Any antitrust economist

Pooling across heterogeneous products inflates the average

The plaintiffs' pooled regression aggregates transactions across cylindrical cells that differ materially in chemistry, capacity, and form factor, and across a class period spanning more than a decade during which nominal cell prices fell sharply in a rapidly scaling industry. Pooling these transactions without validating the homogeneity assumption inflates the average overcharge and masks individual variation.

A substantial share of the class shows no injury under corrected assumptions

My event study finds that when I correct the plaintiffs' model for cost-elasticity assumptions applied outside the conspired input, a substantial share of class transactions yield zero or negative overcharges, indicating that a nontrivial portion of the proposed class was not injured by the alleged conspiracy under any economically defensible specification.

An average overcharge is not a showing of individual injury

The plaintiffs' model produces an average overcharge figure that does not translate into actual damages for individual class members; an aggregate or average estimate cannot substitute for a showing that each class member paid a supracompetitive price.

The model does not separate the cartel from the secular price decline

The plaintiffs' model does not adequately distinguish the effect of the alleged price-fixing from the secular price decline driven by rapid capacity expansion and technology improvement in the cylindrical-cell industry over the class period, a confounding factor that the regression must control for but does not isolate cleanly.

Legacy contract pricing generates false positives

My analysis identifies a false-positive problem: certain transactions priced under pre-conspiracy contractual formulas appear in the class data and generate apparent overcharges that are sections of the pricing formula rather than evidence of conspiracy impact, inflating the share of class members the model classifies as injured.

Reasoning arc

I begin from the observation that the plaintiffs' model is a pooled regression that assumes a common overcharge across all cylindrical-cell transactions in the class. That assumption does the heavy lifting in the common-impact analysis, and it is untested. Cylindrical cells are heterogeneous products sold under varied contractual arrangements over a period when industry prices were falling for reasons unrelated to any conspiracy. When I correct the model's cost-elasticity assumptions and account for transactions priced under pre-conspiracy formulas, the overcharge estimate is not robust: a meaningful share of transactions show zero or negative overcharges. That finding does not require me to name the correct overcharge; it is sufficient to show that the plaintiffs' model cannot reliably establish that all or nearly all class members were injured by a common method. The guilty pleas by two defendants establish that a conspiracy existed, but they do not establish the magnitude of the overcharge or that every direct purchaser paid a supracompetitive price on every transaction. The overcharge magnitude and common-impact questions remain genuinely contested, and the plaintiffs' model as specified cannot resolve them.

Caveats and limits, stated by the witness

The expert states its own vulnerabilities before opposing counsel does, which is what makes the cross-examination section that follows worth reading:

• My critique does not offer an alternative overcharge estimate or a but-for price series. I have not quantified what the overcharge would be under a correctly specified model, and I do not contend that the overcharge was zero.
• My finding that a substantial share of class members show zero or negative overcharges depends on my own corrections to the plaintiffs' model assumptions. If the court or a trier of fact accepts the plaintiffs' original specification, the uninjured-member count will differ.
• I have not run a formal Chow test or equivalent structural-break test to establish that pooling is statistically inappropriate; my pooling objection rests on economic reasoning and sensitivity analysis rather than a formal hypothesis test.
• Whether my critiques go to the admissibility of the plaintiffs' model or only to its weight at the certification hearing is a legal question for the court; I express no opinion on that question.
• The opinions I rely on for methodological benchmarks were decided before the December 1, 2023 amendment to Federal Rule of Evidence 702, which clarified that the proponent must establish admissibility by a preponderance of the evidence; their precedential weight on admissibility questions is therefore somewhat reduced.