Quantitative Techniques for Competition and Antitrust Analysis_1 pdf

1.1 Demand Functions and Demand Elasticities 11.2 Technological Determinants of Market Structure 191.3 Competitive Environments: Perfect Competition, Oligopoly, and 2.1 Multiple Regressi

Quantitative Techniques for Competition and Antitrust Analysis

This page intentionally left blank

Quantitative Techniques for Competition and Antitrust Analysis

Peter Davis and Eliana Garc´es

Princeton University Press Princeton and Oxford

Copyright c
2010 by Peter Davis and Eliana Garc´es

Requests for permission to reproduce material from this work

should be sent to Permissions, Princeton University Press

Published by Princeton University Press,

41 William Street, Princeton, New Jersey 08540

In the United Kingdom: Princeton University Press,

6 Oxford Street, Woodstock, Oxfordshire OX20 1TW

All Rights Reserved

Library of Congress Cataloging-in-Publication Data

Davis, Peter J (Peter John), 1970–

Quantitative techniques for competition and antitrust analysis / Peter Davis, Eliana Garc´es.

p cm.

Includes bibliographical references and index.

ISBN 978-0-691-14257-9 (alk paper)

1 Consolidation and merger of corporations 2 Antitrust law.

3 Econometrics I Garc´es, Eliana, 1968– II Title.

HD2746.5.D385 2010

338.8 0 2015195–dc22 2009005675

British Library Cataloging-in-Publication Data is available

This book has been composed in Times using TEX

Typeset and copyedited by T&T Productions Ltd, London

Printed on acid-free paper. ∞

press.princeton.edu

Printed in the United States of America

10 9 8 7 6 5 4 3 2 1

For Lara, Adrian, and Tristan

For Sara

This page intentionally left blank

1.1 Demand Functions and Demand Elasticities 11.2 Technological Determinants of Market Structure 191.3 Competitive Environments: Perfect Competition, Oligopoly, and

2.1 Multiple Regression 632.2 Identification of Causal Effects 892.3 Best Practice in Econometric Exercises 113

2.5 Annex: Introduction to the Theory of Identification 121

3.1 Accounting and Economic Revenue, Costs, and Profits 1253.2 Estimation of Production and Cost Functions 1313.3 Alternative Approaches 1493.4 Costs and Market Structure 158

4.1 Basic Concepts in Market Definition 1624.2 Price Level Differences and Price Correlations 1694.3 Natural Experiments 1854.4 Directly Estimating the Substitution Effect 1914.5 Using Shipment Data for Geographic Market Definition 1984.6 Measuring Pricing Constraints 201

viii Contents

5 The Relationship between Market Structure and Price 230

5.1 Framework for Analyzing the Effect of Market Structure on Prices 2315.2 Entry, Exit, and Pricing Power 256

9.1 Demand System Estimation: Models of Continuous Choice 4379.2 Demand System Estimation: Discrete Choice Models 4629.3 Demand Estimation in Merger Analysis 491

10 Quantitative Assessment of Vertical Restraints and Integration 502

10.1 Rationales for Vertical Restraints and Integration 50310.2 Measuring the Effect of Vertical Restraints 518

The use of quantitative analysis by competition authorities is increasing around theglobe Whether the quantitative analysis is submitted by external experts, or the com-petition authority itself undertakes the analysis, empirical analysis is now a vitallyimportant component of the competition economist’s toolkit Much of the empiri-cal analysis submitted to, or carried out by, investigators is fairly straightforward.This is partly because simple tools are often very powerful and partly because theneed to communicate with nonexperts sometimes places a natural boundary on thedegree of sophistication which can comfortably be used Of course, one person’s

“cutting-edge” method is another’s basic tool and this difference drives the normalprocess of diffusion of new methods from basic research to applied work The tools

we discuss in this book are broadly the result of the ideas and methods which havedeveloped over the past twenty years in the empirical industrial organization liter-ature and which are either gradually diffusing into practice or, no doubt in a smallnumber of cases, gradually diffusing into obscurity

While the aim of this book is to examine empirical techniques, we cannot stressenough that any empirical analysis in a competition investigation needs to be eval-uated together with the factual, documentary, and qualitative evidence collectedduring the case An empirical analysis will usually be one albeit important element

in a broader evidence base Only in a small minority of cases will quantitative sis alone be sufficiently clear-cut, precise, and robust enough to support a finding,though it will provide one important plank of evidence in a wider range of cases.Even in cases where quantitative analysis is important, a solid qualitative analysisand a good factual knowledge of the industry will provide both a necessary basisfor quantitative work and a source for vital reality checks regarding the conclusionsemerging from empirical work

analy-With those caveats firmly in mind, in this book we discuss the most useful and mostpromising empirical strategies available to antitrust and merger investigators Some

of these techniques are tried and tested, others are more sophisticated and not yetwidely embraced by practitioners Throughout we try to take a careful practitioner’seye to tools that have often been proposed by the academic community The fact

is that practitioners need to understand both the potential uses and the importantlimitations of the available methods before they will, indeed before they should,choose to apply them We do that by closely tying the empirical models and empiricalstrategy used to answer our competition policy questions to the underlying economictheory Specifically, economic theory allows us to define the assumptions required for

a given piece of empirical work to be meaningful Indeed, no solid empirical analysis

x Preface

is entirely disconnected from economic theory and thus theory usually has a veryimportant role in providing guidance and discipline in the design of empirical work.The purpose of this book is not theory for itself but rather the aim is to help compe-tition economists answer very practical questions For this reason the structure of thebook is broadly based around potential competition issues that need to be addressed.The first two chapters provide a review of basic theory and econometrics Specifi-cally, the first chapter reviews the determinants of market outcomes, i.e., demand,costs, and the competitive environment, since those are the fundamental elementsthat need to be very well-understood before any competition policy analysis is pos-sible The second chapter reviews the basic econometrics of multiple regressionwith a particular emphasis on the crucially important problem of “identification.”Identification—the data variation required to enable us to tell one model apart fromanother—is a theme which emerges throughout the book The subsequent chaptersguide the reader through issues such as the estimation of cost and demand func-tions, market definition, the link between market structure and price, the scope foridentifying firms’ competitive conduct, damage estimation, merger simulation, and

we end with the developing approaches to the quantitative assessment of the effects

of vertical restraints Each chapter critically discusses the empirical techniques thathave been used to address that competition policy issue The book does not aim to

be comprehensive, but we do aim to provide practical guidance to investigators.Naturally, sometimes tools which are too simple for the job at hand can result inthe investigator getting a radically wrong answer On the other hand, sophisticatedtools poorly understood will be poorly applied and are more likely to act as a blackbox from which a decision emerges instead of providing a great deal of insight Such

is the challenge faced by antitrust agencies in choosing an appropriate economicmethodology In some instances, we will discuss empirical techniques that an indi-vidual agency may well currently judge to be too complicated, too theoretical, or tootime-consuming to be of immediate practical use for time-constrained investigators.The approach of this book is that these techniques can still be useful in that theywill at least signal the difficulty or complexity of a particular question and even anabstract discussion still provides guidance on the relevant empirical questions thatneed to be investigated if we want to have conclusions on a particular topic In addi-tion, the requisite expertise may be built gradually within an institution rather thanwithin the remit of, say, a particular merger inquiry with a statutory deadline Theultimate objective of this book is not to have economists in competition authoritiesreplicate the examples discussed in these chapters but to help them develop a way

of thinking about empirical analysis which will help them design their own originalanswers to the specific problems they will face given the data that they have Wealso hope that the book will help reduce the amount of concurrent rediscovery ofstrengths and weaknesses of particular approaches currently undertaken in agenciesacross the world

Preface xiFinally, it is important to note that while this book explores the variety of meth-ods available to analysts, the right tool for any particular inquiry will depend on thecontext of that inquiry This book does not aim to explicitly or implicitly set anyrequirements as to how competition questions should be addressed empirically inany particular jurisdiction We do, however, aim to raise awareness among empiri-cal economists of the underlying econometric and economic theory that inevitablyunderpins all empirical techniques Our hope is that increased awareness will bothpromote high-quality work in the relatively simple empirical exercises and alsoreduce the entry barriers hindering the use of more sophisticated approaches wheresuch methods are appropriate.

Dr Peter Davis currently serves as Deputy Chairman of the U.K Competition mission While he is a principal author of the text, he writes as an individual and theviews expressed in this text are solely those of the author and should not be taken toreflect the views of the U.K Competition Commission Indeed, this text has evolvedfrom a project undertaken, before his current appointment, by Applied EconomicsLtd (www.appliedeconomics.com) for the European Commission

Dr Eliana Garc´es previously worked for the Chief Economist team at DG petition before taking on her current role as a Member in the Cabinet of EuropeanCommissioner for Consumer Protection Meglena Kuneva The contribution to thiswork is her own and does not represent the opinions of the European Commission.This book began life as a project in the European Commission to disseminatepractical knowledge and good practices in empirical analysis We would like tothank the European Commission and, in particular, the EC’s Competition ChiefEconomists who served during the making of this book, Lars Hendrik R¨oller andDamien Neven, for their continued support of the project

Com-The book has benefited in numerous ways from contributors Com-The authors wouldlike to thank Richard Baggaley from Princeton University Press for his support,encouragement, and patience and Jon Wainwright from T&T Productions Ltd forhis tireless efforts to typeset the book in the face of numerous corrections andamendments Enrico Pesaresi provided valuable support throughout the process

We also thank Frank Verboven and Christian Huveneers for their detailed comments

on the draft version And, of course, the anonymous reviewers for their importantcontributions This work builds heavily on the work of many authors who have eachcontributed to the literature The authors would, however, like to thank, in particular,Douglas Bernheim and John Connor for allowing them to draw extensively from theirwork on cartel damage estimation Last but by no means least, the book incorporates

in substantial part updated and expanded content from classes and lectures Peter hastaught at MIT and LSE over the best part of a decade and a substantial debt ofgratitude is due to former students and colleagues at those institutions as well as tohis former classmates and teachers at Yale and Oxford In particular, thanks are due

to Ariel Pakes, Steve Berry, Lanier Benkard, Ernie Berndt, Sofronis Clerides, PhilipLeslie, Mark Schankerman, Nadia Soboleva, Tom Stoker, and John Sutton

The Determinants of Market Outcomes

A solid knowledge of both econometric and economic theory is crucial when ing and implementing empirical work in economics Econometric theory provides

design-a frdesign-amework for evdesign-aludesign-ating whether ddesign-atdesign-a cdesign-an distinguish between hypotheses ofinterest Economic theory provides guidance and discipline in empirical investiga-tions In this chapter, we first review the basic principles underlying the analysis

of demand, supply, and pricing functions, as well as the concept and application

of Nash equilibrium We then review elementary oligopoly theory, which is thefoundation of many of the empirical strategies discussed in this book Continuing

to develop the foundations for high-quality empirical work, in chapter 2 we reviewthe important elements of econometrics for investigations Following these first tworeview chapters, chapters 3–10 develop the core of the material in the book Theconcepts reviewed in these first two introductory chapters will be familiar to all com-petition economists, but it is worthwhile reviewing them since understanding thesekey elements of economic analysis is crucial for an appropriate use of quantitativetechniques

The analysis of demand is probably the single most important component of mostempirical exercises in antitrust investigations It is impossible to quantify the likeli-hood or the effect of a change in firm behavior if we do not have information aboutthe potential response of its customers Although every economist is familiar withthe shape and meaning of the demand function, we will take the time to brieflyreview the derivation of the demand and its main properties since basic conceptualerrors in its handling are not uncommon in practice In subsequent chapters we willsee that demand functions are critical for many results in empirical work undertaken

in the competition arena

1.1.1 Demand Functions

We begin this chapter by reviewing the basic characteristics of individual demandand the derivation of aggregate demand functions

2 1 The Determinants of Market Outcomes

Figure 1.1. (Inverse) demand function

1.1.1.1 The Anatomy of a Demand Function

An individual’s demand function describes the amount of a good that a consumerwould buy as a function of variables that are thought to affect this decision such asprice Piand often income y Figure 1.1 presents an example of an individual lineardemand function for a homogeneous product: Qi D 50
0:5Pi or rather for theinverse demand function, Pi D 100
2Qi More generally, we may write Qi DD.Pi; y/.1 Inverting the demand curve to express price as a function of quantitydemanded and other variables yields the “inverse demand curve” Pi D P Qi; y/.Standard graphs of an individual’s demand curve plot the quantity demanded of thegood at each level of its own price and take as a given the level of income and thelevel of the prices of products that could be substitutes or complements This meansthat along a given plotted demand curve, those variables are fixed The slope of thedemand curve therefore indicates at any particular point by how much a consumerwould reduce (increase) the quantity purchased if the price increased (decreased)while income and any other demand drivers stayed fixed

In the example in figure 1.1, an increase in price,
P , of€10 will decrease thedemand for the product by 5 units shown as
Q The consumer will not purchaseany units if the price is above 100 because at that point the price is higher than thevalue that the customer assigns to the first unit of the good

One interpretation of the inverse demand curve is that it shows the maximum pricethat a consumer is willing to pay if she wants to buy Qi units of the good While a

1 This will be familiar from introductory microeconomics texts as the “Marshallian” demand curve (Marshall 1890).

1.1 Demand Functions and Demand Elasticities 3consumer may value the first unit of the good highly, her valuation of, say, the onehundredth unit will typically be lower and it is this diminishing marginal valuationwhich ensures that demand curves typically slope downward If our consumer buys

a unit only if her marginal valuation is greater than the price she must pay, then theinverse demand curve describes our consumer’s marginal valuation curve

Given this interpretation, the inverse demand curve describes the differencebetween the customer’s valuation of each unit and the actual price paid for eachunit We call the difference between what the consumer is willing to pay for eachunit and what he or she actually pays the consumer’s surplus available from thatunit For concreteness, I might be willing to pay a maximum of€10 for an umbrella

if it’s raining, but may nonetheless only have to pay€5 for it, leaving me with ameasure of my benefit from buying the umbrella and avoiding getting wet, a surplus

of€5 At any price Pi, we can add up the consumer surplus available on all of theunits consumed (those with marginal valuations above Pi) and doing so provides

an estimate of the total consumer surplus if the price is Pi

In a market with homogeneous products, all products are identical and perfectlysubstitutable In theory this results in all products having the same price, which isthe only price that determines the demand In a market with differentiated products,products are not perfectly substitutable and prices will vary across products sold

in the market In those markets, the demand for any given product is determined

by its price and the prices of potential substitutes In practice, markets which lookhomogeneous from a distance will in fact be differentiated to at least some degreewhen examined closely Homogeneity may nonetheless be a reasonable modelingapproximation in many such situations

1.1.1.2 The Contribution of Consumer Theory: Deriving Demand

Demand functions are classically derived by using the behavioral assumption thatconsumers make choices in a way that can be modeled as though they have anobjective, to maximize their utility, which they do subject to the constraint that theycannot spend more than they earn As is well-known to all students of microeconomictheory, the existence of such a utility function describing underlying preferencesmay in turn be established under some nontrivial conditions (see, for example, Mas-Colell et al 1995, chapter 1) Maximizing utility is equivalent to choosing the mostpreferred bundle of goods that a consumer can buy given her wealth

More specifically, economists have modeled a customer of type yi; i/ as ing to maximize her utility subject to the budget constraint that her total expenditurecannot be higher than her income:

choos-Vi.p1; p2; : : : ; pJ; yiI i/ D max

ui.q1; q2; : : : ; qJI i/subject to p q C p q C    C p q 6y ;

4 1 The Determinants of Market Outcomes

where pj and qj are prices and quantities of good j , ui.q1; q2; : : : ; qJI i/ is theutility of individual i associated with consuming this vector of quantities, yi is thedisposable income of individual i , and idescribes the individual’s preference type

In many empirical models using this framework, the “i ” subscripts on the V and ufunctions will be dropped so that all differences between consumers are captured

J C 1 equations in J C 1 unknowns: the J quantities and the value of the Lagrangemultiplier, 

At the optimum, the first-order conditions describe that the Lagrange multiplier

is equal to the marginal utility of income In some cases it will be appropriate toassume a constant marginal utility of income If so, we assume behavior is described

by a utility function with an additively separable good q1, the price of which isnormalized to 1, so that ui.q1; q2; : : : ; qJI i/ D Qui.q2; : : : ; qJI i/ C q1and p1D

1 This numeraire good q1is normally termed “money” and its inclusion provides anintuitive interpretation of the first-order conditions In such circumstances a utility-maximizing consumer will choose a basket of products so that the marginal utilityprovided by the last euro spent on each product is the same and equal to the marginalutility of money, i.e., 1.2

More generally, the solution to the maximization problem describes the ual’s demand for each good as a function of the prices of all the goods being soldand also the consumers’ income Indexing goods by j , we can write the individual’sdemands as

individ-qij D dij.p1; p2; : : : ; pJI yiI i/; j D 1; 2; : : : ; J:

A demand function for product j incorporates not only the effect of the own price

of j on the quantity demanded but also the effect of disposable income and theprice of other products whose supply can affect the quantity of good j purchased

In figure 1.1, a change in the price of j represents a movement along the curve while

a change in income or in the price of other related goods will result in a shift orrotation of the demand curve

2 This is called a quasi-linear demand function and gives the result because the first-order condition for good 1 collapses to