D. Cal. 2008), an expert‘s correlation and regression models used the average prices paid by consumers. Id. at 493–95. As the court noted, the expert‘s report did not say how “specific product pricing was correlated across buyers or whether prices paid for multiple products by particular direct purchasers were correlated.” Id. at 493. It further reasoned that “[i]f data points are lumped together and averaged before the analysis, the averaging compromises the ability to tease meaningful relationships out of the data.” Id. And in In re Pharmacy Benefit Managers Antitrust Litigation, No. CV 03-4730, 2017 WL 275398, at *20 (E.D. Pa. Jan. 18, 2017), the court rejected an expert‘s use of national averages in his regression model because “averages cannot demonstrate antitrust impact for individual class members.” Id. The court noted that the regression model was unreliable because by analyzing only average prices the model found damages for class
Thus, again, Amazon‘s concerns thus go to the weight that should be afforded to Dr. Pathak‘s opinion, not its admissibility. Dr. Pathak‘s methodology can be tested through the adversary process—competing expert testimony, other contrary evidence, and the “crucible of cross-examination.” See Encompass Ins. Co. v. Norcold, Inc., No. 2:23-CV-231, 2025 WL 36025, at *3 (W.D. Wash. Jan. 6, 2025) (reasoning that any alleged shakiness in the expert witness‘s opinion “should be addressed through the crucible of cross-examination and the adversarial process“).
IV CONCLUSION
Based on the above, the Court DENIES Amazon‘s motion to exclude testimony of Dr. Parag Pathak, Ph.D. [REDACTED]
Dated this 1st day of July, 2025.
John H. Chun
United States District Judge