Appendix IV. Issues In Scanner Data

• Pages: 431-462

431

APPENDIX IV

ISSUES IN SCANNER DATA

A. Introduction

The past decade has witnessed substantial developments in the

quantitative analysis of horizontal mergers. Several factors account for

these changes. First, the quantity and quality of data available to

estimate the degree of substitutability among competing products has

increased dramatically. Second, this increase has been accompanied by a

substantial decrease in the price of computing power required to analyze

the data. Third, there has been an increased focus on the possibility of

competitive harm from unilateral market conduct, particularly in mergers

involving differentiated products. However, understanding how

consumers substitute among competing products also contributes to the

analysis of possible competitive harm from coordinated market behavior.

How consumers substitute across products as relative prices change

is relevant to understanding the potential price effects of mergers.1 This

information, contained in the own- and cross-price elasticities of demand

at retail,2 is most frequently used as a means for predicting the unilateral

incentives to increase prices postmerger. In the context of differentiated

consumer goods, the term “unilateral effects” refers to the fact that a

1. Of course, the own- and cross-price elasticities of the demand faced by

manufacturers (wholesale demand) are more directly relevant to the

market definition and competitive effects analyses in the U.S. Department

of Justice and the Federal Trade Commission’s Horizontal Merger

Guidelines. The properties of demand of consumers (the demand at

retail) are relevant because they have implications for wholesale demand.

2. The elasticity of some variable x with respect to another variable y is the

percentage change in x that arises from a 1% increase in y. For example,

the own-price elasticity of the demand for widgets is the percentage

change in the quantity of widgets demanded given a 1% increase in their

price.

Econometrics in Antitrust

432

merger of competitors creates an incentive to increase price (all else

being equal) to the extent that there are significant substitution

possibilities between the (now) jointly-owned products. In this setting,

the existence of credible information about demand elasticities is

important. Having a good estimate of how substitutable these products

are helps the analyst predict more accurately whether these unilateral

incentives to raise price are substantial or minuscule.3

Until recently, economists had to rely on relatively simple

quantitative and qualitative information when performing merger

analyses. Resources such as surveys, analyses of shift in share, or

internal company documents did not permit quantification of the degree

of substitutability among the merging firms’ products. While these

3. It has become conventional to analyze these unilateral pricing incentives

using static oligopoly models in which firms compete on the basis of

price. The static oligopoly game that models firms’ pricing decisions is

sometimes referred to as the “Bertrand” pricing game, named after Joseph

Bertrand, Review of “Theorie Mathematique de la Richess Sociale” and

Recherche sur les Principes Mathematiques de las Theorie des Richesses,

499-508 JOURNAL DE SAVANTS (1883), who was the first economist to

study it. This is the framework used most often to examine unilateral

effects in merger analysis. Whether these static models provide an

appropriate benchmark for predicting the consequences of a horizontal

merger is somewhat controversial. See, e.g., Franklin M. Fisher, It Takes

t* to Tango: Trading Coalitions with Fixed Prices, 56 REV. ECON. STUD.

391 (1989); Carl Shapiro, Market Power and Mergers in Durable-Good

Industries: Comments, 32 J.L. & ECON. S227 (1989). Without attempting

to resolve that controversy here, note that several recent studies have

attempted to test the validity of static oligopoly models. See, e.g., Aviv

Nevo, Measuring Market Power in the Ready-to-Eat Cereal Industry, 69

ECONOMETRICA 307 (2001); Joris Pinske & Margaret Slade, Mergers,

Brand Competition, and the Price of a Pint (2001) (mimograph, on file at

the Department of Economics, University of British Columbia); Jerry

Hausman & Gregory Leonard, The Competitive Effects of a New Product

Introduction: A Case Study (2000) (mimeograph, on file at the

Massachusetts Institute of Technology); David Genesove & Wallace

Mullin, Testing Static Oligopoly Models: Conduct and Cost in the Sugar

Industry, 1890-1914, 29 RAND J. ECON. 355 (1998); Catherine Wolfram,

Measuring Duopoly Power in the British Electricity Market, 89 AM.

ECON. REV. 805 (1999). Most of these studies suggest that the static

oligopoly models yield reasonably accurate predictions of pricing

behavior.

Issues in Scanner Data

433

sources of information are useful and continue to play an important role

in merger analysis, a well-executed econometric analysis of demand may

enable an economist to infer not only that a set of goods are substitutes,

but also what volume of sales will switch from product X to product Y

given a specific price increase for product X.

The value of information on demand elasticities is not limited to

studies of unilateral pricing incentives. Information about the degree of

substitutability among potentially competing products also is important

in determining the incentive and ability to engage in coordinated post-

merger pricing.

It is interesting to note that marketing professionals also use

elasticity information taken from econometric analyses of scanner data.

Manufacturers of consumer products estimate systems of demand

equations to help determine optimal prices for their products. Scanner

data drawn from consumers’ actual purchases provide a wealth of

information that can be used to describe and analyze consumer demand.

While the quantitative estimation of demand relationships can make

substantial contributions to merger analysis, practitioners nevertheless

must confront and resolve difficult econometric and conceptual issues.

This appendix attempts to identify a number of those issues that

researchers and practitioners may consider to make the quantitative

estimation of demand relationships using scanner data more applicable to

merger review. Specifically:

1. What is the best way to aggregate data across observational units

and across time?

2. Is it necessary to address the possible endogeneity of explanatory

variables?

3. Is it possible to construct meaningful measures of the accuracy

of predicted price changes?

4. Can one easily translate elasticities estimated with retail-level

data into wholesale-level elasticities?

These are difficult questions, and this appendix does not attempt to

provide definitive answers to them. Its purpose is to provoke further

discussion and research into these issues, with the ultimate goal of

improving the quality of antitrust analysis.