Searching over 5,500,000 cases.


searching
Buy This Entire Record For $7.95

Download the entire decision to receive the complete text, official citation,
docket number, dissents and concurrences, and footnotes for this case.

Learn more about what you receive with purchase of this case.

Kleen Products LLC v. International Paper

United States District Court, N.D. Illinois, Eastern Division

May 31, 2017

KLEEN PRODUCTS LLC, Individually and on Behalf of all those similarly situated, Plaintiffs,
v.
INTERNATIONAL PAPER, et al., Defendants.

          MEMORANDUM OPINION AND ORDER

          Harry D. Leinenweber, Judge.

         Before the Court are Motions to Exclude the testimonies of four of the Plaintiffs' experts and seven of the Defendants'. For the reasons stated herein, the Court denies Plaintiffs' Motions to Exclude the testimony of Kevin Murphy [ECF No. 1101], Steven Davis [ECF No. 1100], Robert Topel [ECF No. 1103], John Huber [ECF No. 1109], Mark Ready [ECF No. 1110], and Donald Skupsky [ECF No. 1111]. It likewise denies Defendants' Motions to Exclude the testimony of Mark Dwyer [ECF Nos. 1082 and 1089], Douglas Zona [ECF Nos. 1104 and 1090], and Michael Harris [ECF Nos. 1125 and 1094]. It grants in part and denies in part the Motion to Strike Lawrence Cunningham's opinion [ECF No. 1096]. It grants Plaintiffs' unopposed Motion to Exclude part of the testimony of Dennis Carlton [ECF No. 1105].

         I. BACKGROUND

         This case is an antitrust class action in which Plaintiffs accuse Defendants International Paper, Temple-Inland, Georgia-Pacific, Westrock (f/k/a Smurfit-Stone or RockTenn), and Weyerhaeuser of conspiring to fix prices. Plaintiffs were direct purchasers of containerboard products from Defendant paper companies. They allege that, in between February 15, 2004 and November 8, 2010 (“the Class Period”), Defendants engaged in a series of agreed-upon actions to raise the price of containerboard products. These include lock-step announcements of price increases and reductions in the supply of containerboard achieved by “cutting capacity, slowing back production, taking downtime, idling plants, and tightly restricting inventory.” Kleen Prods. LLC v. Int'l Paper, 306 F.R.D. 585, 589 (N.D. Ill. 2015) (internal quotation marks omitted).

         To support their contention that these actions were the result of an illegal agreement, or conspiracy, and not legal tacit collusion, Plaintiffs bring evidence that Defendants had the motive, means, and opportunity to conspire. Crucially, much of this evidence comes in the form of expert testimonies. For example, Plaintiffs rely on experts who would testify that Defendants operated in a concentrated industry where “a conspiracy among the Defendants could succeed in increasing prices”; that Defendants used their public announcements as a means to coordinate their price increases and supply reductions; and that Defendants had many opportunities to come to an agreement, as they attended the same trade shows and had documented contacts with each other. See, Kleen Prods. LLC v. Int'l Paper Co., 831 F.3d 919, 922 (7th Cir. 2016). Plaintiffs also rely on their experts to calculate the amount of damages they say they suffered as a result of the elevated prices.

         Defendants counter with their own expert testimonies. Defendants' experts generally opine that Defendants' actions were consistent with actions taken in unilateral self-interest and inconsistent with conspiracy. They offer direct rebuttals to Plaintiffs' expert testimonies, pointing out the weaknesses in their counterparties' methodologies and calling into question their conclusions. Primarily on the strength of these rebuttals, Defendants seek to bar Plaintiffs from making use of their experts at summary judgment or trial. Plaintiffs respond in kind, asking the Court to strike Defendants' expert reports.

         Both Plaintiffs and Defendants argue that the other side's proffered expert testimonies do not pass muster under the standard set in Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993). This is the Court's first opportunity to apply Daubert scrutiny to the parties' expert reports, even though some of the reports had been introduced into evidence earlier for the purpose of class certification. As the Seventh Circuit noted when it affirmed this Court's decision to certify the class, Defendants did not challenge Plaintiffs' expert testimonies at the certification stage. See, Kleen Prods., 831 F.3d at 922. As such, both the Court and the Seventh Circuit took the evidence at face value. See, Id. The case no longer allows for that luxury, and the Court must decide whether the parties' expert testimonies are admissible under Federal Rule of Evidence 702 and the Daubert line of cases interpreting that rule.

         II. LEGAL STANDARD

         While the details of the standard for the admission of expert testimony are fleshed out as the Court examines the specific arguments raised in regard to each expert's testimony, the Court here notes the general outlines of that standard. First, before it may admit any expert's testimony, the Court “must ascertain whether the expert is qualified, whether his or her methodology is scientifically reliable, and whether the testimony will ‘assist the trier of fact to understand the evidence or to determine a fact in issue.'” Bielskis v. Louisville Ladder, Inc., 663 F.3d 887, 893 (7th Cir. 2011) (quoting Fed.R.Evid. 702). Second, even evidence that meets this standard may be excluded under other rules of evidence, most notably Rule 403's weighing of the probative value of the evidence against the danger of the expert misleading or confusing the jury. See, Daubert, 509 U.S. at 595. Third, the Court must make sure not to abrogate the role of the jury as it examines the admissibility of the evidence. See, Bielskis, 663 F.3d at 894. In particular, “[t]he soundness of the factual underpinnings of the expert's analysis and the correctness of the expert's conclusions based on that analysis are factual matters to be determined by the trier of fact, or, where appropriate, on summary judgment.” Smith v. Ford Motor Co., 215 F.3d 713, 718 (7th Cir. 2000). Thus, even when the Court is convinced that one side's experts have the better argument than the other, it is to let “[v]igorous cross-examination, presentation of contrary evidence, and careful instruction on the burden of proof” expose the weaker side's “shaky but admissible evidence.” Daubert, 509 U.S. at 596.

         The Court exercises discretion in deciding to admit or exclude evidence under this standard. See, GE v. Joiner, 522 U.S. 136, 138-39 (1997) (holding that abuse of discretion is the standard an appellate court should apply in reviewing a trial court's decision to admit or exclude expert testimony under Daubert). Ultimately, it is the proponents of the evidence who must persuade the Court by the preponderance of the evidence that the expert testimony should be admitted. See, Lewis v. CITGO Petroleum Corp., 561 F.3d 698, 705 (7th Cir. 2009).

         III. ANALYSIS

         Many of the experts in this case offer testimonies that either overlap with or directly respond to the opinions of other experts. The Court thus begins the analysis with the economic experts who opine on broad issues of liability and damages, follows the thread to those who build on or challenge those conclusions, and ends with experts who testify as to specific facets within the broader framework.

         A. Mark Dwyer

         Mark Dwyer (“Dwyer”) is Plaintiffs' main damages expert. He offers a calculation on the amount of damages that Defendants owe to Plaintiffs, assuming that Defendants are found liable. While Dwyer insists that he does not give an opinion as to whether Defendants indeed are liable, he nonetheless says that the amount of damages he arrives at - a large, positive number of roughly $3.9 billion - is “consistent” with Defendants' having engaged in a conspiracy. Dwyer's opinion is rebutted by Defendants' primary economics expert, Kevin Murphy (“Murphy”).

         The methodology Dwyer used to arrive at his damages number was discussed at length in the Court's opinion on class certification. See, Kleen Prods., 306 F.R.D. at 603-05. Dwyer has since supplemented his reports to respond to various criticisms, but he stands steadfast by his original method. The Court here recaps the essential details of his analysis as well some of the ways in which it falls short as pointed out by Murphy.

         Dwyer's ultimate goal is to quantify how much Plaintiffs were harmed by having to pay artificially inflated containerboard prices during the Class Period. To do so, he regresses the price of containerboard products on a set of control variables and the variable of interest, a Class Period Dummy. The technique that Dwyer uses is called a regression analysis. See generally, Daniel Rubinfeld, Reference Guide on Multiple Regression, in Reference Manual on Scientific Evidence (3d ed. 2011). The price of containerboard is called the dependent variable; and the control variables and Class Period Dummy are variously referred to as regressors, explanatory variables, or independent variables. See, Id. at 352-56. Running a regression produces a set of estimated coefficients on the independent variables, where the coefficients describe a “line” having the property that the sum of the squared differences between the line and the dependent variable is as small as possible. See, ATA Airlines, Inc. v. Fed. Express Corp., 665 F.3d 882, 890-91 (7th Cir. 2011) (outlining the basics of a regression and citing sources). The coefficients are interpreted as the effects that the independent variables have on the dependent variable. See, id.

         In this case, the estimated coefficient on the Class Period Dummy is intended to capture the increase in Defendants' prices during the Class Period that is unexplained by any of the economic factors that drive prices in a competitive market. (The variable is called a dummy because it takes only two values, a value of 1 for the months during the Class Period and 0 otherwise. See, David Kayne & David Freedman, Reference Guide on Statistics, in Reference Manual on Scientific Evidence, 283, 286, 294-95 (3d ed. 2011).) Assuming that it does so and assuming that the unexplained difference in prices is due to illegal conspiracy, the coefficient yields a supra-competitive overcharge that serves as a measure of Plaintiffs' damages.

         Defendants challenge Dwyer's testimony on reliability grounds. To beat back the challenge, Plaintiffs must show that Dwyer's testimony “is the product of reliable principles and methods, ” which is “based on sufficient facts and data, ” and that Dwyer “has reliably applied the principles and methods to the facts of the case.” Fed.R.Evid. 702(b)-(d). In determining whether Plaintiffs have met this challenge, the Court may consider such factors as: (1) whether the methods that Dwyer employs “can be (and ha[ve] been) tested, ” (2) whether they “ha[ve] been subjected to peer review and publication, ” (3) whether the techniques command widespread acceptance within the relevant scientific community, (4) whether there are “standards controlling the technique's operation, ” and (5) the “known or potential rate of error” of the methods. See, Daubert, 509 U.S. at 593-94; see also, Fed. R. Evid. 702, Advisory Committee's Notes (listing additional factors that courts have found “relevant in determining whether expert testimony is sufficiently reliable to be considered by the trier of fact”). However, as the Supreme Court has explained, “the law grants a district court the same broad latitude when it decides how to determine reliability as it enjoys in respect to its ultimate reliability determination.” Kumho Tire Co. v. Carmichael, 526 U.S. 137, 141-42 (1999) (emphasis in original). As such, the Court need not apply all the factors “to all experts or in every case, ” and it may consider factors other than those listed. Id.

         Defendants point to three features of Dwyer's methodology that they argue show the method to be unreliable. First, they argue that Dwyer's regression produces estimated effects that are “absurd.” In particular, Defendants highlight the fact that Dwyer obtains negative coefficients on variables that reflect the costs of producing containerboard, e.g., variables like hourly wages, pulp prices, and energy costs. See, ECF No. 1093, Ex. 6 (Murphy's April 2016 Report) ¶ 43, App'x S13; ECF No. 1093, Ex. 2 (Dwyer's December 2014 Report) at Ex. 3. This means that the costs of the inputs and the price of the output are negatively related - the more it cost Defendants to make containerboards, the more cheaply they sold those boards. Defendants argue that such results cannot be right.

         Plaintiffs, inter alia, respond that the negative coefficients should not be interpreted as the causal impact of costs on prices. According to Plaintiffs, “Dr. Dwyer has constructed a ‘reduce form' model where individual variables do not express specifically supply-side effects or demand-side effects. Instead, they express complex relationships between supply and demand effects on the structure of a market.” ECF No. 1207 at 9-10. The Court understands that Plaintiffs are saying that even though Dwyer chooses the original, cost variables because “they address costs of Containerboard Product production and delivery, ” because of the way the different variables interact with one other in his regression model - i.e., because of the “complex relationships” - he can no longer say that any of the coefficients on those variables actually capture the causal effect of costs on prices - i.e., the “supply-side effects.” ECF No. 1093, Ex. 1 ¶ 62. The long and short of it is that Plaintiffs have some argument as to why the negative coefficients are not “absurd.”

         The Court agrees that the estimated effects, even in reduced form, are counter-intuitive. However, the Court is mindful that it “usurps the role of the jury, and therefore abuses its discretion, if it unduly scrutinizes the quality of the expert's data and conclusions rather than the reliability of the methodology the expert employed.” Manpower, Inc. v. Ins. Co. of Pa., 732 F.3d 796, 806 (7th Cir. 2013). The conclusion that Dwyer comes to - that costs are negatively related to price - is suspect, but the Court better trains its eyes on the method that produces that conclusion. In this case, Defendants do not challenge the regression itself; instead, they say that the methods by which Dwyer generates his independent variables and selects the specific variables to include in the regression are indefensible. The Court thus directly examines those methods, called principal components and forward selection, rather than strike Dwyer's testimony on the indirect evidence that the “wrong” signs on his estimated coefficients suggest that the model he uses is misspecified.

         Second, Defendants argue that Dwyer's results are not robust, as small changes in how the variables are defined produce large changes to the results. Such criticism goes to the credibility of Dwyer's conclusion, and it calls into question the reliability of the method for arriving at it. See, Joiner, 522 U.S. at 146 (stating that “conclusions and methodology are not entirely distinct from one another”). The call is thus a close one. However, as the Court believes that Defendants can readily explain such weaknesses to a jury (if the case goes to trial), the Court will not strike the testimony on this ground. See, Stollings v. Ryobi Techs., Inc., 725 F.3d 753, 766 (7th Cir. 2013) (“The judge should permit the jury to weigh the strength of the expert's conclusions, provided such shortcomings are within the realm of a lay juror's understanding.”). (Throughout the remainder of this memorandum, the Court will drop the conditional “if the case goes to trial” and speak as if the next stage of the proceeding is a trial by jury. This is done for ease of exposition and to account for the possibility that the experts may testify at trial. The Court expresses no opinion at this time on whether Plaintiffs' case will, in fact, survive summary judgment and proceed to be tried.) Furthermore, as the Court explained previously, it ought to look at the methods that produced this arguably unstable result and not at the instability itself.

         Finally, and perhaps most complicated to explain to a jury, Defendants say that the methods Dwyer employs to generate his independent variables and select which of them to include in the final regression are “unreliable and create omitted variable bias.” This takes some explaining as to what these methods are. Based on his economic judgment, Dwyer has opined that 150 economic variables (not including the Class Period Dummy) drive containerboard prices. However, these economic variables are highly correlated, and regression analysis generally cannot estimate with precision coefficients on such highly correlated (or “collinear”) variables. Dwyer deals with the problem by applying a technique called principal components. The use of this technique allows him to obtain independent variables from the original, correlated variables.

         The technique works as follows. Suppose there are two variables, X and Y, with X measuring the price of crude oil and Y the price of heating oil. Because energy prices tend to move together - for instance, if OPEC decides to cut production, then the price of crude and heating oil would both rise - X and Y are correlated. By applying some linear algebra, the principal component procedure delivers two different variables, call them P1 and P2, that consist of those “components” of X and Y that do not move together. That is, unlike X and Y, P1 and P2 are independent of each other. Roughly speaking, this means that when OPEC cuts production, P1 and P2 move in such a way such that knowing P1 does not help one to predict what P2 will be (and vice versa).

         In addition, while his economic judgment calls for 150 explanatory variables, Dwyer has only 144 observations of containerboard prices to explain. His economic judgment thus results in an over-specified model - something like a whack-a-mole game in which there are more mallets than mole holes. Such a model is impossible to estimate by a regression.

         To solve the problem that he has a theoretical model (in which all 150 variables have some causal relationship to prices) that cannot be empirically estimated, Dwyer neither collects more data nor offers an alternative model that could be estimated given the data limitation. Instead, he turns to a technique by the name of forward selection. This technique (subject to some restrictions imposed by Dwyer that will be discussed below) selects from all the independent variables a subset that will best explain, or “fit, ” the price data. Forward selection enables Dwyer to run a regression with fewer than 144 independent variables while accounting for much of the variation in containerboard prices. However, both experts agree that forward selection selects the variables based on their mechanical fit with the data to be explained (in this case, containerboard prices) rather than any causal relationship between those variables and prices.

         In addition, Murphy points out that Dwyer essentially “stacks the deck” by starting forward selection with a base model in which the Class Period Dummy is already present before allowing the procedure to select other variables. According to Murphy, doing things in this manner favors a regression specification that delivers a statistically significant coefficient on the Class Period Dummy even if prices were not elevated beyond a level justified by economic conditions during the Class Period. This is because with the Class Period Dummy in the base model, forward selection tends to omit variables that, if included, would absorb the effects now captured by the dummy and thus result in a smaller or statistically insignificant coefficient on the dummy being estimated. (In layman's terms, a statistically insignificant coefficient means a coefficient estimate that is indistinguishable from zero, indicating that prices did not rise above competitive levels during the Class Period.) The problem of omitted variables causing the coefficients in the regression to be systematically and wrongly estimated is known as omitted variable bias. Murphy further argues that the use of principal components worsens the omitted variable bias in this case.

         The most persuasive evidence that Murphy offers to illustrate this point is to show how principal components and forward selection spuriously pick up a “conspiracy-consistent effect” even on data containing no such conspiracy. Murphy simulates 100 data sets in which, by construction, there was no elevation in prices during the Class Period. He then applies the principal component and forward selection procedures to the simulated data and shows that, on this “clean” data, the estimation still produced a statistically significant coefficient on the Class Period Dummy 66 to 89% of the time (depending on how exactly the data is simulated). This means that the methods are prone to producing false positives, showing that prices changed in a statistically significant way the majority of the time even when there was no actual change in prices. If one were to look only at statistically significant and positive coefficients (i.e., those interpreted as reflecting damages), then one would still find them in as many as 35 to 47 simulations.

         The Court thinks that a method that produces false results the majority of the time cannot be reliable. An error rate of 60% signifies that, given that prices were the same during the Class Period and outside it, the method would more often than not lead one to conclude wrongly that prices changed in a significant way during the Class Period. However, due to various technical hurdles, Murphy did not quite replicate Dwyer's methodology in his simulated data analysis. The Court thus cannot say that Dwyer's exact methodology is unreliable. As such, although Murphy's criticism exposes weaknesses in Dwyer's analysis, the Court is not prepared to conclude that the analysis is so flawed as to be inadmissible. See, Fed. R. Evid. 702, Advisory Committee's Notes (“A review of the caselaw after Daubert shows that the rejection of expert testimony is the exception rather than the rule.”).

         The Court's conclusion that Dwyer's methods are closer to shaky than unreliable is bolstered by comparing the methods used in this case with those that the Seventh Circuit has found so lacking as to be justifiably excluded. In ATA Airlines, for instance, the court eviscerated an analysis where the expert opined that revenues explained costs on no other basis than that he had data on revenues but no other “more plausible variables”; estimated the relationship between revenues and costs with a “tiny sample” of 10 observations; and “improperly implemented” the flawed model he had. See, ATA Airlines, 665 F.3d at 893-96. Likewise, in Blue Cross & Blue Shield United v. Marshfield Clinic, the court found an expert's testimony to be “worthless” when his regression analysis included only the variable of interest and a single control. See, Blue Cross & Blue Shield United v. Marshfield Clinic, 152 F.3d 588, 593 (7th Cir. 1998). Finally, in Zenith Elec. Corp. v. WH-TV Broad. Corp., the Seventh Circuit affirmed the lower court's decision to exclude the expert's testimony when the expert offered no explanation as to why he did not employ the “extensively used” method of regression analysis but rather resorted to “my expertise” to justify his estimation. See, Zenith Elec. Corp. v. WH-TV Broad. Corp., 395 F.3d 416, 418-20 (7th Cir. 2005).

         In contrast to these experts, Dwyer here performs extensive quantitative analysis. He chooses economic variables that plausibly explain prices, relies on a sample size of 144 observations, and includes a number of controls in his regressions. The Court thus finds that the identified weaknesses in Dwyer's methodologies affect the probativeness of his testimony rather than its admissibility. See, In re Titanium Dioxide Antitrust Litig., No. RDB-10-0318, 2013 U.S. Dist. LEXIS 62394, at *56-58 (D. Md. May 1, 2013) (“[I]nadequacies in a multiple regression analysis normally ‘affect the analysis' probativeness, not its admissibility.'”) (quoting Bazemore v. Friday, 478 U.S. 385, 400 (1986)).

         The Court nonetheless is concerned that a lay jury may not be able to grasp the techniques of principal components and forward selection and so may be easily misled by the expert. See, Daubert, 509 U.S. at 595 (“Expert evidence can be both powerful and quite misleading because of the difficulty in evaluating it.”) (internal quotation marks omitted). The techniques are complicated, as evidenced by the many pages that the experts took to explain them, as well as the pages that the Court just now devoted to lay out its understanding of the methods. To alleviate the problem and avoid exclusion of the testimony under Rule 403, the Court suggests that Plaintiffs build up to their preferred model in incremental steps so as to allow the jury to see what each layer of the methodology is delivering. Plaintiffs should first show graphs of the raw data of containerboard prices before, during, and after the Class Period. This will help the jury to understand whether elevated prices are visible to the “naked eye” or isolated by the econometrics to follow. Plaintiffs should then show an ordinary least squares regression with the original economic variables as independent variables, followed by a regression with the principal components of the economic variables, and then a regression with regressors chosen by the forward selection technique.

         Subject to the above, the Court denies Defendants' Motions to exclude Dwyer's testimony. Defendants remain free to impeach Dwyer with the various arguments they raised in the Motions.

         B. Douglas Zona

         Like Dwyer, Plaintiffs' next expert, Douglas Zona (“Zona”), calculates damages that Plaintiffs allegedly suffered. Unlike Dwyer, however, Zona aims to link Plaintiffs' damages - the increase in prices that Plaintiffs had to pay - to Defendants' reductions in containerboard supply. To do so, Zona performs a two-step analysis. In the first step, he assesses how much Defendants, as a group, reduced their capacity during the Class Period over and above reductions predicted for a comparison group not accused of having engaged in a conspiracy. In the next step, he estimates how containerboard prices changed in response to changes in containerboard supply. By combining the results from these two stages, Zona is able to calculate how much containerboard prices rose because of the supply reduction he identified. Assuming that the additional, more-than-predicted reduction is attributable to Defendants' conspiracy, this increase in prices is a measurement of damages.

         Defendants raise multiple challenges to Zona's analysis. The Court takes the analysis, and its criticisms, piece by piece. In the first step of his analysis, Zona employs a method called multinomial logit to show that Defendants reduced the capacity at their paper mills by more than is predicted for a comparison group. Multinomial logit is a method for modeling behavior that involves discrete, as opposed to continuous, choices, e.g., whether a person played a whack-a-mole game in the last month (the discrete choices are yes/no), or where the person last played a whack-a-mole game (the discrete choices may be at home/at a friend's/at an arcade). One can think of multinomial logit as an alternative estimation method to ordinary least squares regression when one is dealing with discrete data. See, The Tao of Pleading: Do Twombly and Iqbal Matter Empirically? 59 Am. U.L. Rev. 553, 616 n.280 (explaining that “[a] multinomial logistic regression['s] . . . purpose is the same as the more commonly known multiple linear regression, except that in multiple regression, the dependent variable is linear (also called ‘continuous' or ‘quantitative'), while in logistic regression, the dependent variable is categorical”) (citing Damodar Gujarati, Essentials of Econometrics, 451-53 (2d ed. 1999)).

         Of course, capacity does not present such discrete data. A firm's production capacity, for all practical purposes, can take any numerical value from zero to whatever is the upper range imposed by total resources. Nonetheless, Zona uses multinomial logit because he models Defendants' capacity as four discrete choices: maintain capacity, increase capacity, reduce capacity, or close mill. Specifically, Zona looks at capacity data, as published by a third-party industry watch group called RISI, for Defendant and non-Defendant mills during the years 1998-2010. Based on this capacity data, he categorizes a mill as “maintain” if its capacity changed by no more than 20% from one year to the next; “increase” if its capacity jumped by more than 20%; “decrease” if its capacity dropped by more than 20%; and “closed” if the mill shut down production altogether. After categorizing the data in this way, Zona sums up the number of mills falling into each category for all of the Defendants. He then uses multinomial logit as a means to compare how Defendants ran their mills during the Class Period against a benchmark group that consists of Defendants outside the Class Period, non-Defendants outside the Class Period, and non-Defendants during the Class Period.

         Defendants raise a number of issues with this part of the analysis. First, they object that by lumping all Defendants into one group, Zona leaves no room for any particular Defendant to exonerate itself by showing that it did not reduce its containerboard capacity more than the benchmark group. Defendants present data indicating that five of seven Defendants actually had capacity level above that of the benchmark group, and Georgia-Pacific and Temple-Inland in particular emphasize how high their production was relative to their competitors.

         The Court agrees that Zona's analysis cannot establish that any particular Defendant restricted supply in a manner consistent with either tacit collusion or conspiracy. However, this does not mean that the analysis is irrelevant. If believed, the analysis indicates that Defendants, as a group, behaved differently than how they behaved outside of the period of the alleged conspiracy and how non-Defendants behaved. This piece of evidence increases the probability that there was a conspiracy during the Class Period, even if it does not shed light on who among the Defendants actually cut supply in accordance with the alleged conspiracy. Defendants remain free to put on individual defenses at trial that call into doubt the strength of the evidence against any of them, and the jury is entitled to find that, say, two of the Defendants conspired but not the other five. See, Alexander v. Phx. Bond & Indem. Co., 149 F.Supp.2d 989, 1000 (N.D. Ill. 2001) (“[E]ven in a conspiracy case, liability remains individual and is not a matter of mass application.”) (citing Kotteakos v. United States, 328 U.S. 750, 772 (1946)). In short, the evidence is not inadmissible on this basis. See, Smith, 215 F.3d at 720 (“[E]xpert testimony need only be relevant to evaluating a factual matter in the case. That testimony need not relate directly to the ultimate issue that is to be resolved by the trier of fact.”).

         Second, Defendants argue that Zona constructs the wrong benchmark with which to compare their capacity. Recall that Zona is here comparing Defendants' behavior during the Class Period to that both of Defendants outside of the Class Period and non-Defendants during all of the years for which he has data. Defendants point out that, because the benchmark group mixes in non-Defendants, any identified differences between Defendants' capacity choices during the Class Period and the benchmark group could be driven solely by the differences between Defendants and non-Defendants that have nothing to do with the alleged conspiracy. In fact, Defendants demonstrate that this is the case by running what is called a difference-indifference analysis. The analysis controls for any inherent difference between Defendants and non-Defendants, that is, any difference found outside the Class Period and so is unrelated to the alleged conspiracy. With such an adjustment in the benchmark group, Defendants show that any difference in capacity attributable to the alleged collusion dissipates.

         Without quibbling with any part of this difference-indifference analysis, the Court nonetheless concludes that whether Zona “might have done a better job is not the test for the admissibility of his testimony.” Traharne v. Wayne/Scott Fetzer Co., 156 F.Supp.2d 697, 712 (N.D. Ill. 2001). Even if Zona had only compared Defendants to non-Defendants during the Class Period and opined that any difference between them is attributable to Defendants' alleged collusion, still, the Court would not find his method unreliable. A comparison between firms accused of conspiracy and the firms not accused conspiracy appears to the Court to be a valid method to assess the impact of the alleged conspiracy where, as here, the two groups are from the same industry, face the same economic conditions, and produce the same “standardized” containerboard products. See, Kleen Prods., 831 F.3d at 923 (“Containerboard is a commodity, sold in standardized compositions and weights. The final products are also standardized. . . .”).

         Even if it turned out that, despite these similarities, the two groups are different in important ways such that the expert's conclusion is subject to doubt, still such doubts are best left to the jury. See, Stollings, 725 F.3d at 766 (“An expert may provide expert testimony based on a valid and properly applied methodology and still offer a conclusion that is subject to doubt. It is the role of the jury to weigh these sources of doubt.”). In this case, Defendants can easily point out the ways in which Defendants and non-Defendants are different and so argue that the difference between the two groups had nothing to do with the claimed illegal collusion - precisely as they have done in their motions. The jury can then decide whom to believe. The matter is not particularly complicated, and the Court does not believe that Zona's testimony is “too complex for the jury to appreciate important issues of credibility.” Id. The Court thus will let the testimony stand.

         Third, Defendants press that Zona's method is unreliable by pointing to several results produced by the model that are demonstrably false. In particular, Zona's model predicts capacity levels for Defendants that are lower than their actual capacity for the period during the alleged conspiracy. That is, Defendants' real-world capacity levels, allegedly depressed by the conspiracy, were in fact higher than what Zona predicts them to be in the absence of such conspiracy. Zona explains that such discrepancies occurred because his model is intended to estimate capacity growth rates rather than their levels. While the Court is not convinced this explanation puts the matter to rest, it thinks that the push and pull of the adversarial process will best expose the truth as to the correctness of Zona's conclusions. The testimony can be tested and the potential for error exposed, as the Defendants have shown. See, Daubert, 509 U.S. at 594.

         Fourth, Defendants contend that the expert makes a conceptually unjustifiable choice by measuring capacity on a mill-by-mill instead of firm-by-firm basis. Defendants argue that capacity decisions are made at the firm level and should be modeled as such. After all, a firm decides how to run individual mills based on the mills' contribution to the firm's overall profitability and not just the profitability at a particular mill. The Court conceives of this as an attack on the data that Zona uses, essentially an argument that he should have aggregated his mill data to the firm level.

         To put things this way is to give away the answer that the testimony should go to the jury. The Seventh Circuit has made clear that unless there was no rational connection between the data used and the conclusion arrived at, “arguments about how the selection of data inputs affect the merits of the conclusions produced by an accepted methodology should normally be left to the jury.” Manpower, 732 F.3d at 808-09; see also, Hannah's Boutique, Inc. v. Surdej, No. 13-cv-2564, 2015 U.S. Dist. LEXIS 79501, at *19-21, 25-27 (N.D. Ill. June 19, 2015). To illustrate where no such rational connection exists, the court gave an example where a hypothetical expert uses “changes in the size of the white rhino population in Africa” to project earnings for a recruiting firm headquartered in Milwaukee. See, Manpower, 732 F.3d at 808-09. Similarly, it described a situation where an expert relies on “sales to only one customer” to estimate average gross sales. See, id. In contrast to such irrational methodologies, Zona here uses mill-level capacity data to study capacity changes. The Court thus leaves the evaluation of his data-input choice to the jury.

         Likewise, other criticisms that Defendants levy at Zona's testimony are data-related. For example, Defendants say that Zona should have used the Defendant-provided data rather than third-party data from RISI. But even if the RISI data are flawed, “Rule 702 [] does not condition admissibility on . . . a complete and flaw-free set of data.” Lees v. Carthage Coll., 714 F.3d 516, 524-25 (7th Cir. 2013) (internal quotation marks omitted).

         As another example of a data issue, Defendants criticize how Zona chooses to code mill closures in his dataset. The Court does not find his choices to be so outrageous so as to justify exclusion. The decision to code a mill that closed in the middle of the year as a “decrease” for that year and a “closure” for the next year is a choice made to deal with the fact that a mill's capacity is not zero in the very year that it closed. While Zona could have done things differently, the choice that he did make is reasonable (even if not the most conservative). If the choice was driven by the ex post discovery that, without such coding of the data, any looked-for effect would have disappeared, then this is something Defendants are free to point out during cross-examination or presentation of their own expert witnesses. See, Miller UK Ltd. v. Caterpillar, Inc., No. 10-cv-03770, 2015 U.S. Dist. LEXIS 147843, at *11 (N.D. Ill. Nov. 1, 2015) (“[A] party who finds an expert's conclusion disagreeable is entitled to challenge the expert and his or her opinion through cross-examination and, of course, to put on his own expert to offer a counter opinion.”). The attack goes to the robustness and hence probativeness of Zona's testimony, not its reliability.

         Criticisms relating to the second part of Zona's analysis - that devoted to quantifying how decreases in containerboard supply raise the product's price - are simpler. In this part of the analysis, Zona used an “off-the-shelf” model based on a paper published in the Journal of Forest Economics. However, Zona made some modifications to the model, including by dropping some control variables during his estimation. Defendants show that adding the controls back in undermines the magnitude and ...


Buy This Entire Record For $7.95

Download the entire decision to receive the complete text, official citation,
docket number, dissents and concurrences, and footnotes for this case.

Learn more about what you receive with purchase of this case.