Econometrics and Regression Analysis

• Pages: 123-208

CHAPTER 6

ECONOMETRICS AND REGRESSION ANALYSIS

Econometrics refers to a set of statistical methods or tools that

economists use to assess antitrust damages. Regression analysis is the

most common tool of econometrics. It enables one to uncover the

relationship between a dependent variable (the outcome being modeled,

such as prices) and one or more explanatory variables (the potential

influences on the dependent variable, such as anticompetitive conduct,

supply factors including costs or capacity, and demand factors).1

Regressions can prove useful in answering the basic “counterfactual”

question: What would market outcomes (e.g., prices) have been in the

absence of the anticompetitive act? By their very nature, these but-for

market outcomes are unobserved and, as such, must be estimated by the

damages expert.

Market outcomes are often the result of a complex interaction among

a large number of factors. For example, market prices were likely affected

by a wide variety of demand and supply factors unrelated to the alleged

anticompetitive act. Isolating and measuring the effect of an alleged

anticompetitive act on price (or on other market outcomes) requires

properly accounting for these other factors. Otherwise, one might attribute

to the alleged anticompetitive act the effect of one or more of these other

factors or miss the effects of the act in question by failing to control for a

relevant factor.

Econometric models—particularly those involving regression

analysis—are uniquely suited to isolating the effect of a single factor on

the market factor of interest, while properly accounting for other relevant

factors. The measured effect of an explanatory variable, controlling for

other factors, is referred to as an estimate of the “partial effect” of the

explanatory variables on the dependent variable. In other words, an

explanatory variable’s partial effect is the change in the dependent variable

that would result from a change in that explanatory variable, holding all of

the other explanatory variables constant.2 Thus, when correctly

implemented, econometric techniques can isolate and measure the effect

of a single explanatory factor—such as the impact of the alleged

1. See Jeffrey M. Wooldridge, Econometric Analysis of Cross Section and

Panel Data 3-4 (2d ed. 2010).

2. See William Greene, Econometric Analysis 36 (7th ed. 2010).

123

124 Proving Antitrust Damages

conduct—on the economic outcomes that are relevant when estimating

damages.

The key role of econometrics in legal proceedings is to use the

available data to provide accurate and reliable measures of the economic

impact of the alleged conduct for the finders of fact (judges or juries). This

chapter describes the legal requirements, the methods, and a wide range of

issues in the use of econometrics and statistics in expert testimony. To

begin, the damages methodology must be designed to causally determine

the impact of the alleged conduct. Implementation must address issues of:

data, model specification, estimation, interpretation of estimation results,

and hypothesis testing. The following discussion explains the use of

econometrics to implement before-during-after and benchmark analyses

in general and in the specific context of a case study.3 Not all of the tests

or analyses discussed here need to be performed in every damages

analysis; the ones that are important for a particular case will depend on

the specific facts of the case.

A. Legal Requirements

The legal requirements for methods used to estimate damages,

including all econometric methods, fall under the rules for testimony by

experts. Under Federal Rule of Evidence 702:

If scientific, technical, or other specialized knowled ge will assist the trier

of fact to understand the evidence or to determine a fact in issue, a

witness qualified as an expert by knowledge, skill, experience, training,

or education, may testify thereto in the form of an opinion or otherwise,

if (1) the testimony is based upon sufficient facts or data, (2) the

testimony is the product of reliable principles and methods, and (3) the

witness has applied the principles and methods reliably to the facts of the

case.4

Regression analyses have met these requirements many times in

litigation for a wide range of issues, including the estimation of antitrust

damages,5 as well as showing at the class certification stage that such

3. See Chapter 4, Section C for an overview of the different approaches used

to quantify damages, including before-during -after and benchmark (or

yardstick) analyses.

4. Fed. R. Evid. 702.

5. See, e.g., Conwood Co. v. U.S. Tobacco Co., 290 F.3d 768 (6th Cir. 2002);

In re Urethane Antitrust Litigation, 768 F.3d 1245 (10th Cir. 2014).

Econometrics and Regression Analysis 125

damages are consistent with the plaintiff’s theory of liability.6 A

competent expert should be able to tell the litigant whether sufficient facts

and data are available for econometric analysis, explain the types of

econometric analyses that are applicable, and reliably apply these

econometric techniques.

Using regression analysis does not, by itself, guarantee that an analysis

will be viewed as reliable.7 In general, econometric results will be more

reliable as the amount and quality of data increase.8 If sufficient data are

available, econometric analysis has been found necessary to achieve the

minimum scientific standard for establishing lost sales and price changes.

For example, in Zenith Electronics Corp. v. WH-TV Broadcasting Corp.,9

expert opinion and internal forecasts for sales growth were excluded

because data to estimate sales growth via regression analysis were

available and not used. The regression must be in a form that assists in

determining a material fact, such as the amount of lost sales or the size of

price changes.10 The analysis also must be based on data “reasonably relied

upon by experts in the field.”11 Because experts typically verify data for

accuracy in consulting work or academic research, experts presenting

6. Comcast Corp. v. Behrend, 133 S. Ct. 1426 (2013); See also In re Rail

Freight Surcharge Antitrust Litigation, 725 F.3d244 (D.C. Cir. 2013)

(remanding in light of Behrend so that the district court could analyze

whether the regression model proffered by plaintiffs’ expert passed

sufficient muster at the class certification stage).

7. In re Hydrogen Peroxide Antitrust Litigation, 552 F.3d 305 (3d Cir. 2008);

Kottaras v. Whole Food Market, Inc., 281 F.R.D. 16, 25-26 (D.D.C. 2012);

In re Live Action Antitrust Litigation, 863 F. Supp. 2d 966 (C.D. Cal.

2012).

8. Roy J. Epstein, An Econometrics P rimer for Lawyers, ANTITRUST,

Summer 2011, at 29.

9. 395 F.3d 416 (7th Cir. 2005); see also In re High-Tech Employee Antitrust

Litigation, 289 F.R.D. 555, 582 (N.D. Cal. 2013) (rejecting as evidence

proposed factors analysis and compensation charts, but accepting that a

regression analysis provided plausible support for the class-wide theory

espoused by plaintiffs).

10. Daniel Rubinfeld, Reference Guide on Multiple Regression, in FEDERAL

JUDICIAL CENTER, REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 303 (3d

ed. 2011), available at http://www.fjc.gov/public/

pdf.nsf/lookup/sciman03.pdf/$file/sciman03.pdf.

11. Fed. R. Evid. 703.