Seventh Circuit Provides Primer On Regression Analysis And What Can Go Wrong

||Seventh Circuit Provides Primer On Regression Analysis And What Can Go Wrong

Seventh Circuit Provides Primer On Regression Analysis And What Can Go Wrong

March 2012

After a jury trial, ATA Airlines (ATA) won a nearly $66 million verdict against Federal Express Corporation (FedEx). FedEx appealed that ruling.  In ATA Airlines, Inc. v. Federal Express Corp., No. 11-1382 (7th Cir. Dec. 27, 2011), a distinguished panel consisting of Justices Easterbrook, Posner, and Wood criticized the complex damages analysis accepted by the trial judge and jury.  The case provides some interesting illustrations and lessons in presenting complex damages calculations.

At the end of its ruling, the Appellate Court identified perhaps the most important mistake made by defendant Federal Express – FedEx apparently presented no damages estimate of its own.  Instead, the defendant relied on its searing cross-examination to so cripple the plaintiff’s expert that the jury would award nothing at all, regardless of a liability finding.  Cross-examination alone was not enough, and the jury took the plaintiffs number without any adjustment.  The Appellate Court summarized the situation as follows:

“The “principles and methods” used by expert witnesses will often be difficult for a judge to understand. But difficult is not impossible. The judge can require the lawyer who wants to offer the expert’s testimony to explain to the judge in plain English what the basis and logic of the proposed testimony are, and the judge can likewise require the opposing counsel to explain his objections in plain English.

This might not have worked in the present case; neither party’s lawyers, judging from the trial transcript and the transcript of the Rule 702 hearing and the briefs and oral argument in this court, understand regression analysis; or if they do understand it they are unable to communicate their understanding in plain English. But a judge can always appoint his own expert to assist him in understanding and evaluating the proposed testimony of a party’s expert. Fed. R. Evid. 706;” …

“Both because FedEx tendered no estimate of damages and because neither Morriss  [the plaintiff’s expert] nor the lawyers nor the judge presented an intelligible damages analysis to the jury, it is no surprise that, having decided that ATA should win, the jury simply awarded the exact figure that ATA had asked for in damages. …Evidence unintelligible to the trier or triers of fact has no place in a trial.”

As noted in the above quote, the plaintiff’s damage witness calculated damages with the assistance of a regression analysis.  This is a commonly-used and accepted tool which itself should not have caused problems.  The Appellate Court started with the following summary of what a regression analysis is:

“A simple linear regression (that is, one involving only two variables—one the dependent variable, the variable to be explained, and the other the independent variable, the variable believed to explain the dependent variable) is easily visualized by plotting the data points on a graph. The regression line is a straight line that minimizes the aggregate of the squared vertical distances from the points to the line. The equation that generates that line can be written as Y = a + bX + u, where Y is the dependent variable, a the intercept (explained below), X the independent variable, b the coefficient of the independent variable (that is, the number that indicates how changes in the independent variable produce changes in the dependent variable), and u the regression residual—the part of the dependent variable that is not explained or predicted by the independent variable and the intercept, or in other words is “left over,” like the change you receive after paying for a 99-cent item with a $1 bill.”

The Seventh Circuit then provided examples and illustrations of regression analysis, including charts showing illustrative data, and the data in this case.  Interestingly, the Appellate Court found it necessary to perform its own charting of the data, meaning that apparently the plaintiff had not already done so on its own at trial.  This is a problem.  A proper statistics analysis should make its methodology and application of its methodology clear to the trier of fact.  Doing so would make the correctness of the position clear, as well as identify weaknesses or alternatives that should be part of the presentation.   Instead, the Appellate Court was left to make its critique without the benefit of a more detailed explanation from the Plaintiff of why its presentation was correct.

The plaintiff’s calculation was painfully mechanical, and failed to address the reasonableness of the mathematical conclusion.  In this case, the plaintiff considered ten years of data, but did not address the trend of this data over time.  Stated otherwise, the plaintiff’s regression calculations ignored time, causing any such trend to be similarly ignored.  Consequently,  the projected costs and profits in the last/damage year bore no relationship whatsoever to what had occurred in recent years.  The result was a silly conclusion that lost revenues would have caused the plaintiff’s total revenues to increase, yet total costs in that year would decrease at the same time.

Putting aside the trends in the data, the data used in the analysis showed variations that caused the cost predictions to be quite unreliable.  The Seventh Circuit described the issue and result as follows:

“Confidence intervals (familiar as the “margins of error” reported in predictions of election outcomes) are statistical estimates of the range within which there can be reasonable confidence that a correlation or prediction is not the result of chance variability in the sample on which the correlation or prediction was based; 95 percent confidence is the standard criterion of reasonable confidence used by statisticians. Consider our hypothetical regression of wages on experience. A regression based on a sample of 10 workers would yield a less precise prediction of what the average relation of wages to experience was for the workers in a plant that had 1000 workers than a regression based on a sample of 50 or 100 of the workers.

The 95 percent confidence interval for Morriss’ prediction of ATA’s 2008 costs was correctly calculated in the report of FedEx’s expert to be $90 million. This means that Morriss’s estimate that ATA would have costs of $254 million was the midpoint of a range from $299 million at the top ($254 million + $90/2 million) to $209 million at the bottom ($254 million – $90/2 million)—and if its costs were at the top of the range the result would have been a $12.5 million annual net loss for ATA rather than Morriss’s predicted $32.7 million profit (before the adjustment for fixed costs). All else aside, the confidence interval is so wide that there can be no reasonable confidence in the jury’s damages award.”

Courts do not often do as good a job as this case does in explaining statistical concepts and damages calculations.  The case is worthwhile for its willingness to address these topics in an understandable manner.

Fulcrum Inquiry performs damages analyses in complex commercial litigation.

2018-12-20T11:39:00+00:00Commercial Damages|

Monthy Archives