Lecture Notes on Industrial Organization

Xavier Martinez-Giralt CODE and Department of Economics Universitat Aut` onoma de Barcelona

Lecture Notes on Industrial Organization

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Contents
Preface 1 Introduction 1.1 A historical appraisal. . . . . . . . . . . . . . . . . . . 1.2 Oligopoly theory vs. the SCP paradigm. . . . . . . . . 1.3 Variations of prices and welfare. . . . . . . . . . . . . 1.3.1 Price indices. . . . . . . . . . . . . . . . . . . 1.3.2 Welfare variations . . . . . . . . . . . . . . . 1.3.3 Consumer surplus . . . . . . . . . . . . . . . 1.4 Producer surplus and deadweight welfare loss. . . . . . 1.5 Market and market power. . . . . . . . . . . . . . . . 1.5.1 Denition of the market. . . . . . . . . . . . . 1.5.2 Concentration measures. . . . . . . . . . . . . 1.5.3 Degree of Monopoly. . . . . . . . . . . . . . . 1.5.4 Concentration indices and degree of monopoly. 1.5.5 Volatility measures . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Monopoly Pricing 3 Homogeneous Product Oligopoly Models 3.1 Cournot oligopoly. Quantity competition . . . . . . . . . . . . . . 3.1.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2 Equilibrium. . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3 Cournot equilibrium and Pareto optimality. . . . . . . . . 3.1.4 Existence of Cournot equilibrium. . . . . . . . . . . . . . 3.1.5 Uniqueness of the Cournot equilibrium. . . . . . . . . . . 3.1.6 An alternative approach to the existence and uniqueness of Cournot equilibrium. . . . . . . . . . . . . . . . . . . 3.1.7 Strategic complements and substitutes. . . . . . . . . . . 3.1.8 Cournot and conjectural variations. . . . . . . . . . . . . vii xiii 1 1 3 6 7 8 10 14 16 16 18 20 22 24 25 27 29 29 29 32 34 36 40 43 45 46

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

viii

CONTENTS 3.1.9 The geometry of the Cournot model. . . . . . . . . . . . . 3.1.10 Cournot and the competitive equilibrium. . . . . . . . . . 3.1.11 Stability of the Cournot equilibrium. . . . . . . . . . . . . 3.2 Price competition. . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Bertrands model. . . . . . . . . . . . . . . . . . . . . . . 3.3 Cournot vs. Bertrand. . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Variations 1. Capacity constraints. . . . . . . . . . . . . . . . . . 3.4.1 Rationing rules. . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Exogenous capacity constraints: Edgeworths cycle. . . . 3.4.3 Endogenous capacity constraints. Kreps and Scheinkman model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Variations 2. Contestable markets. . . . . . . . . . . . . . . . . . 3.6 Variacions 3. Sticky prices. Sweezys model. . . . . . . . . . . . 3.6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . 3.6.2 Sweezys model. . . . . . . . . . . . . . . . . . . . . . . 3.7 Variations 4. Commitment. . . . . . . . . . . . . . . . . . . . . . 3.7.1 The Stackelberg model. . . . . . . . . . . . . . . . . . . . 3.8 Too many models? How to select the good one? . . . . . . . . . 3.9 Variacions 5. Conjectural variations. . . . . . . . . . . . . . . . . 3.9.1 Bowleys model. . . . . . . . . . . . . . . . . . . . . . . 3.9.2 Consistent conjectural variations. . . . . . . . . . . . . . 3.9.3 Statics vs dynamics. Marschack and Selten models. . . . . 3.10 Variations 6. Dynamic models. . . . . . . . . . . . . . . . . . . . 3.10.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . 3.10.2 Supergames. . . . . . . . . . . . . . . . . . . . . . . . . 3.10.3 Multiperiod games. . . . . . . . . . . . . . . . . . . . . . 3.11 Variations 7. Supermodular games. . . . . . . . . . . . . . . . . . 3.12 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 52 63 67 67 68 71 74 74 76 81 86 89 89 90 93 93 95 97 98 101 102 104 104 105 107 109 109 117 119 121

4 Differentiated Product Oligopoly Models 5 Collusion References

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

List of Figures
1.1 1.2 1.3 1.4 1.5 1.6 1.7 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 3.20 3.21 3.22 3.23 History of IO. . . . . . . . . . . . . . . . . . . The Structure-Conduct-Performance paradigm . Equivalent variation and compensating variation Consumer surplus . . . . . . . . . . . . . . . . Consumer surplus as an approximation . . . . . Taxation and deadweight welfare loss . . . . . Monopoly and deadweight welfare loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 5 10 11 14 15 16 32 34 35 37 38 41 42 49 50 51 53 62 66 66 67 69 70 72 73 75 76 77 78

The space of outcomes. . . . . . . . . . . . . . . . . . On the meaning of Cournot equilibrium . . . . . . . . The Cournot model in extensive form. . . . . . . . . . Existence of Cournot equilibrium (i). . . . . . . . . . . Existence of Cournot equilibrium (ii). . . . . . . . . . Browers xed point theorem. . . . . . . . . . . . . . An instance of non-uniqueness of equilibrium . . . . . Firm is isoprot curves. . . . . . . . . . . . . . . . . Firm j s isoprot curves. . . . . . . . . . . . . . . . . The loci of Pareto optimal points. . . . . . . . . . . . . Cournot with increasing number of rms. . . . . . . . Cournot, monopoly and perfect competition equilibria. Examples where f is a contraction. . . . . . . . . . . . Stable equilibrium. . . . . . . . . . . . . . . . . . . . Unstable equilibrium. . . . . . . . . . . . . . . . . . . Firm is contingent demand. . . . . . . . . . . . . . . Firm is contingent prots. . . . . . . . . . . . . . . . Bertrand equilibrium. . . . . . . . . . . . . . . . . . . Cournot vs. Bertrand. . . . . . . . . . . . . . . . . . . The efcient rationing rule. . . . . . . . . . . . . . . . The proportional rationing rule. . . . . . . . . . . . . . Efcient vs. proportional rationing rule. . . . . . . . . Price wars with capacity constraints. . . . . . . . . . . ix

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

x 3.24 3.25 3.26 3.27 3.28 3.29 3.30 3.31 3.32 3.33 3.34 3.35

LIST OF FIGURES Prots over residual demand. . . . . . . . . . . . . Minimum market price. . . . . . . . . . . . . . . . Prices and residual demand. . . . . . . . . . . . . Edgeworth cycle. . . . . . . . . . . . . . . . . . . Reaction functions . . . . . . . . . . . . . . . . . Residual demand . . . . . . . . . . . . . . . . . . Sustainability vs. long-run competitive equilibrium. Demand and marginal revenue. . . . . . . . . . . . Marginal revenue and marginal cost. . . . . . . . . Reaction functions. . . . . . . . . . . . . . . . . . Stackelberg equilibrium. . . . . . . . . . . . . . . Markov perfect equilibrium. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 . 79 . 80 . 82 . 84 . 85 . 88 . 91 . 92 . 92 . 94 . 110

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

List of Tables
3.1 Payoff matrix of the Cournot game. . . . . . . . . . . . . . . . . 36

xi

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Preface

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Chapter 1 Introduction
Industrial Organization1, Industrial Economics, Oligopoly, Imperfect Competition, ... All these are well known labels to address one of the oldest problems in economics, namely how prices arise in the market when there are few competitors. We will start with a review of the ideas of the founding fathers of the oligopoly problem, Cournot, Bertrand, Edgeworth, Stackelberg, Chamberlin, Robinson, and Hotelling. Next we will present the contrast between the so-called classical Industrial economics with the modern industrial economics. Finally, we will address the issue of the adequacy of the partial equilibrium framework where most of the contributions in this area are developed.

1.1 A historical appraisal.

A complete account of the early ideas in what today is known as Industrial Organization can be found in Schumpeter (1958). For our purpose, Cournot (1838) was the rst in proposing a solution concept to determine market prices under oligopolistic interaction. By means of an example of two producers of mineral water deciding production levels and competing independently, Cournot proposes that the price arising in the market will be determined by the interplay of aggregate supply and demand. Also, such a price will be an equilibrium price when every producers production decision maximizes its prots conditional on the expectation over the production of the rival. It is worth noting that this equilibrium involves a price above the marginal cost of production. This concept of equilibrium is precisely what Nash (1950) proposed as solution of a non-cooperative game when we consider quantities as strategic variables. Next, Cournot tackles the case of complementary products. Interestingly enough he assumed in this case
This chapter is based on Martin (2002, ch. 1), Tirole (1988, Introduction), and Vives (1999, ch. 1)
1

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

1.1 A historical appraisal.

that producers would choose prices and applied the same solution concept, namely a Nash equilibrium with prices as strategic variables. In this case, the equilibrium price is larger than the monopoly price. Cournots contribution was either ignored or unknown for 45 years until Bertrand (1883) published his critical review where he claims the the obvious choice for oligopolists competing in a homogeneous product market such as the proposed by Cournot would be to collude, given that the relevant strategic variables must be prices rather than quantities. In particular, in Cournots example, the equilibrium price will equal marginal cost, i.e. the competitive solution. The criticism of the Cournot model continued with Marshall (1920) and Edgeworth (1897). Marshall thought that under increasing returns, monopoly was the only solution; Edgeworths main idea was that in Cournots set up the equilibrium is indeterminate regardless of products being substitutes or complements. For substitute goods with capacity constraints (Edgeworth (1897)) or with quadratic cost (Edgeworth (1922)) he concludes that prices would oscillate cycling indenitely. For complementary products the indetermination of the equilibrium is at least very probable (Vives (1999 p. 3)). This demolition of Cournots analysis was called to an end by Chamberlin (1929) and Hotelling (1929) after the observation that neither assumption of quantities or prices as strategic variables are correct in an absolute sense: Equilibrium in the Bertrand model with a standardized product is quite different from equilibrium in the Cournot model. The Cournot model emphasizes the number of rms as the critical element in determining market performance. Bertrands model predicts the same performance as in long-run equilibrium of a perfectly competitive market if as few as two producers supply a standardized product. The qualitative nature of the predictions of the Cournot model are robust to the introduction of product differentiation. The same cannot be said of the Bertrand model. (Martin (2002, p.60)). From that point Cournots model served as a departure point to other analysis. Hotelling (1929), Chamberlin (1933), and Robinson (1933) introduced product differentiation. Hotellings segment model introduces different preferences in consumers and provides the foundation for location theory by assuming consumers buying at most one unit of one commodity; Chamberlin and Robinson considered a large number of competitors producing slightly different versions of the same commodity (thus allowing them to retain some monopoly power on the market) and assumed that consumers had convex preferences over the set of varieties. Stackelberg (1934) considered a sequential timing in the rms decisions, thus incorporating the idea of commitment.

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

Some years later, von Neumann and Morgenstern (1944) and Nash (1950, 1951) pioneered the development of game theory, a toolbox that provided the most ourishing period of analysis in oligopoly theory along the 1970s. Renements of the Nash equilibrium solution like Seltens subgame perfect equilibrium (1965) and perfect equilibrium (1975), Harsanyis Bayesian Nash equilibrium (1967-68), or Kreps and Wilsons sequential equilibrium (1982) have proved essential to the modern analysis of the indeterminacy of prices under oligopoly. Also, the study of mechanisms allowing to sustain (non-cooperative) collusion was possible with the development of the theory of repeated games lead by Friedman (1971), Aumann and Shapley (1976), Rubinstein (1979), and Green and Porter (1984). Figure 1.1 summarizes this discussion.

1.2 Oligopoly theory vs. the SCP paradigm.

Industrial Economics, as we have already mentioned, deals with the study of the behavior of rms in the market. The eld as a separate area within microeconomics appears after the so-called monopolistic competition revolution, linked to the names of Mason (1939) and Bain (1949, 1956) (Harvard tradition). Barriers to entry was the central concept giving rise to market power. The approach is essentially motivated by stylized facts arising from an empirical tradition seeking how the structural characteristics of an industry determine the behavior of its producers that, in turn, yields market performance. This framework of analysis is known as the Structure-Conduct-Performance paradigm. Martin (2002), p.6 reproduces the gure 1.2 from Scherer (1970) showing the SCP paradigm. Schmalensee (1989) provides a very nice survey of this approach. This paradigm dominated the evolution of the eld for three decades. During these years research was mainly discursive and informal and independent of the formal microeconomic analysis of imperfect markets. Basically, the SCP provided a general framework allowing the implementation of public policies from empirical regularities observed in many industries. The early seventies witnessed a major revolution in the analysis, leading to the so-called new industrial economics. Following Martin (2002), p.8, three factors are behind this evolution. (i) the conclusions of the formal microeconomic models are not qualitatively different from those of the SCP paradigm.; (ii) empirical economists held that market structure should be treated as endogenous rather than exogenous with respect to conduct and performance. This raised the need for a theoretical foundation of the econometric models (to be found in the microeconomic models of oligopoly); (iii) last but not least, the application of game theory to the modeling of oligopolistic interaction provided the denite element to replace the SCP paradigm and place Oligopoly Theory (understood as the analysis of strategic interactions being cen-

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

1.2 Oligopoly theory vs. the SCP paradigm.

Figure 1.1: History of IO.

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

Figure 1.2: The Structure-Conduct-Performance paradigm .

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

1.3 Variations of prices and welfare.

tral to the determination of market performance) and the standing methodology. The two decades from the early 1970s until the late 1980s has been the most ourishing period of theoretical development in industrial organization. The main methodological difference with respect to the SCP paradigm is that gametheoretical models are rather specic and their predictions about equilibrium behavior often not robust to minor changes in the set of underlying assumptions. Most of the literature on oligopoly theory has been developed using models of partial equilibrium2. That is the model focuses in an industry and the interactions with the rest of the economy are neglected. This approach goes back to Marshall (1920). His idea is that the partial equilibrium model only makes sense when the income effects are small. In this case the share of consumer expenses in the industry under analysis will be small and small changes in the industry should not give rise to variations in the other markets of the economy. Vives (1999, section III.2) presents a rigorous foundation for these ideas. Following Cournot, it is generally assumed that rms face a downward sloping demand curve (except in the models of spatial competition). Also, it is assumed that welfare changes are adequately measured by variations in consumer surplus, a concept introduced by Dupuit (1844). We will study this last concept and will see that the consumer surplus is a precise measure of the change in consumer welfare only when preferences are quasilinear. Then we will verify that such assumption is meaningful only when the income effects are small, that is when the share of consumer expenses in the industry under analysis is small.

1.3 Variations of prices and welfare.

Variations in the economic environment (price changes, taxes, etc) give rise to variations in the consumers welfare. Thus, it is reasonable to try to obtain quantitative estimations of those changes in prices and welfare with clear economic interpretations. The classic and most used measure of welfare variation is the consumer surplus. The problem with this measure is that it is precise only in the special case of quasilinear preferences. En general, consumer surplus only gives an approximation of the impact on welfare of a variation of some basic magnitude of the economy. Therefore, before focussing the attention in the consumer surplus, we
There is however a whole line of general equilibrium models of oligopoly started by Negishi (1961) and continued by Gabszewicz and Vial (1972), Shitovitz (1973), Novshek and Sonnenschein (1978), Mas-Colell (1982), and Codognato and Gabszewicz (1991). See also a survey paper by Bonanno (1990).
2

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

will examine some more general methods. These are the compensating variation and the equivalent variation.

1.3.1 Price indices.

To start at the beginning, let us suppose that the basic economic magnitude suffering a variation are prices. Thus, we have to construct some measures of that variation that will prove useful in the study of the impact on welfare. These are the price indices3 . Denition 1.1 (Price index). A price index measures the impact on the welfare level following a price variation.
0 Let us consider two price vectors p0 , p1 IRl + , where p represents the initial situation and p1 the new price level after the variation. How can we measure the impact of this price variation on the cost of living? A rst approach consists in considering a reference consumption bundle, xR i and evaluate it at both systems of prices. We obtain an index of the following type:

P I (p0, p1 , xR i ) =

p1 xR i . p0 xR i

(1.1)

1 This index measures the cost of the bundle xR i at prices p with respect to the cost of this very bundle at prices p0 . The relevance of the index thus obtained depends on how representative is the bundle xR i in the economy. Two indices built in this fashion are linked to the names of Laspeyres and Paasche. The difference between them is the reference consumption bundle. The former uses the bundle in the initial period x0 i , the latter the resulting bundle after 1 the price variation, xi :

p1 x0 i , p0 x0 i p1 x1 i P IP (p0 , p1 , x1 ) = . i p0 x1 i P IL (p0 , p1 , x0 i) = The drawback of this family of indices is that given that the consumption bundle is xed, they cannot capture the substitution effects associated with the price change. An alternative family of indices overcoming this problem should estimate the impact of the price change on the utility level. The natural way of constructing such an index should use the expenditure function, that measures the cost of
3

see Villar (1996, pp.72-73)

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

1.3 Variations of prices and welfare.

reaching a certain utility level at a given prices. Therefore, from a utility level of reference uR i we can construct the so-called true price index as T P I (p 0 , p1 , u R i ) = ei (p1 , uR i ) . R 0 ei (p , ui )

Naturally, the usefulness of this index depends on how representative is the utility level uR i . With the same logic behind the Laspeyres and Paasche indices, we can obtain the corresponding true Laspeyres price index and true Paasche price index. ei (p1 , u0 i) , 0 ei (p , u0 i) ei (p1 , u1 i) . T P IP (p0 , p1 , u1 ) = i 0 ei (p , u1 i) T P IL(p0 , p1 , u0 i) = It is easy to prove (this is left to the reader) that the true Laspeyres price index is a lower bound of the Laspeyres index, and that the true Paasche price index is an upper bound of the Paasche price index, that is, T P IL P IL, T P IP P IP . Finally, note that when preferences are homothetic, the true Laspeyres and Paasche price indices coincide (again, the proof of this statement is left to the reader).

1.3.2 Welfare variations

We will examine the effect of a price change (due, for instance to a variation in taxes) on the welfare level of an individual. Let us consider, as before, two price 0 1 vectors p0 , p1 IRl + , where p represents the initial situation and p the new situation. Let us also assume that wealth remains constant in both scenarios. This is a simplifying assumption. To see the effect of the price change on consumers we only have to compare the utility levels in both situations evaluated at the cor1 responding consumption bundles, ui(x0 i ) and ui (xi ). Similarly, we can compare 0 1 the indirect utility levels vi (p , wi) and vi (p , wi ). These comparisons are ordinal, that is they only tell us whether the consumer is better off or worse off after the price variation, but do not tell us anything about how much better off or worse off the consumer is. To overcome this limitation of the analysis, we can consider the expenditure function as representation of the indirect utility function. Thus, let us consider a reference price vector pR together with the price vectors p0 y p1 . Next, let

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

us compute ei [pR , vi (pj , wi)], j = 0, 1. These functions tell us the amount of j money, given prices pR , necessary to achieve the utility level uj i = vi (p , wi ). Given that the expenditure function is strictly increasing in ui , we can think of the expenditure function as a monotonic increasing transformation of vi and therefore an alternative representation of the individual utility. This argument allows us to express the indirect utility in monetary units (Euros), and thus obtain a quantitative measure of welfare variation. In other words, the difference, ei [pR , vi (p1 , wi )] ei [pR , vi (p0 , wi )]

tells us how much does our welfare level vary when prices change from p0 to p1 in Euros relative to the price vector pR . Of course, the selection of pR is crucial to obtain a meaningful interpretation. The obvious candidates are the prices corresponding to the initial situation or to the nal situation. As in the case of the price indices, this gives rise to two different measures of welfare variation. Before introducing these measures, let us recall that we are assuming that wealth remains constant form one situation to the other, namely ei [p0 , vi (p0 , wi )] = ei [p1 , vi (p1 , wi )] = w. Denition 1.2 (Equivalent variation). The equivalent variation is the change in consumer wealth equivalent, in terms of welfare, to the price change: EVi (p0 , p1 , w ) = ei [p0 , vi (p1 , wi )] ei [p0 , vi (p0 , wi)] = ei [p0 , vi (p1 , wi )] w. The equivalent variation tells us that the price change from p0 to p1 has the same impact on welfare as an income change from wi to (wi + EVi ). Therefore, EVi will be negative when the price change will worsen the situation of the consumer and positive otherwise. Denition 1.3 (Compensating variation). The compensating variation is the change in consumer wealth necessary to maintain that consumer in the initial welfare level after a price change has occurred: CVi (p0 , p1 , w ) = ei [p1 , vi (p1 , wi)] ei [p1 , vi (p0 , wi )] = w ei [p1 , vi (p0 , wi)]. The compensating variation is a modication in the wealth of the individual to maintain him (her) in his (her) initial utility level. Hence, that modication will be negative (an increase in income) when the change of prices will worsen the situation of the consumer and positive otherwise. We can thus conclude that when we compare two scenarios, both measures go in the same direction, although not in the same magnitude given that the price vectors at which both scenarios are evaluated are different. This statement does not hold true when we compare more than two situations. In that case the EVi turns out to be a better measure than the CVi (see Villar (1996, p.75)).

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

10

1.3 Variations of prices and welfare.

Figure 1.3: Equivalent variation and compensating variation Figure 1.3 illustrates the argument when the price of good 1 decreases from p0 to p1 . Section (a) in the gure represents the equivalent variation of income, i.e. how much additional money is needed at the price vector p0 to maintain the consumer at the same welfare level as with prices p1 . Part (b) represents the compensating variation of income, that is how much money do we have to subtract from the consumer to maintain him (her) at the same welfare level as with prices p0 .

1.3.3 Consumer surplus

The concept of consumer surplus gives an approximation of the impact of a change of prices on consumer welfare. In contrast with the equivalent variation and the compensating variation, it is easier to compute because it uses the demand function. However, it only allows to obtain an approximation to the true value (except in one particular case that we will examine below). To illustrate the idea of consumer surplus, let us consider a market of a good where a monopolist knows the demand curve of the consumer. This monopolist by setting a price p0 would sell x0 units, so that its revenue would be p0 x0 . Let us assume that the monopolist would like to sell precisely these x0 units to the consumer. In an effort to maximize prots, and given that the monopolist knows the demand function, it can sell every unit separately until it reaches the quantity x0 . According to demand function, it can sell the rst unit at a much higher price than the competitive price, the second unit to a slightly lower price, and so on until reaching the x0 th unit that sells at the price p0 . The difference in revenues obtained by the monopolist using this discriminatory mechanism and the uniform

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

11

Figure 1.4: Consumer surplus price is called consumer surplus. This surplus captures the rents the the consumer saves because the rm cannot set a price for every unit that the consumer buys. In a similar way, we can compute the consumer surplus when the price of the good varies. Figure 1.4 illustrates the argument. Consider an initial situation where the price of the good is p0 . At this price, the consumer, given his (her) demand function, buys x0 units. Assume that for some reason, the price increases to p1 so that the consumer, whose income remains constant, reduces its demand to x1 . The colored area illustrates the variation of the surplus of our consumer and offers an idea of the impact of the price change of his (her) welfare. Formally, the consumer surplus in the case of gure 1.4 is given by,
p1

CSi =
p0

x(t)dt

The consumer surplus coincides with the equivalent variation and the compensating variation when preferences are quasilinear. For any other preferences, the consumer surplus only offers an approximate measure bounded by the equivalent variation and the compensating variation. We will analyze both situations in turn. Quasilinear utility. Assume our consumer lives in a world of two goods with prices p1 = 1 and p2 . Also, assume his (her) income is wi and the utility function can be represented as Ui (xi1 , xi2 ) = xi1 + ui (xi2 ), so that the utility function is linear in one good. Assume the function ui (xi2 ) is strictly concave.

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

12 The problem of the consumer is

xi1 ,xi2

1.3 Variations of prices and welfare.

max xi1 + ui (xi2 )

s.a xi1 + p2 xi2 = wi xi1 0. This problem may have two types of solution according to the consumption of good xi1 be positive or zero. Consider rst xi1 > 0. We can reformulate the problem as,
xi1 ,xi2

max xi1 + ui (xi2 )

s.a xi1 + p2 xi2 = wi or also, max wi p2 xi2 + ui (xi2 )

xi2

The rst order condition, = p2 , tells us that demand of good 2 only depends on its own price and is independent of income. In other words, we can write its demand as xi2 (p2 ). We obtain the demand of good 1 from the budget constraint xi1 = wi p2 xi2 (p2 ). wi When xi1 = 0, demand of good 2 is simply xi2 = . p2 How does the consumer decide his (her) consumption plan? Given that the utility (sub)function on good 2 is strictly concave, the consumer will start consuming good 2 until the marginal utility of an additional euro spent in good 2 will be equal to p1 = 1. From that point on, the increases in income will be devoted to consumption of good 1. For the sake of the argument, let us assume that initially our consumer has zero income and we increase it marginally. The increase in util u (wi /p2 ) ity is i . If this increase in utility is larger than 1 (the price of good 1), p2 the consumer obtains more utility consuming good 2. This behavior will remain the same until the marginal increase in income make the marginal utility of that income spent in good 2 equal to the price of good 1. Then our consumer will be indifferent between consuming either good. From that point on, further increases in income will be devoted to increase the consumption of good 1. The level of utility (welfare) obtained by the consumer is simply the sum of the utility derived from the consumption of every good, i.e., ui (xi2 ) Ui (xi1 , xi2 ) = wi p2 xi2 (p2 ) + ui (xi2 (p2 )). To illustrate this welfare level on the demand curve of good 2, we only need to integrate, wi p2 xi2 (p2 ) + ui(xi2 (p2 )) = wi p2 xi2 (p2 ) +
xi2

p(t)dt.
0

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

13

Leaving aside the constant wi , the expression on the right hand side of this equation is the area under the demand curve of good 2 and above price p2 . The general case When the utility function representing the preferences of our consumer is not quasilinear, the consumer surplus can only offer an approximation to the welfare variation associated to a price change. Recall that the equivalent variation and the compensating variation of the consumer when the price of a good varies from p0 to p1 (given the prices of the other goods and consumer income) are: EVi (p0 , p1 , w ) =ei [p0 , vi (p1 , wi )] ei [p0 , vi (p0 , wi )] CVi (p0 , p1 , w ) =ei [p1 , vi (p1 , wi )] ei [p1 , vi (p0 , wi )]. Also, recall that the compensated demand function is the derivative of the exi , so that we can rewrite the equivalent variation penditure function, hi (p, ui) e p and the compensating variation as,
p0

EVi (p , p , w ) =ei [p CVi (p , p , w ) =ei [p

0 1

, u1 i] , ui 0]

ei [p ei [p

, u0 i] , u0 i]

=
p1 p0

hi (p, u1)dp, hi (p, u0)dp,

p1

that is, the compensating variation is the integral of the compensated demand curve associated at the initial utility level while the equivalent variation is the corresponding integral associated at the nal utility level. The correct measure of welfare is thus an area given by a compensated demand function. The problem, as we already know, is that such demand is unobservable. This is the reason why the consumer surplus, obtained on the (observable) marshallian demand is often used as an approximation. The question that remains is how good this approximation is. To answer this question we start by recalling the Slutsky equation, hij (p, u) xij xij = xik (p, wi ). pk pk wi When the good is not inferior, i.e. xij > 0, we obtain wi

xij hij (p, u) < . pk pk

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

14

1.4 Producer surplus and deadweight welfare loss.

Figure 1.5: Consumer surplus as an approximation that is, the slope of the compensated demand is larger that the slope of the marshallian demand. Figure 1.5 shows the relationship between the equivalent variation, the compensating variation, and the consumer surplus. The initial situation is given by a price p0 , so that consumer is at a on the marshallian demand curve. The nal scenario appears after a decrease in price to p1 , so that consumer is at point b on the marshallian demand curve. The compensating variation is computed from the initial utility level, u0 , and is given by the area below the compensated (hicksian) demand at the point a, that is the area p0 acp1 . The equivalent variation is computed from the nal utility level, u1 , and is given by the area below the compensated (hicksian) demand at the point b, that is the area p0 dbp1 . The consumer surplus is the area below the marshallian demand curve between points a and b, that is the area p0 abp1 . Comparing these areas we realize that CV CS EV . In particular, if xij = 0, (this is the case of quasilinear utility) the there are no income effects wi three areas will coincide. This means that for small income effects, the consumer surplus represents a good approximation to the equivalent variation and to the compensating variation.

1.4 Producer surplus and deadweight welfare loss.

The producer surplus is the prot of the rm in the industry (net of x costs). The next two gures show the marginal cost curve (i.e. the supply curve under perfect

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

15

Figure 1.6: Taxation and deadweight welfare loss

competition). Prot is the difference between revenues (p0 x0 ) and cost, where total cost is the integral of marginal cost (recall we are assuming away xed/sunk costs). Accordingly, prot is the area between the marginal cost curve and the horizontal line at price p0 . The aggregate welfare of the economy is the sum of the consumer surplus and the producer surplus. Looking at gures 1.6 and 1.7 it is easy to understand that total surplus is maximized when price equals marginal cost. Any deviation of the price away from the marginal cost represents a welfare loss. A monetary measure of this loss of welfare is the so-called dead-weight loss. Figure 1.6 shows the dead-weight loss associated with the introduction of a tax on a commodity. Assume an initial state where the economy is perfectly competitive, so that the equilibrium price is p0 . Then a unit tax tis imposed on each unit sold. This raises the equilibrium price to p1 and lowers the equilibrium consumption to x1 . The welfare loss is thus the difference in total surplus between both situations. This is the area of the blue/green triangle. It is given by t(x0 x1 )/2. The second example shows the dead-weight loss associated with the transition from a competitive economy to a monopolized economy. As before the initial situation is a perfectly competitive economy with an equilibrium price p0 . A monopoly would set a price equation marginal revenue and marginal cost, that is p1 . The dead-weight loss is the area of the blue/yellow triangle.

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

16

1.5 Market and market power.

Figure 1.7: Monopoly and deadweight welfare loss

1.5 Market and market power.

1.5.1 Denition of the market.
Dening a market is not an easy task. It is obvious that we do not want to restrict ourselves to the case of a homogeneous good. A rst approximation could be the principle that two goods belong to the same market only if they are perfect substitutes. That would be too a restrictive denition because it would have as consequence that there would be only one rm in each market. But very few rms have an absolute monopoly power. A common feature in real markets is that consumers after the increase of the price of one good react (partially) substituting that good by purchasing other alternative goods. The denition cannot be too general either. Considering any two substitute goods as belonging to the same market would lead to an economy with a single market since any good can be directly or indirectly a potential substitute of any other good. Such a denition would not allow partial equilibrium analysis. At this point thus, we realize that the correct denition of a market has to be contingent to the problem we want to tackle. For instance, let us consider the case of coal. If the problem we face is the design of energy policy, the relevant market is the energy market (including coal, petrol, nuclear, . . . ); to evaluate the effects of a merger between two providers, we would need a narrower denition of the relevant market. An ideal scenario to dene a product market consists in having a set of goods with very high cross price elasticities (in absolute value) among them and very low with respect other commodities not in the set. This denition refers to demand

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

17

elasticities but also in a subsidiary way to supply elasticities. Let us consider some examples: (i) cross elasticity between lead free 95 and 98 octanes gasoline is very high. They are two close substitutes that clearly should belong to the same market; (ii) cross elasticity between consumption of gasoline and mineral water is very low. They are two independent products. They should not belong to the same market; (iii) cross elasticity between shoes for the right foot and the left foot is (in absolute value) very high: they are two perfect complements that must belong to the same market (actually, in this extreme case the correct denition of the market should be that of pairs of shoes). This rule involving elasticities even though contains clear ambiguities (it is not clear what does sufciently high cross elasticity means) is not always easy to apply. First, product differentiation is a gradual phenomenon and therefore determining the critical value of the cross elasticity is not easily determined. Second, we should be aware that two products may be substitutes indirectly through a chain of substitution as it appears often in the pharmaceutical products: a certain drug is effective in treatments A and B ; another one is effective in treatments A and C . Both the strict denition of market based on therapeutical effectiveness (A, B, C ) as the more general denition (A + B + C ) are inconsistent. It is also worth mentioning that a market denition using a geographical criterion raises similar problems. For example, what is the proper denition of the market for wine: the world, Europe, Spain, Catalonia, Barcelona? Despite all these difculties, for statistical purposes there exist classication systems of the economic activities in every country. In Spain we have the so called Clasicaci on Nacional de Actividades Econ omicas, CNAE among others (see appendix). We can also mention the classication system NACE proposed by the Eurostat. These classications are divided in sections from one digit, until four digits. Often these classication systems are used as proxies for market denitions. This is a problem because usually the grouping in these systems are done using criteria from the supply side of the market, while the denition of the market emphasizes the demand side of the market. For example, production of wine and production of cava belong to different groups (because the technology is different) but from the demand point of view should be considered as belonging to the same market. The supply side aspects in the denition of a market offers some advantages from the industrial organization point of view. A well known example is provided by McKie (1985) (see Cabral (1994), p.21): In 1964 the US Air force opened a

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

18

1.5 Market and market power.

contract for the provision of a certain type of radar. The contract was assigned to Bendix, enterprise that maintained a monopoly status for some years. This situation led a second rm, Wilcox, to sue Bendix for abuse of monopoly position. The Federal Trade Commission voted in favor of Bendix. The reason was that from the demand point of view, looking at the elasticity of demand, Bendix can be considered a monopoly. Nevertheless, according to the classication of industrial activities we nd a certain number of rms with similar technological capacity to Bendix. Therefore, any of them could win the next contract when it would become public. Actually, this is what happened in 1969 when Honeywell obtained the new contract.

1.5.2 Concentration measures.

Once we have dened the market (the industry) both from the cross elasticities and from the supply side perspectives, we also need a measure of the relative importance of every rm in the market, and a statistical method to compute an index giving information on the degree of concentration of the market. Even though there are obvious difculties to agree upon a criterion to measure the relative size of a rm (see e.g. Hay and Morris (1996) or Eraso Goicoechea and Garcia Olaverri (1990)), one possibility is to use the market shares. Consider an industry with n rms producing a homogeneous good. The distribution of these rms according to their market share is given by a vector m ordered from biggest to smallest: m = (m1 , m2 , , mn ), mi = qi
n i=1 qi n

0,

mi = 1.
i=1

That is, we are working in a unit simplex Sn1 in IRn . For every number n a concentration measure is a real application Cn dened on Sn1 . Following Encaoua and Jacquemin (1980), any proper concentration measure has to satisfy some requirements and properties. Requirements Unidimensionality. The concentration measure must take values in [0, 1]. Independence of the market size. Symmetry: the concentration measure has to be invariant to permutations of the market shares of the rms, that is, Cn (m1 , m2 , , mn ) = Cn (m(1) , m(2) , , m(n) )

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction for any permutation de {1, 2, , n} Properties n x

19

P1 (Transfer principle): Transferring part of the production from a smaller rm to a bigger one cannot decrease the concentration measure, i.e. Cn (m1 , . . . , mj , , mk , , mn ) Cn (m1 , . . . , mj , , mk , , mn ), where mj > mj , mk > mk , and mj mj = mk mk . P2 (Homogeneity principle): Given the number of rms in the industry, the concentration measure takes its minimum value when all rms have the n times same market share, i.e. min Cn (m1 , m2 , , mn ) = Cn (m, m, , m). P3 (Lorenz criterion): consider two industries with the same number of rms. Let the aggregate production of the k, (k = 1, 2, n) bigger rms in the rst industry be larger than or equal to the corresponding aggregate production of the k bigger rms in the second industry. Then, this inequality must also hold between the concentration measures of the two industries, k 1 1 2 2 2 1 Formally, Cn (m1 1 , m2 , , mk ) Cn (m1 , m2 , , mk ) when i=1 mi k 2 i=1 mi . n variable P4: If two or more rms merge, the concentration measure cannot diminish. Formally, Cn (m1 , , mj , , mk , , mn ) Cnl (m1 , . . . , mjl , , mn ) P5: If in an industry all rms market shares are equal (mi = mj , i = j, i, j = 1, 2, , n), the concentration measure cannot be increasing in the number of rms, i.e. Cn (m, , m) Cn+l (m, , m) These properties are not independent. Encaoua and Jacquemin show the following implications: If P1 holds, then P2 i P3 also hold (P 1 P 2, P 1 P 3) If P2 and P4 hold, then P5 also holds (P 2 P 4 P 5) Denition 1.4 (Concentration measure). Let h(mi ) be a function dened in [0, 1] assigning a weight to the relative production of a rm i, i.e. mi mi h(mi ). A concentration measure is dened as,
n

Cn (m1 , m2 , , mn ) =

mi h(mi )
i=1

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

20

1.5 Market and market power. We can recall now the most usual concentration indices. Concentration ratio: Ck =
i=1 k

mi .

This is a measure that considers h(mi ) = 1. This index measures the relative production of the k largest rms in the market over the total production k of the industry. The index takes values in the interval Ck [ n , 1]. The problem of this index is the arbitrariness of k . Herndhal index: CH =
i=1 n

m2 i.

This index denes h(mi ) = mi . Therefore, it is a measure that overestimates the relative importance of the large rms against the smaller ones. 1 , 1]. Also, the number C1 The index takes values in the range CH [ n H is called Adelmans equal number associated to the Herndhal index. It represents the number of equal sized rms whose distribution results in the same concentration measure as the one given by CH . Note that the Herndhal index is dened over all rms in the industry. Entropy index: CE =
i=1 n

mi loga mi , a > 1.

This index denes h(mi ) = loga mi . Therefore, it is a measure that underestimates the relative importance of the large rms against the smaller ones. 1 The index takes values in the range CE [loga n , 0]. Also, the number 1 is called the equal number associated to the entropy index. It repreaCE sents the number of equal sized rms whose distribution results in the same concentration measure as the one given by CE .

1.5.3 Degree of Monopoly.

The most popular measure of monopoly power of a rm was proposed by Lerner (1934) and thus called Lerners index: Li = P MCi MC = 1 P P

The index takes values in the range Li [0, 1). When a rm behaves competitively, its price equates its marginal cost, Li = 0. As the rm increases its ability

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

21

to set a price above marginal cost the index increases. In the limit, when the margin of the price over the marginal cost is innitely large, Li 1. From the individual Lerner indices in an industry, we can obtain an aggregate index of monopoly power. Let n (L1 , L2 , , Ln ) be the aggregate index of monopoly power in an industry with n rms. This index has to satisfy three properties: 1.- The value of n (L1 , L2 , , Ln ) must lay in the range dened by the extreme values of the distribution of individual Lerner indices (L1 , L2 , , Ln ) i.e. max{L1 , L2 , , Ln } n (L1 , L2 , , Ln ) min{L1 , L2 , , Ln }. If L1 = L2 = = Ln = L, then n (L1 , L2 , , Ln ) = L. This result has two interpretations. Either the industry is perfectly competitive or it is a perfect cartel. In the latter case, the optimal assignment of market shares is such that the marginal costs of the different rms coincide; even if market shares are not equal, the individual Lerner indices coincide. Hence, it is important to make the distinction between the distribution of market shares and the distribution of monopoly power in an industry. 2.- In an industry, some of its members may be price takers so that their monopoly power are nil, i.e. Li = 0. Even though, these components must also be included in n (L1 , L2 , , Ln ), because each member of the industry has its weight in the computation of the aggregate index. In other words, if Ln = 0, n (L1 , L2 , , 0) = n1 (L1 , L2 , , Ln1 ). 3.- If two or more rms merge, the aggregate index of monopoly power must not decrease, i.e. n (L1 , L2 , , Ln ) n1 (L1,2 , L3 , , Ln ). Some indices satisfying these properties are:
k

Aggregate concentration ratio: 1 k = n Aggregate Lerner index a =

n

Li
i=1

mi Li
i=1

This is an arithmetic average so that large rms get more weight. Aggregate entropy index g =
n

(Li )mi , Li = 0
i=1

This is a geometric average so that large rms get less weight.

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

22

1.5 Market and market power.

1.5.4 Concentration indices and degree of monopoly.

Consider a homogeneous industry with n rms. Market (inverse) demand is given by p = f (q ) where q = n i=1 qi . Every rm has a technology described by Ci (qi ) and a production capacity limit vi . Firm is prots are, i (q1 , q2 , , qn ) = qi f (q ) Ci (qi ), where qi [0, vi ]. Assuming the proper conditions on demand and cost functions, a noncooperative Cournot equilibrium4 will be interior and unique, i.e. 0 qi (0, vi ), i, so that the system of rst order conditions of the prot maximization problem evaluated at the equilibrium satises,
0 0 0 f (q 0 ) + qi f (q ) Ci (qi ) 0 (i = 1, 2, , n),

where q 0 =

n 0 i=1 qi .

Rearranging terms, we can write,

0 0 0 f (q 0 ) Ci (qi ) = qi f (q )

(i = 1, 2, , n).

Dividing both sides by f (q 0) and multiplying and dividing the right hand side q 0 , we obtain,
0 0 f (q 0 ) 0 qi f (q 0) Ci (qi ) = q f (q 0 ) f (q 0 ) q 0

(i = 1, 2, , n).

The left hand side of this expression is rm is Lerner index evaluated at the equilibrium output vector. The right hand side contains two elements:
0 qi m0 i i.e. rm is equilibrium market share, and q0

f (q 0 ) 0 q0 f (q 0 ) q 0 1 0 q = f ( q ) = = . 0 0 0 0 f (q ) f (q ) q f (q )

Therefore, we can write the rst order condition evaluated at the equilibrium production plan as 1 0 m. (1.2) L0 i = i Now we can show the (direct) relationship between a concentration index and an adequate aggregate index of monopoly power:
4

see chapter 3 on the concept of noncooperative equilibrium.

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction Concentration ratio 1 k = n Herndhal index a = Entropy index 5

n n
i Lm = i

23

i=1

11 Li = n

mi =
i=1

11 Ck ; n

i=1

1 mi Li =

m2 i =
i=1

1 CH ;

g =

i=1

i=1

1 mi 1 mi = aCE .

All these indices are reasonable in the sense that they satisfy the axioms proposed by Encauoa and Jacquemin (1980), but nothing is said on their usefulness. Are they a useful instrument for policy design? To answer we can think of relating these indices with the productivity of the industry. In particular, we can examine the relationship between concentration and industry prots. Start assuming all rms in the industry with the same market shares (symmetry). The only reasonable concentration measures are equivalent to the number of k 1 1 rms in the industry: Ck = n , CH = n , CE = ln n . Bertrands model tells us that market price and industry prots are independent of the number on rms in the industry. Thus, concentration and protability are not related. Cournots model (with xed number of rms) shows a negative correlation between concentration and protability. If rms are asymmetric, because they may have different costs for instance, the concentration measure is not ambiguous any more. Assume, to illustrate, that rms have constant marginal costs, Ci (qi ) = ci qi , and compete in quantities. The aggregate industry prots are:
n n n

=
i=1
5

i =
i=1

(f (q ) ci)qi =
n

i=1

f (q )mi qi f (q )q =
n

m2 = i
i=1
n

f (q )q CH ,

CE =
i=1

mi loga mi =
i=1

i loga mm = loga i

i mm i ;

i=1 n

taking antilogs a
CE

=
i=1

i mm i

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

24

1.5 Market and market power.

where we have used (1.2). Assume also that consumers spend a x proportion of their incomes in this market, that is, pq = k where k is a positive constant. Then, demand elasticity = 1, and the above expression reduces to = kCH . In this particular case, the Herndhal index gives an exact measure (up to a proportionality constant) of industry protability. The asymmetries between rms tend to generate high correlation between concentration indices and industry protability. This is so because asymmetries in costs generate asymmetries in production levels and thus increase the concentration indices. Also, the most efcient rms in the industry obtain rents increasing the global industry prot. For instance, under Bertrand competition with constant marginal cost the rm with the lowest marginal cost gets all the market (so that concentration index will be highest) and obtains positive prots. When rms are more symmetric concentrations indices usually are not so high and rms obtain small prots. This phenomenon also appears under Cournot competition. Summarizing, concentration indices are useful because they are easy to compute and give an economic interpretation of how competitive an industry is. Unfortunately, there is no systematic relation between these indices and the relevant economic variables to evaluate changes in technology, demand, or economic policies. Even if it would be possible, we should be aware that the indices are endogenous measures, so that correlations could not be interpreted in a causal sense.

1.5.5 Volatility measures

A limitation of the concentration measures is their static character. The introduction of dynamic considerations in the analysis of market structure leads to the so-called volatility measures. The degree of competitiveness in a market is related not only with the distribution of market shares, but also with the relative position of the rms as time goes by. Assume that a certain market has at any point in time a dominant rm, but that rm varies in the different periods. It is quite likely that this market approaches more a competitive behavior than another market with less concentration but with more stable position of the rms in the market. One of the most popular measures of dynamic competitiveness (or stability of market shares) is the instability index. It is dened as, 1 I 2
n

i=1

|mi2 mi1 | [0, 1),

where mi2 and mi1 represent rm is market shares in periods 1 and 2 respectively, and n denotes the number of rms in any period. It is easy to verify that I varies between 0 (minimum instability) and 1 (maximum instability). When

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

Introduction

25

market shares remain constant along time we obtain I = 0. The situation where I = 1 corresponds to a scenario where all rms present in the market in the initial period have zero market share in the next period (i.e. none remains in the market). Naturally, this instability index also presents some problems of measurement and interpretation.

Exercises
1. Show a) T P IL P IL and T P IP P IP .

b) when preferences are homothetic, then T P IL = P IL and T P IP = P IP . 2. Consider a quasilinear utility function Ui (xi1 , xi2 ) = xi1 + ui (xi2 ) where ui (xi2 ) is strictly concave. Show that CV = CS = EV . 3. Show that CH = 1 + nV (mi ) where CH denotes the Herdahl index, n n the number of rms and V (mi ) the variance of market shares. From the equation above, provide an interpretation to the Adelman number dened 1 as CH .

4. Compute the extreme values of the (i) concentration ratio, (b) Herndhal index, and (iii) entropy index. 5. Consider a n-rm Cournot oligopoly where each rm has a constant marginal cost ci . Market demand is described by a well-behaved function p = f (Q). Show that the ratio between industry prots and industry revenues equals the ratio between the Herndhal index and the elasticity of demand. 6. Show that under the conditions of problem 2, the average Lerner index ( i mi Li ) equals the Herndhal index divided by the demand elasticity. 7. Consider a market with linear demand Q = 1 P . Two rms operate with constant marginal costs, c1 and c2 such that c1 + c2 = 2c where c is a constant. Show that when rms become more asymmetric (i.e. ci moves away from c) Cournot competiton yields a higher concentration index and a higher industry prot. 8. Consider a market of a certain (homogeneous) product described by the list of rms with market shares above 2% as shown in the following table:

Lecture Notes on Industrial Organization - Xavier Martinez-Giralt -- http://pareto.uab.es/xmg/Docencia/IO-en/IO-Introduction.pdf

26 Firm 1 2 3 4 5 6 7 Share 14.19 12.71 11.02 10.56 9.50 7.92 3.00 Firm Share 8 2.60 9 2.54 10 2.50 11 2.14 12 2.10 13 2.03

Exercises

These 13 rms together cover 82.81% of the market. Compute the upper and lower bounds of the Herndhal index for this market.