Slides Motta

Economics of Competition Policy Massimo Motta

Master in Competition and Market Regulation Barcelona GSE January-Feruary 2009 1

Outline of the lectures Massimo Motta: 1. Introductory elements 2. Cartels and horizontal agreements 3. Horizontal mergers 4. Vertical restraints and vertical mergers 5. Price discrimination - A brief introduction to abusive practices (this topic will be then developed by Jorge Padilla in the second part of the course) 2

Methodological note: in these lectures I will try to explain relatively sophisticated economic concepts by relying on very simple models and intuitions. I will also occasionally refer to relevant cases.

Textbook: Motta, Massimo Competition Policy. Cambridge U.P., 2004 3

1) Introductory elements a) Introductory remarks (why do we need Competition Policy?) b) Market power and welfare c) Market power and market definition

4

1a) Competition Policy: Introduction What is Competition Policy? Free entry does not solve all the problems Objectives of competition policy

5

What is competition policy? Definition: Competition Policy aims at ensuring that competition in the marketplace is not restricted in a way that is detrimental to society Why do we need a competition policy? Market failure also in markets without natural monopoly features. Even if entry is possible, dominant positions might persist, due to: • sunk costs industries (see below) • lock-in effects and switching costs • network effects (see below) 6

What is competition policy, II We need competition policy also because: Un-monitored, firms may resort to actions that increase their profits, but harm society, such as: • • • •

Collusion Mergers which lessen competition Predatory behaviour Exclusionary behaviour 7

Will the market fix it all? Contestable market theory: does free entry eliminate all concerns about market power of incumbents? Persistence of dominance under free entry • Endogenous sunk costs industries: finiteness property • Network externalities (definition, direct and indirect, coordination effects, interoperability) and two-sided markets • Switching costs (definition, natural v. artificial, competitiveness of switching cost markets) • Predatory and exclusionary practices 8

Contestable markets Assume an incumbent I and a potential entrant E are equally efficient and produce homogenous goods. Cost of production is F+cq Baumol et al (1982): at equilibrium I will not set monopoly price, but price equal to average costs (AC: p=c+F/q). This is because if the price was higher than AC, profits would attract entry, in turn lowering prices (‘hit and run’ strategy).

9

Contestable markets: discussion The theory of contestable markets would have strong implications: if entry is free, we should not care about monopolists, as efficient outcome is reached. Critique: the theory hinges on two strong assumptions: • Unrealistic timing of the game (I cannot change price as E enters the market) • No fixed sunk costs of entry (hit-and-run strategy not profitable for E if some costs are non-recoverable) But the theory has the merit to stress the role of free entry in limiting market power: crucial insights for merger analysis. 10

A game theoretic version of the discussion Contestable market game. 2) Firms I and E simultaneously set prices 3) E (as efficient as I) decides whether to pay F or not. At the equilibrium, I sets pI=AC and E does not enter. Proof. At the last stage. If at stage 1) ACAC, then E would undercut and get all demand) 11

A game theoretic version , cont’d When fixed sunk costs exist. 2) E decides whether to pay F or not. 3) Active firms simultaneously set prices At the equilibrium, I sets pI=pM and E does not enter. Proof. At the last stage. If at stage 1) firm E entered, then standard Bertrand game: pI=pE=c; if at stage 1) firm E did not enter, firm I can set monopoly price pM. At stage 1), the only equilibrium is that firm E does not enter: if it did enter, it would get profits –F (zero Bertrand profits, but it has to pay the sunk costs). It if does not enter, it gets zero profits, which is better than negative. 12

Finiteness Property (endogenous sunk costs) Consider the following model (a very simplified version of Shaked-Sutton’s (1982) There exist n firms each with a product of quality uk (labelled so that u1>u2>…>un) and a price pk There exists a continuum of consumers with identical tastes but different incomes t. t is uniformly distributed with density S (S=size of the market) on a support [a,b], with a>0. Consumers buy one unit of the good (the market is covered), and have utility U(t,k)=uk (t-pk) 13

The game 1. Firms decide on entry (fixed cost ε>0) 2. They decide on quality of the good 3. They decide prices and sell (zero marginal costs) Proposition: If b<2a, only one firm will enter the industry at equilibrium (whatever S) (As income becomes less concentrated, more firms can enter; e.g., if 2a
Proof of the proposition We show that two firms cannot co-exist at equilibrium. Firms’ demand is derived by finding the consumer indifferent between the two qualities: From: u1 (t-p1)≥ u2 (t-p2), we obtain: u1 p1 − u2 p2 t ≥ t12 ( p1 , p2 , u1 , u2 ) = u1 − u2

All consumers with income t≥t12 will buy 1, all others will buy 2. Therefore: q1 = S (b − t12 ); q2 = S (t12 − a ).

15

Proof (cont’d) Profits can be written as:  u p −u p  u p −u p  Π1 = S  b − 1 1 2 2  p1 ; Π 2 = S  1 1 2 2 − a  p2 u1 − u2    u1 − u2 

By setting dΠi/dpi=0 we obtain the best reply functions: R1 : p1 =

b( u1 − u2 ) + u2 p2 a( u1 − u2 ) + 2u2 p2 ; R2 : p1 = ; 2u1 u1

Equilibrium prices are given by: p = * 1

( 2b − a )( u1 − u2 ) 3u1

;

p = * 2

( b − 2a )( u1 − u2 ) 3u2

Therefore, if b<2a there exists no equilibrium with positive p2, and firm 2 will not enter the industry (whatever S). 16

Equilibrium, when b>2a p1

R2

R1 p1E

E

0 p2

E

p2 17

No equilibrium, when b<2a p1

R2’(a’>a) R2

When a increases, 2’s best reply function shifts upwards and the equilibrium should involve a negative price of firm 2. R1

0

p2 18

Generalisation The finiteness property holds if the cost of producing a higher quality does not fall upon variable costs It holds across a number of different specifications (see e.g., Shaked-Sutton, 1987) Sutton (1991) puts the result to an empirical test. It shows that in advertising-intensive industries as S increases the industry does not become fragmented (when S increases, firms have incentive to increase Ad, which in turn raises fixed costs and limit the number of firms in the market). 19

Network effects: miscoordination Assume that consumers value a network good i as: Ui=vi (n)-pi, Where vi (n) is valuation if n consumers buy good i. vi (n) is non-decreasing and concave, with vi (1)=0 and vi (z) = vi (z+j) for any j>0 (all externalities exhausted at size z) There are an incumbent I and an “entrant” E, with cE
The game 1. Firms I and E set (uniform) prices, pI and pE 2. The z ‘new’ buyers decide btw. network I and E Assume the two networks are incompatible. This game admits two types of equilibria: • “Entry” equilibria, where buyers buy from the efficient “entrant” • Miscoordination equilibria, where the inefficient incumbent remains a monopolist

21

2nd stage: Buyers’ equilibria If pE≤pI≤v(z). There are two equilibria: 1) Miscoordination: all new consumers buy from I. Each buyer receives v(z)-pI≥0. If one deviated, he would get v(1)-pE≤0: no incentive to deviate. 2) “Entry”: all new consumers buy from E. Each buyer receives v(z)-pE≥. If one deviated, he would get v(z)-pI≤v(z)-pE. If pIv(z)-pE and disrupt the equilibrium.)

Entry equilibria There is an entry equilibrium where (pI,pE)=(cI,cI), and all z new consumers join E’s network whenever pE≤pI. Proof. A consumer would have no incentive to deviate. At (candidate) equilibrium, its surplus is v(z)-cI. By deviating and buying from I, it also gets v(z)-cI. Firm I has no incentive to deviate: zero profits also if it raises price (at the chosen continuation equilibria), negative profits if reduces it. Firm E neither: zero profits if it raises price, lower profits if it reduces it. 23

Miscoordination equilibria There is a miscoordination equilibrium where I sets monopoly price pI=v(z), and all z new consumers join I’s network even when pE
Welfare implications In this case (networks externalities do not increase above the critical threshold), the miscoordination equilibrium leads to lower welfare (productive inefficiency). However, in a model where network externalities are continuously increasing, welfare of old consumers matter as well. Entry may have an adverse effect, as old consumers are worse off relative to a situation where all buyers buy from the incumbent (limit case: old consumers are ‘stranded’). 25

Full interoperability If the two networks were perfectly compatible with each other, the entry equilibrium where (pI,pE)=(cI,cI), and all z new consumers join E’s network is the unique one. Proof. Suppose there exists an equilibrium where all new buyers buy from I at a price pI≥cI. Then firm E could set a price pE≤cI and it would attract all buyers (a deviant buyer would prefer to buy from E and get v(z)-pE>v(z)-pI – recall networks are fully compatible!). 26

Exclusion in network markets Incumbents can use their customer basis to exclude more efficient entrants. For instance: • By using price discrimination the incumbent can exclude more easily (Karlinger and Motta 2006) • Making a network incompatible with other networks, may lead consumers not to buy the latter • Since coordination of consumers play important role, incumbent may manipulate expectations so as to deter entry

27

Two-sided markets Consumers on a side of the market care about the number of buyers on the other side of the same platform, and (possibly) vice versa. Examples: Credit cards, discos, newspapers and TVs, shopping malls etc. Market failures are possible: below-cost price on one side to get one side on board, with the other side being attracted as well However, incumbent platforms may also benefit from miscoordination and ‘asymmetric pricing’ and exclude more efficient entrant platforms (Motta and Vasconcelos, 2008). 28

A model of two-sided markets A consumer on side i values a platform k as: Ui=vi (bjk+Njk)-pik, Where vi is externality from the other group j, bjk is the established base of k, Njk is new consumers. vi (n) is non-decreasing and concave, with vi (0)=0. There are an incumbent I and an “entrant” E, with cE
The game 1. Firms I and E set (uniform) prices, piI and piK 2. The N ‘new’ buyers decide btw. network I and E Assume the two platforms are incompatible. This game admits two types of equilibria: • “Entry” equilibria, where buyers buy from the efficient “entrant” • Miscoordination equilibria, where the inefficient incumbent remains a monopolist

30

2nd stage: Buyers’ equilibria If pE≤pI≤v(2N). There are two equilibria: 1) Miscoordination: all N side-i consumers buy from I. Each buyer receives v(2N)-pI≥0. If one deviated, one would get v(0)-pE≤0: no incentive to deviate. 2) “Entry”: all N side-I consumers buy from E. Each buyer receives v(N)-pE≥. If one deviated, one would get v(N)-pI≤v(N)-pE. If pIv(N)pE and disrupt the equilibrium.) 31

Entry equilibria There is an entry equilibrium where (pI,pE)=(cI,cI), and all N new consumers join E’s network whenever pE≤pI. Proof. A consumer would have no incentive to deviate. At (candidate) equilibrium, its surplus is v(N)-cI. By deviating and buying from I, it also gets v(N)-cI. Firm I has no incentive to deviate: zero profits also if it raises price (at the chosen continuation equilibria), negative profits if reduces it. Firm E neither: zero profits if it raises price, lower profits if it reduces it. 32

Miscoordination equilibria There is a miscoordination equilibrium where I sets monopoly price pI=v(2N), and all N new consumers join I’s network even when pE
Negative prices in 2-sided markets Suppose now there is a monopolist, and that the market does not exist yet (bIi=0). As before, Ui=vi (Njk)-pik. Under uniform pricing, there is a market failure eq’m. Proof. The lowest price I can make is p=cI. At candidate equilibrium, UI=0. By deviating and buying from I, given that nobody buys on side j, side i’s utility is Ui=vi (0)-cI<0. Under asymmetric prices, firm I can disrupt a market failure eq’m by setting, e.g., piI=-ε, pjI= vj (N). Side-i consumers would all buy, as Ui=vi (0)-(-ε)>0. Hence, side-j’s utility would be: Uj=vj (N)-pjI (=0). A side is ‘subsidised’ to have both sides ‘on board’. 34

Possible objectives of Comp. Law Economic Welfare (Total Surplus) Definition: W=CS+PS (+ …) If price falls, welfare rises Size of the pie, not how slices are distributed Dynamic aspects important (future W matters) Consumer Surplus CS v. W: lobbying arguments; who owns the firms?; If price equals marginal cost, who pays the fixed costs?; Who innovates and invests? Anyhow: usually, W and CS move together 35

Other possible objectives Defence of smaller firms Promoting market integration Protection of economic freedom Fighting inflation Fairness and equity Public policy considerations affecting competition Social, political, environmental reasons Strategic reasons (trade and industrial policies) 36

1b) Market power, competition, and welfare

1. Allocative efficiency 2. Productive efficiency 3. Dynamic efficiency 4. Public policies, and incentives to innovate

37

1. Allocative efficiency Definition of market power: the ability of a firm to profitably raise price above marginal costs A matter of degree, not of existence The deadweight loss (see Figure 2.1) Inverse relationship between market power and welfare An additional loss of monopoly: rent-seeking activities (see Figure 2.2)

38

Figure 2.1. Welfare loss from monopoly p O

R

pm

pc

T

qm

S

qc

c

O′

q 39

2. Productive efficiency Additional welfare loss if monopolist has higher costs (see Figure 2.3) “Quiet life” and managerial slack Principal-agent models: market competition helps, but too fierce competition may decrease efficiency Nickell et al.: individual firms’ productivity higher in competitive industries Darwinian arguments: competition selects more efficient firms Olley-Pakes, Disney et al.: industry productivity mostly increases through entry/exit 40

Figure 2.3. Additional loss from productive inefficiency p O R′

p′m pm

pc′ pc

R

W

V Z T′ T qm′ qm

qc′

c′ S

c O′

qc

q 41

Productive efficiency, II Number of firms and welfare: trade-off between allocative and productive efficiency As number of firms increases, market power decreases, but also welfare Important: defending competition, not competitors! (else, inefficiencies, and fixed cost duplications)

42

3. Dynamic efficiency U-shaped relationship between market power and welfare: trade-off between appropriability and competition in R&D investment Lower incentives to innovate of a monopolist: innovation introduced if additional profits higher than costs Appropriability matters: no (little) innovations if no patent protection, compulsory licensing etc...

43

4. Public policies and incentives to innovate Ex ante (incentives) v. ex post (diffusion): IPR protection guarantees market power Essential facilities (EF) doctrine Necessary, non-reproducible inputs Ex.: airport slots, port installations, local loop… EC accept EF doctrine, but ECJ: Bronner case Important to preserve incentives to innovate! Apply EF doctrine only when owner has not invested to create the facility 44

1c) Market power and market definition Market definition is only an intermediary objective Modern econometric tools allow in some cases to assess market power directly: if so, defining the market is not needed; if not, two steps in the analysis: i. Market definition (product and geographic) ii. Market power assessment 45

i. Market definition Product Market: the set of products “substitutes enough” to each other (in the sense of exercising a competitive constraint on each other, not of ‘resembling’ each other). a) Demand sustitutability and supply substitutability(*) b) How to find this set? SSNIP (Small Significant Non-transitory Increase in Prices) test: could a hypothetical monopolist selling products X profitably raise prices by 5-10%? If yes, X is the product market If no, apply SSNIP test again with X and Y, etc... (*) Entry may also be considered at the market power stage of investigation if not as likely and easy.

46

Market definition, II How to implement this ‘thought experiment’? Own-price elasticities, cross-price elasticities, price correlation tests NO: Price differences and product characteristics A problem with this test: raise prices relative to what? Merger cases: current prices Art.82 cases: competitive price? (more difficult to check; also, cellophane fallacy – from Du Pont case where it was find that high cross elasticity of demand was compatible with cellophane and other flexible wrapping materials to be in the same market) Geographic definition: SSNIP test, transport costs 47

Market definition, III Temporal markets: restaurants and bars may be in the same market for lunch meals, not for dinner meals Seasonal markets: oranges and bananas may be in the same market only for some months of the year Multiple markets: markets may be defined differently according to the starting point (e.g., trains, buses, planes) After-markets: secondary products (e.g. spare parts) may be in the same market as primary ones depending on: 1. price of secondary product relative to primary; 2. probability of replacement; 3. sophisticated buyers. (Examples: Kodak, Kyocera, Hugin) Consistency of market definitions over time. 48

ii. Assessment of market power The traditional approach Central role of market shares Which thresholds for market shares? Measurement and relative strengths (reserves, capacities, persistence of shares) Ease and likelihood of entry Buyers’ power Econometric techniques Estimation of residual demand elasticity Logit models 49

2. Cartels and horizontal agreements a. Cartels a. Economics of collusion b. Practice c. Woodpulp

b. Horizontal agreements a. Joint-ventures b. Cross-licensing c. Patent pools

50

2a. Cartels Antitrust laws prohibit agreements aimed at fixing prices, sharing markets etc. Rationale: such agreements allow firms to exercise market power they would not have otherwise Collusion may take different forms, and laws may differ as to what is collusion and what is evidence required to prove it Plan: what is collusion, what helps firms to sustain it, what actions can be taken to fight it 51

Collusion, I: “enforcing collusion” Collusion refers to a situation where firms set prices which are higher than some competitive benchmark (or prices close enough to monopoly situation) For economists, collusion is an outcome What ingredients are necessary to enforce collusion? Timely detection of deviations from collusive actions Credible mechanism for the punishment of deviations Threat of punishment prevents firms from deviating Examples (explicit agreement not necessary) 52

Collusion, II: Coordination Which collusive price? The problem of coordination Tacit collusion: costly experimentation to coordinate on a collusive outcome, risk of triggering price wars Explicit collusion: firms coordinate on collusive outcome and avoid problems due to shock adjustments (Market sharing schemes: possible to adjust to cost and demand shocks without triggering price wars) Firms will try to talk in order to coordinate! 53

Collusion, III: formalisation Firms meet in the marketplace for an infinite number of periods (or with uncertain final date) Collusion if following Incentive Compatibility Constraint (IC) holds for each firm: Πc+ δVc > Πd+ δVp Or: δ > (Πd- Πc)/(Vc-Vp) (RHS is called critical discount factor) Formalisation: see SEPARATE SET OF SLIDES. 54

Factors that facilitate collusion I. Structural factors Concentration Entry Cross-ownership and other links with competitors Regularity and frequency of orders Buyers’ power Evolution of demand Symmetry Multi-market contacts 55

Facilitating factors, II II. Price transparency and exchange of information Observability of firms’ actions facilitate enforcement – Green and Porter’s model – Exchange of information on past/current data

Coordination and the role of communication – Focal points – Exchange of information on future prices and outputs (private v. public announcements) Examples: ATP and collusion in auctions 56

Facilitating factors, III III. Pricing rules and contracts Most-Favoured Nation clause (ambiguous effect) Meeting-Competition clauses (helps collusion, by eliciting information on rivals and discouraging deviations in the first place) Resale price maintenance (enhances cartel stability by eliminating variation in retail prices) Uniform delivered pricing; basing point pricing 57

PRACTICE: How to detect and fight collusion? For economists, collusion as an outcome Both tacit and explicit agreements may sustain collusion So, why not inferring collusion from market data? Inferring collusion from data. Problems, I: price levels Price data availability (list v. effective prices) Difficult to estimate ‘monopoly price’ and marginal costs Where to set the threshold level? A dangerous principle: firms guilty because able to set a high price…(market power not a problem per se) 58

Standards of proof, II: data Inferring collusion from data. Problems, II: evolution of prices Price parallelism: not a proof of collusion (common shocks) Which legal certainty if firms are found guilty for independent business practices? ‘Parallelism plus’ not convincing either, unless there is proof of coordination on facilitating factors (eg., RPM, info exchange) Periods of ‘price wars’ not sufficient condition for collusion either (new capacity, new competitors, demand shocks…)

Conclusion Econometric tests as complementary evidence, not proof of collusion (results sensitive to different techniques used) 59

Standards of proof, III: hard evidence Hard evidence only (of communication on prices and/or coordination on facilitating practices) as proof (focus on observable elements verifiable in courts, to preserve legal certainty: fax, e-mail, phone calls, video etc.)

Too lenient with the firms? (Since collusion can be reached tacitly, focusing on ‘hard evidence’ amounts to permitting collusion?) Not necessarily: firms will try to coordinate to avoid costly market experimentation and will leave ‘traces’ More active policies can be used, ex ante and ex post 60

Ex ante policies to fight collusion Black list of facilitating practices might deter collusion and free resources for cartel detection – Private announcements of future prices/outputs – Exchange of disaggregate current/past information – Meeting competition, RPM and other clauses, if adopted by coordination – Cross-ownership among competitors not to be allowed – Merger control (joint dominance)

Deterrence of collusion - more severe fines; administrative penalties and directors’ disqualification, criminal sanctions - higher probability of discovery 61

Ex post policies to fight collusion Surprise inspections Leniency programmes The US and EU experience: Leniency must be clear and certain (not discretionary) Leniency should be extended to firms that report after an investigation has started

62

A Case: Wood pulp (ECJ, 1993) 1984: the EC finds that 40 wood pulp producers and 3 trade associations infringed article 85 of the Treaty (now art. 81) by concerting on prices 1993: the ECJ annulled a large part of the EC Decision (both on procedural and substantial grounds) Three accusations made by the EC: 1) Price concertation with KEA (a US export cartel) 2) Price concertation within Fides (a trade-association) 3) Parallel behaviour. 1) and 2) concern a subset of firms/years and are supported by clear hard evidence; 3) more interesting 63

Parallel behaviour in Wood pulp Parallel behaviour consisted of: (i) a system of quarterly price announcements (ii) (quasi-) simultaneity of price announcements (iii) the fact that announced prices were identical EC: this parallel behaviour is proof of collusion ECJ: “parallel conduct cannot be regarded as furnishing proof of concertation unless concertation constitutes the only plausible explanation for such conduct” Also: “[article 81] does not deprive economic operators of the right to adapt themselves intelligently to the existing and anticipated conduct of their competitors” 64

Analysis of parallel behaviour, I ECJ commissions an experts’ study of the industry, and finds that parallel behaviour in this case might be due to normal oligopolistic interdependence among firms: (i) System of quarterly price announcements ECJ finds that the practice of announcing quarterly prices beforehand had been requested by buyers, who wanted to estimate their costs (wood pulp accounts for 50-75% of costs of a paper manufacturer) and fix the prices of their downstream products. (All prices quoted in US dollars also welcome by buyers; not necessarily a practice to enhance collusion.) 65

Analysis of parallel behaviour, II (ii) Near-simultaneity of price announcements. According to the EC, it can only be explained by concertation. ECJ finds that there is a possible innocent explanation for immediate diffusion of price information: 1. Each buyer is in contact with several suppliers (for diversification) and may have incentive to reveal prices 2. Most wood pulp producers are also paper manufacturers, and (again because of diversification) purchase some of their inputs from upstream rivals 3. There are common agents who work for several wood pulp producers 66

Analysis of parallel behaviour, III (iii) Parallelism of announced prices. EC: very similar prices are proof of concertation; very high prices in some years and low prices in others (punishment phase) also indicate collusion ECJ: 1. average price evolution compatible with evolution of world demand and supply (e.g., storage subsidy schemes in Sweden, excess capacity in North America) 2. Similarity of prices across producers is compatible with oligopolistic behaviour (ECJ refers to kinkeddemand curve story; but tacit collusion may also explain parallelism) 67

2b. Other horizontal agreements: JVs Joint-Ventures: as for mergers, trade-off between market power and efficiency A special case: Research Joint-Ventures Because of spillovers and non-rivalry, R&D unlikely to attain socially optimal levels RJV may promote R&D by sharing costs and avoiding duplications, but: R&D may fall absent competition, and… collusion may extend to marketing and production Only RJV on basic research should be allowed 68

Co-operative R&D (d’Aspremont-Jacquemin ‘88) Homogeneous goods, demand: p=a-Q. Marginal costs: ci=C-xi-sxi, with i=1,2. with xi=investment, 0≤s≤1 is spillover rate. R&D costs gxi2/2, with g>4/3 (for stability and SOC). Game: 6. Firms choose R&D levels, xi, with i=1,2. 7. Firms choose output levels, qi, with i=1,2. – –

Version (a) R&D set non-cooperatively; Version (b) R&D set cooperatively (R&D J.V.)

Firms always compete downstream. 69

Version (a): competition in R&D Last stage first. Each firm’s problem is: Maxqi Πi(qi,qj)=[a-qi-qj-ci(xi, xj)]qi, given xi,xj, for i,j=1,2. From FOCs, we obtain the Cournot outputs as: qic=[a-C+xi (2-s)+xj(2s-1)]/3, and by substitution: Π i =[a-C+xi (2-s)+xj(2s-1)]2/9. First stage. Each firm’s problem is Maxxi Πi (xi,xj)- gxi2/2. From FOCs, and imposing symmetry: xc=[2(a-C)(2-s)]/[9g-4-2s+2s2], and: qc= =[3(a-C)g]/[9g-4-2s+2s2]. 70

Cooperation in R&D (version b); comparisons Last stage is the same as before. First stage. Firms Maxxi,xj Πi (xi,xj)+ Πj (xi,xj)-gxi2/2-xj2/2. From FOC: xJV=[2(a-C)(1+s)]/[9g-2(1+s)2], and: qJV =[3(a-C)g]/[9g-2(1+s)2]. Comparisons: xJV > xC, qJV > qC, πJV > π C, WJV > WC, iff s>1/2. If s large enough, firms know they appropriate little of their investments in R&D, resulting in lower R&D. Under R&D cooperation, the spillover is internalised. 71

Cross-licensing Cross-licensing: when two firms allow each other to use their technology. When technologies are complementary, crosslicensing may be indispensable. Suppose that two firms have ‘blocking’ (i.e., essential) patents. Then, production or new innovation requires both patents. Also, an agreement whereby two rivals set cooperatively the licensing fees for complementary technologies allows internalisation of the Cournot effect (see below) and results in lower fees. 72

Patent pooling When patents are complementary, better to have a single owner of all patents (“Cournot effect”: better a multiproduct monopolist than two independent monopolists when products are complementary). Patent pool: firm or organisation which owns the patent rights and licenses them to third parties as a package. If patents are complementary (a fortiori if essential), this will keep royalties down (best if two-part pricing). Patent pooling may also save on transaction costs (rather than having to negotiate with multiple parties, a firm has to deal with one party only). 73

A simple model of patent pooling Suppose n manufacturers face a market demand Q=1-p and need two complementary technologies owned by firm A and B, who separately license them at royalty rate wA and wB. (After, we look at patent pooling.) 2. Patent holders simultaneously set wA and wB. 3. The n firms compete in prices with cost c+wA+wB. 2nd stage. Bertrand comp.: p= c+wA+wB. 1st stage. Patent holders maxwi Πi=wi(1-c-wA-wB). From FOCs and symmetry: w*=(1-c)/3; p*=(2+c)/3; Π*= (1-c)2/9. 74

Patent pool 1. Patent-holders set wi, wj to maximise joint profits. 2. As before. 2nd stage. Bertrand comp.: p= c+wA+wB. 1st stage. Patent holders maxwi,wj Πi=(wi+wi)(1-c-wiwj). From FOCs and symmetry: wp=(1-c)/4; pp=(1+c)/2; Πp= (1-c)2/8. Clearly, patent poling Pareto-dominates the situation where licenses are set independently: royalty fees and prices are lower, profits are higher. 75

3. Horizontal Mergers a. Single-firm dominance (unilateral effects) Merger increases market power Efficiency gains b. Joint dominance (pro-collusive - or coordinated - effects) c. How to proceed: A "check-list” d. EU Merger Regulation e. Merger remedies in the EU f. A case: Airtours/First Choice 76

Why modelling mergers is difficult Brief explanation of what happens with homogenous goods and cournot competition (salant, switzer, reynolds, QJE 1981) – see separate slides Need of asset-based model • Product differentiation (Motta, 2004) • Homogenous goods with capacity constraints (Perry-Porter, 1985)

77

Horizontal mergers: unilateral effects (One-shot Nash equilibrium before and after the merger.) If there are no efficiency gains, merging firms increase prices:

→

consumer and total surplus decrease.

Intuitions. 78

Unilateral effects: A model* n  n 2 γ  n 2  U = v ∑ q1 −  ∑ qi +  ∑ qi   + y 2(1 + γ ) i =1 n  i =1   i =1 n

γ ∈ [ 0, ∞ )

Whence:

(1)

is degree of substitution.  n 1   pi = v − nq + γ q ∑  i j  1 + γ  j =1 

(2)

By inversion, direct demand functions: 1 γ qi = v − pi (1 + γ ) + n n

p j  j =1  n

∑

(3)

79

Properties of demand function - Both price and quantity competition can be studied - Aggregate demand does not depend on γ and n. Firms have identical technologies: with c
C ( qi ) = cqi ,

80

The effects of a merger 1. Equilibrium with all single - product firms 2. Equilibrium with a multi - product firm with m products. Lemma 1 The merger increases prices and decreases consumer surplus. Lemma 2 A merger always benefits the merging firms. The result holds unless one assumes (i) quantity competition, (ii) homogenous goods and (iii) no efficiency gains. 81

RO

p1

RI′ M

pI

pB

RI

B

pB

pO

p3

Figure 5.1. Effects of a merger absent efficiency gains: Strategic complements 82

q1 RO

qc

C

qI

π 1 M π 1′ qc qO

RI′

RI q3

Figure 5.2. Effects of a merger absent efficiency gains: Strategic substitutes 83

Lemma 3 The merger increases outsiders' profits. This result does not depend on whether firms compete on prices or quantities. Lemma 4 The merger increases producer surplus. Lemma 5 The merger reduces net welfare. 84

Efficiency gains If savings from the merger are large enough, they will outweigh the increase in market power and result in lower prices. Assessment of efficiency gains. • Distinction between cost savings that will affect variable production costs (and prices), and cost savings that affect fixed costs. •Efficiencies from technical rationalisation are easier to demonstrate. •Efficiencies should be merger-specific. • Sinergies, not mere reallocation of production •Independent studies to try and evaluate efficiency considerations. 85

Efficiency gains from mergers* A merger creates a larger firm. Possible cost savings: the merged entity has unit cost ec, with e ≤ 1. The lower e, the higher the efficiency gains. Lemma 6 The merger is beneficial to consumers if and only if it involves enough efficiency gains, i.e. if and only if: e ≤ e .

86

Synergies v. reallocation of output Farrell-Shapiro: mere reallocation of production not enough to increase consumer surplus. Example. Simple Cournot model with two firms, having cost c1=0 and c2<1/2. Demand p=1-Q. Absent merger: q1=(1+c2)/3; q2=(1-2c2)/3;p=(1+c2)/3; PS=(2-c2+5c22)/9; W=(8-8c2+11c22)/18. Merger: qm=1/2; pm=1/2; PSm=1/4; Wm=3/8. The merger always raises prices (pm>p for c2<1/2), and thus decreases consumer surplus. However, it increases W (due to the increase in profits) if c2>5/22. Note: here consumer welfare standard may lead to different conclusions than a welfare standard. 87

RO

p1

RI′ (e = 1) M

pI

RI

RIe B

pB E

pB

pO

p3

Figure 5.3. Effects of a merger with efficiency gains 88

Lemma 7 A merger always benefits the merging firms. Lemma 8 The merger increases outsiders' profits if efficiency gains are small enough, i.e. if e>e:

Only if there are important efficiency gains will the outsiders lose from the merger. Lemma 9 The merger always increases producer surplus. Lemma 10 The merger improves net welfare if it involves enough efficiency gains, i.e. if e ≤ e . 89

Horizontal mergers: pro-collusive effects The merger might create the structural conditions for the firms to (tacitly or explicitly) collude. Two main reasons. Reduced number of firms. More symmetric distribution of assets.

90

How to proceed in horizontal mergers: a check-list 3. Unilateral effects Market definition: Product and geographic market Market power: Market shares and distribution of capacities; demand elasticities; elasticity of supply of rivals; potential entrants; switching costs; buyer power… If possible, econometric analysis. Efficiency gains 91

Two possible outcomes: 1. The merger enables firms to significantly raise

→

prices beyond the current level. Prohibition or remedies.

2. Might collusion arise after the merger?

92

Joint dominance Number of firms and concentration Distribution of market shares and capacities Potential entrants (and switching costs) Buyers' power Observability of other firms' behaviour (exchange of information, competition clauses, resale price maintenance) Frequency of market transactions and magnitude of orders…. 93

EU Merger Policy Preventive authorisation system (originally MTF, but recent re-organisation) One-stop shop for mergers (subsidiarity principle) Reasonably quick and effective, with certain time horizon

94

The EU Merger Regulation 4064/89 was source of inefficient biases. 1) Restricting attention to mergers which create dominance implies that some welfare detrimental mergers might be approved. (Joint dominance to cover unilateral effects: not a good approach. Airtours judgment.) 2) Failure to consider efficiency gains might result in beneficial mergers being blocked by the EU authorities. 95

New Merger Regulation Compromise between “dominance” and “SLC” test. It prohibits mergers that “would significantly impede effective competition, in the common market or in a substantial part of it, in particular as a result of the creation or strengthening of a dominant position”. Merger guidelines clarify DG-COMP’s approach to mergers. (They also include an efficiency defence.)

96

Merger remedies in the EU Merger remedies increasingly important in the EU and US Structural remedies : they include divestiture of an entire ongoing business or partial divestiture (possibly a mix and match of assets of the different firms involved. Non-structural remedies : engagements not to abuse of certain assets available to them, including compulsory licensing or access to property rights.

97

Airtours/First Choice EC prohibited the merger in 2000. CFI annulled decision in 2002. First merger where joint dominance applied to more than two oligopolists. This case sheds light on the tension between the dominance test and economic analysis of mergers (discussions following it contributed to introduce the new Merger Regulation) First, main arguments in the EC Decision, Then, main arguments in the CFI Judgment 98

UK short-haul package holiday market MMC review, 1997: Rapid changes in the sector; low barriers to entry; market shares volatility: 1992: Thomson 24% ; Airtours 11%; First Choice 6%; Thomas Cook 4%. 1998: Thomson 30.7%; Airtours 19.4%; First Choice 15%; Thomas Cook 20.4%; Cosmos/Avro 2.9%

EC: after 1997, M&As: four large firms vertically integrated (from charter airlines to retail distribution), and many small ones (not vertically integrated). EC: after V.I. process, no longer mobility and openness. Purchase of package tours via Internet not significant in 2000. 99

Main characteristics of the industry Firms' decisions at two stages. Planning stage: overall capacity (seats on charter flights, rooms in hotels) is set for the following 1218 months. Selling season: firms compete under a capacity constraint, and have a strong incentive to fill capacity: a given package loses all value after departure date.

100

Analysis of joint dominance, I Standard collusive arguments could not be sustained: Temptation to deviate from collusive prices would be strong: discounts on catalogue prices as departure dates approach, to fill capacity. Threat of punishment within selling period has little credibility due to the capacity constraint. Also, package holidays are heterogeneous goods: packages differ in terms of destination, type of hotel, additional services...: difficult to reach agreement on collusive prices and adjust to shocks. 101

Analysis of joint dominance, II EC: firms' profitability determined by overall level of capacity (number of holidays) Collusion not on high prices (selling season), but on low capacity (planning season). (Deviation: high capacity in planning season; punishment: high capacity for one/more periods.) In general, such collusion is unlikely: capacity decisions constrain firms for a long time: punishments are very costly and much delayed. This industry: capacity reviewed periodically, collusion on capacities not a priori impossible 102

Analysis of joint dominance, III Staiger and Wolak (1992). Repeated game, each period of three stages. 1. firms choose capacity without knowing demand; 2. demand level disclosed; 3. firms choose market prices. The model fits the package tour industry very well. Collusive equilibria where firms collude by restricting capacity levels but then set prices which are not the fully collusive ones. Lower prices (or even price wars) in selling season are compatible with collusion on lower capacities in planning period. 103

Analysis of joint dominance: other factors (+) High concentration (after merger only three majors) (+?) EC: transparency in planning period (few charter operators, large tour operators trade seat capacity, purchase of planes unlikely to be kept hidden) (-) High demand volatility: difficult to distinguish btw. deviations and genuine mistakes in predicting demand. Collusion more difficult (EC disagrees) (?) Supply substitution and barriers to entry: hot debate (+) Symmetry among large operators after merger Conclusion: Tacit collusion on capacities conceivable, but is it sufficiently likely? 104

Economic analysis v. dominance test Economic analysis: will prices rise after the merger? High concentration; small firms or potential entrants unlikely to fully discipline the major operators. Efficiency gains unlikely to outweigh market power: “...the merger is only expected to lead to overall synergies of less than 1% of the overall costs of the combined entity. Furthermore, the cost savings mostly relate to overhead and other fixed costs.”

Under a SMP test: prohibition decision easier to defend. Under dominance test: controversial, but EC had no choice but use joint dominance. 105

The CFI Judgment For tacit coordination sustainable, three conditions: 1. Sufficient market transparency (to monitor each other and see whether there are deviations) 2. There must be an incentive not to depart from the common policy, i.e., there must be a credible mechanism of retaliation if deviations occur 3. Current and prospective rivals, as well as consumers, must not jeopardise coordination Therefore, concept of joint dominance used by the CFI is same as in economic analysis 106

CFI Judgment: Analysis of the industry M&As did not change industry structure (Th. Cook’s rise only important change since 1997) MMC Report still valid: high volatility of market shares not conducive to collusion High demand growth (not low, as claimed by EC): collusion less likely High demand volatility: collusion less likely Market not transparent: overall capacity difficult to observe, as made of hundreds of separate decisions (routes, destinations, hotels) often going in different directions. 107

CFI’s analysis of the industry, II Even if total capacity levels were meaningful, difficult to monitor each other’s capacity decisions: 1. Hotels unlikely to be means of monitoring (they prefer tour operators from different countries) 2. Purchase of airline seats by tour operators minor and comes at late stage of planning period 3. Decisions about investments to increase capacities are observed with delays Therefore, transparency is low during planning period 108

CFI’s analysis of the industry, III No credible punishment mechanism (also: EC decision confused about importance of retaliation) Increasing capacity in selling season not a deterrent: 1. Innate tendency to caution in capacity decisions 2. Since deviations are not detected timely, reactions would take time 3. Late-added package holidays would be of poor quality (inconvenient flight times, poor-quality accommodation) Increasing capacity following season is poor retaliatory measure (demand unpredictable) 109

CFI’s analysis of the industry, IV Underestimated reaction of competitors and consumers 1. Small tour operators would increase capacity if prices rise: adequate access to airline seats (charters, scheduled, low-cost) and to distribution, both internet and traditional (EC said large operators would foreclose; 40% in any case independent agencies) 2. Sizeable European operators would enter UK 3. Consumers compare before buying. Long-haul foreign package holidays increasingly attractive  EC Decision vitiated by a series of errors of assessment of joint dominance 110

4. Vertical restraints and vertical mergers Vertical agreements (or restraints): clauses to control for the externalities arising between firms operating at successive stages of an industry Plan a. Different types of vertical restraints b. Intra-brand competition: double marginalisation and freerider problem c. Inter-brand competition d. Exclusive dealing e. Policy implications f. Vertical mergers 111

4a. Types of vertical restraints Different vertical restraints are used (according to observability, absence of arbitrage etc.): Franchise fee (FF) contracts Resale price maintenance (RPM); Quantity fixing Exclusivity clauses: exclusive territories (ET), exclusive dealing (ED), selective distribution Tying Royalties 112

4b. Intra-brand competition The problem of double marginalisation: vertical integration (VI), RPM (ceiling), and FF: efficient; ET: inefficient.

Upstream firm (manufacturer)

Downstream firm (retailer)

Consumers

113

Double marginalisation U has to sell to D, who sells to consumers: q=a-p. U has marginal cost c;ΠUs+ ΠDs. 114

Intra-brand competition, II Interaction among retailers may create negative externalities

Upstream firm (manufacturer)

Downstream firm (retailer)

Downstream firm (retailer)

Consumers 115

Intra-brand competition, III Free-riding among retailers and under-provision of services: VI, ET and RPM (floor): efficient (N.B.: Substitutability of vertical instruments) Quality certification: RPM and selective distribution Free-riding among producers: exclusivity clauses Exclusivity may also remove opportunistic behaviour and promote specific investments 116

Intra-brand competition, IV Possible anti-competitive effects of vertical restraints: the commitment problem of an input monopolist. Example of commitment problem: franchise Solution of the monopolist’s problem: Vertical integration Exclusive territories Resale price maintenance Most-Favoured Nation clauses and antidiscrimination laws (transparency is bad) 117

4c. Inter-brand competition

Illustration of inter-brand competition

Upstream firm (manufacturer)

Upstream firm (manufacturer)

Downstream firm (retailer)

Downstream firm (retailer)

Consumers

118

Strategic use of vertical restraints Upstream firms may use vertical restraints so as to make their own retailers more or less aggressive (depending on the nature of competition in the marketplace). For instance, assume that market competition is such that when a retailer increases prices, the best response of its rival retailer is also to raise prices. Then, it is optimal for the manufacturer to use a two-part tariff contract and sell to the retailer at a higher price than it would be optimal absent strategic considerations. The retailer would behave less aggressively, the rival retailer’s price also increase, and market profits will go up for both (by using the fixed fee these profits are appropriated by the manufacturer). 119

Strategic effects of VR: intuitions R1 ( w1 = c) R1′( w1′ > c)

p2

R2′ ( w2′ > c ) E′

R2 ( w2 = c )

E

p1 120

Strategic use of vertical restraints Two upstream firms U1,U2 sell differentiated goods. Demand is given by: qi = (1 / 2)[ v − pi (1 + γ / 2 ) + (γ / 2) p j ]

Each upstream firm needs retailer (resp. R1,R2) to sell the good Zero production and retail cost, for simplicity We show that vertical restraints (in the form of delegation with two-part tariffs) can be used to increase profits

121

Integration v. delegation Vertical integration. If R1, R2 are owned by U1, U2, one can find equilibrium by solving: max π = p q ( p , p ) p i i i i j i From FOCs one obtains:

2v ( 2 + γ )v 2 pVI = ; π VI = . 2 4 +γ (4 + γ ) 122

VR: Two-part tariffs 1st stage: Ui sets Fi+wiqi for Ri. Contracts are observable. 2nd stage: Ri chooses pi. We solve by backward induction. Last stage: each Ri maxpiπiR=(pi-wi)qi (pi,pi). Whence, pi*(wi,wj), qi*(wi,wj). First stage: each Ui earns Fi+wiqi . Therefore, Ui sets Fi so as to appropriate Ri’s profit: maxwiπiU=(pi*-wi)qi*+wiqi*. At equilibrium: wi*>0 and: 2 2 4 ( 2 + γ ) v 2 ( 2 + γ )( 8 + 8 γ + γ ) v VI FF VI p FF = > p ; π = > π 16 + 12γ + γ 2 (16 + 12γ + γ 2 ) 2 123

Strategic effects of VR: intuitions R1 ( w1 = c) R1′( w1′ > c)

p2

R2′ ( w2′ > c ) E′

R2 ( w2 = c )

E

p1 124

Exclusive territories Rey and Stiglitz (1988): exclusive territories allow manufacturers to relax competition. Suppose each (differentiated) Ui has two or more retailers perceived as homogenous by consumers. Intra-brand competition: pi=wi, and solution as if Ui are vertically integrated. Suppose now each retailer is given an ET. Then in each territory, the game is as the one above, and prices will be higher.

125

Vertical restraints might facilitate collusion An example: Resale price maintenance. Consider possible collusion among manufacturers. Assume that manufacturers need to sell through local retailers. Absent RPM, when local demand shocks occur, difficult to see if retail prices differ before of deviations by manufacturers or demand shocks. RPM increases observability of deviations and facilitate collusion upstream. 126

4d. Exclusive dealing Outline: Definition, and motivation Brief historical perspective Recent contributions on exclusionary effects Pro-competitive effects Policy implications Advanced session: Anti-competitive models of exclusive dealing

127

Exclusive dealing: contracts that require to purchase products or services for a period of time exclusively from one supplier. Efficiency gains

Anti-competitive - stimulate investments effects into retailers’ services (free riding problem). to

- allow a dominant firm deter efficient entry.

- stimulate specific vestments (opportunistic haviour) LONG SERIES OF HIGH PROFILE CASES. E.g.: United States v. Microsoft (1995). Mars v. Langnese/Schöller - art. 81 case (1996) US v. Dentsply International (2005) 128

Traditional argument

I E

Foreclosure of a crucial input (ex. distribution network) I: incumbent B: unique buyer E: potential entrant

B

129

“Chicago school” critique (Posner 1976, Bork 1978)

Why does the buyer sign the exclusive deal?

cE
cI

A B=π m C

πm < CS(cI)-CS(pm) D

=x*

Incumbent’s

buyer’s loss

gain

The incumbent cannot profitably use exclusive contracts to deter entry. Efficiency considerations explain the use of exclusive contracts. 130

Why exclusive dealing can be anticompetitive 1. Upstream competition matters: if the buyer expects weak competition, they will accept ED for very small compensation 2. If there are several buyers and scale economies (i.e. the entrant needs a minimum number of orders), then buyers’s miscoordination may arise. 3. With several buyers (or markets), depriving the entrant of some buyers (market) obliges all to buy from incumbent, and hence deters entry 4. Exclusive dealing as a rent-extraction mechanism (advanced, but lower practical relevance) 131

1. Upstream competition matters In the Chicago school’s theory, the buyer expects fierce competition if the entrant will actually enter. Therefore, he will require a large compensation (x*=CS(cI)-CS(pm) ), that incumbent cannot profitably offer. If instead there would be weak competition if entry took place, the buyer would require lower compensation, which the incumbent may be able to offer. In the limit, if entrant and incumbent colluded (and had siimilar efficiency), then he would accept ED even at x*=CS(pm)-CS(pm)=0. 132

2. Buyers’ miscoordination Suppose there are several uncoordinated buyers, and that the entrant needs a certain number of them for entry to be profitable (scale economies on production or demand side) It is possible that exclusive contracts are accepted by all buyers (even behind a very small compensation) simply because each buyer expects all the others to do so (and if one rejected exclusivity, it anticipates it would not be sufficient to trigger entry) 133

3. Discriminatory offers facilitate exclusion

pm

cI

A

pm

B=π m C D Buyer 1

cI

A B=π m C D Buyer 2

If 2 πm >∆CS (= B+C), then the incumbent persuades one buyer, and thus excludes the entrant for both buyers. 134

Sequential offers Stage 1: Incumbent offers ED to buyer 1 (including compensation x1); no price commitment. Buyer 1 decides whether to accept or reject.

Stage 2: Incumbent offers ED to buyer 2 (including compensation x2); no price commitment. Buyer 2 decides whether to accept or reject.

Stage 3: Entry decision. Stage 4: Active firms name prices. Assumption: 2πm>x* 135

Sequential offers B1

accept

reject

B2 accept E In π I = 2π − x1 − x2 m

π E = −F

B2 reject

π I = 2π − x1 − x2 πE = 0 m

In

E

E

E Ou t

reject

accept

Ou t

In

π I = π m − x2

Ou t

In

π I = 2π m − x2

πE=0 π E = (c I − c E ) q (c I ) − F < 0

πI=0

Ou t

π I = 2π m πE=0

π E = 2(cI − cE )q(cI ) − F > 0

• If at least a buyer signed: Entry does not follow 136

Sequential offers I offers x1=

0

accept I offers x2=

0

B2 reject

accept E

E Ou t

reject

B2

accept E

B1

π I = 2π m − x1 − x2

Ou t

π I = 2π m − x2

Ou π I = 2π mt− x2

π bi = CS ( p m ) + xi

π b1 = CS ( p m ) + x1

π b1 = CS ( p m )

π b2 = CS ( p m )

I offers x2=x*

reject E In πI = 0 π bi = CS (cI )

π b2 = CS ( p m ) + x2

137

Sequential offers Buyer 2’s decision: If B1 accepted, • B2 accepts behind x2>0 (entry does not follow in any case) • The incumbent offers x2=0 and B2 accepts.

If B1 rejected, • B2 accepts behind x2 =x*. • The incumbent offers x2 =x* (2πm –x*>0) and B2 accepts.

Buyer 1’s decision: Whatever it chooses, entry does not follow: • It accepts behind x1>0. • The incumbent offers x1=0 and B1 accepts.

In equilibrium both buyers accept behind negligible compensations: profitable exclusion!!!!

138

Scale economies and exclusion More generally, the same mechanism will take place whenever the rival of a dominant firm will need a certain size of the market to be profitable. The dominant firm may take actions allowing it to ‘capture’ the number of buyers (or markets) enough to make the rival unprofitable. Price discrimination, rebates, (tying) are other tools which achieve exclusion in similar way as exclusive dealing (see more below).

139

Anti-competitive exclusive dealing: summary Exclusive deals might indeed be used to deter entry Hypothesis of buyer fragmentation and scale economies most convincing (but check market features in actual cases) Central Purchasing Agencies might help? Discriminatory offers help exclude Of course: - Dominance crucial - Proportion of buyers covered by ED? - Length of the ED important (but also for procompetitive rationale) 140

Pro-competitive effects of exclusive dealing Exclusivity clauses may enhance investments. 1) Avoid free-riding on manufacturers’ investments to promote retailers’ sales (Besanko and Perry, 1993). Suppose manufacturers can invest to help retailers’ sales. If retailers carry several brands, low investment because of spillovers. Exclusivity raises investments in retailers’ services, promotion, training etc.

141

Pro-competitive effects of exclusive dealing, II 2) Segal and Whinston, 2000: When are exclusive contracts likely to protect relation-specific investments against opportunistic behaviour? Incumbent (I) and/or buyer (B) can make a (noncontractible) investment; buyer may later buy from external seller (E). An exclusive contract prevents B from buying from E, but exclusivity is renegotiable. Their analysis relies on a number of special assumptions, and the following conclusions are very tentative (more research needed before policy implication can be drawn). 142

Segal-Whinston, cont’d Incumbent invests

Buyer invests

ED increases Complementary investment (absent Investments (increase ED, I would invest value of trade btw. little, because E and B) investment favours E)

ED decreases investment (absent ED, B would invest more, to increase value of trade with E)

ED decreases Substitutable investment (absent Investments (decrease ED, I would invest value of trade btw. more, to worsen E’s E and B) position)

ED increases investment (absent ED, B would invest less, not to reduce value of trade with E)

143

Balancing anti- and pro-competitive effects Currently, in the EU per se prohibition of exclusive contracts by a dominant firm Better to adopt a rule of reason approach: iii. Are exclusionary effects likely? iv. What are the likely pro-competitive effects of such a practice? v. Are the latter likely to outweigh the former?

144

A case: Ice-Cream (1995) Long series of similar cases in Europe. Here, a EU case (German market). Mars complained that exclusive agreements of Langnese-Iglo (LI) and Schöller hindered its sales in the German market for industrial impulse icecream (excludes craft trade). Commission of the EC (CEC) finds that LI and Schöller had infringed art. 81 and prohibited the two firms from using such agreements. CFI upholds the CEC’s decision. 145

The market Growing market after reunification LI and Schöller have over 45% and 20% of market as defined by CEC; roughly half if craft ice-cream included. Other manufacturers less than 10%. Sales occur directly, through agents, or brokers (the latter have no exclusivity agreements): – Grocery trade (supermarkets, food stores…) – Traditional trade, i.e.: specialised trade (kiosks, petrol stations, cake shops, cinemas, theatres) and catering trade (hotels, restaurants, cafés…) 146

The agreements Manufacturers often use exclusive agreements: buyer cannot sell rivals’ ice-cream in its outlet Manufacturers also provide (on loan) the customer with one or more freezer cabinets, with the clause that they cannot store rivals’ ice-cream. Duration of these agreements: on average 2.5 years. Percentage of sales made by tied outlets: disputed With CEC’s definition, tied distributors account for 15% (LI) and 10% (Schöller) of total market. With broader definition, much less. 147

Foreclosing effects of exclusivity Do these exclusive agreements foreclose rivals? Entry not easy: large sunk costs for advertising and distribution; several outlets tied to existing sellers; [would shops take new cabinets?] But Mars is not a minnow, and is already established with other products in Germany. Theory: exclusionary effects possible (large number of uncoordinated retailers; presumably little competition in the downstream market)

148

Efficiency effects of the agreements Manufacturers make large investments to provide shops with freezers (this explains high growth). Standard free-riding arguments call for rivals not to make use of another seller’s investments! But this applies to freezer cabinet exclusivity. Doubtful that outlet exclusivity is also crucial (to guarantee certainty and stability of demand?) Unclear that outlet exclusivity necessary to enhance retailer’s services (retailers do little or no advertising, training, promotion to sell ice-cream) 149

Conclusions Neither the CEC nor the CFI make a distinction between freezer exclusivity and outlet exclusivity. While outlet exclusivity has a higher foreclosing potential and no clear efficiency effects, freezer cabinet exclusivity has little foreclosing potential (Mars and others could give retailers other freezers) and clear efficiency effects.

150

Case II: Mobile Telephony Dealers in Italy “Costituzione Rete Dealer GSM”, Italian Antitrust case (2 May 1996) Italian market for mobile communications. Former monopolist incumbent (Telecom Italia Mobile), imposes exclusivity on its dealers as soon as a new rival (Omnitel) is allowed to enter. Here, unlikely that efficiency defences would hold, and the risk that liberalisation process is blocked at the outset is enormous. Correct to find exclusivity clauses abusive. 151

4e. Policy implications Strong presumption VR enhance efficiency Possible anti-competitive effects only when enough market power exists Market power, not the type of agreement adopted, matters (=> change in the approach to VR in Europe in the late 90’s) Large enough market power: rule of reason, balancing efficiency with (possible) adverse effects 152

4f. Vertical mergers When two firms at successive production stages merge. Possible efficiency effects from vertical integration: Avoiding double marginalisation problems Generally, avoiding (or reducing, since agency problems might still exist) vertical externalities, as joint control of upstream and downstream firms Avoiding opportunistic behaviour and promote specific investments Better coordination of investments and innovations 153

Vertical mergers: Possible exclusionary effects? If an upstream monopolist integrates downstream, would it foreclose input to downstream rivals? Chicago School: No foreclosure effects, because of “Single Monopoly Profit” theory (SMP theory: if downstream firms are perfectly competitive, all profits from the vertical industry can be reaped by the input monopolist: no incentive to merge for foreclosure) Modern IO: yes, foreclosure might be both feasible and profitable, e.g. because of the “commitment problem” (Hart-Tirole) 154

Two-step analysis of vertical foreclosure Step 1. Does the vertical merger harm downstream rivals? (does the input price paid by them rise?) a. Will integrated firm cease to supply? (raise input prices?) No, if downstream rivals serve different markets and/or other efficient upstream firms exist b. Assume integrated firm ceases to supply. Input price paid by downstream rivals will not necessarily rise: (i) other upstream firms might increase supply; (ii) lower demand for input (downstream affiliate withdraws from input market) tends to reduce prices 155

Two-step analysis of vertical foreclosure, II If foreclosure will occur (if not, investigation stops): Step 2. Does the vertical merger harm competition? (i.e., which final effect on welfare?) a. Because of removal of double marginalisation problems, downstream affiliate will tend to reduce prices b. If downstream rivals pay higher input price, but are competitive enough, difficult for downstream affiliate to raise prices even if it wanted c. If indeed vertical firm has market power, this may be balanced by efficiencies from vertical integration 156

5. Price discrimination Important, both in itself and to understand other practices Three types of price discrimination (PD): 1st degree (perfect) PD 2nd degree PD: self-selection of consumers 3rd degree PD: if different observable characteristics Two main ingredients of price discrimination - ability to “sort out” different consumers and charge them different prices - no arbitrage opportunities (=no parallel imports) 157

Welfare effects of PD PD is not always bad: the extreme case of 1st degree PD, under which the first-best is attained (provocative but unrealistic example) Quantity discounts (2nd degree PD). Some consumers have higher demand than others. If consumers are charged T+pq, the unit price (T/q+p) decreases with the number of units bought. Welfare increases because the fixed fee is used to extract surplus, allowing for a lower variable component than under linear pricing (Proxy of the first-best case where p=c and T is used to extract all the surplus) 158

3rd degree PD and parallel imports Suppose h and l are two regions (h=rich; l=poor) or identifiable consumer groups, with different demands. If price discrimination is allowed, the firm sets ph>pl. If price discrimination is prohibited, two cases may arise: 4) Under uniform pricing, sales in both markets. In this case: Πd > Πu , but Wd<Wu. 5) Under uniform pricing, one market is not served: the firm may prefer to set ph even if this implies no sales in country l. (This happens when country l is relatively unimportant, for instance.) In this case: Πd >Πu and Wd >Wu.

General result: PD welfare detrimental if qPD does not increase. 159

Further remarks PD and investments. Since PD increases the firms’ profits, the uniform pricing policy may have long-run negative effects (on investments, innovations etc.) PD and market power. Both small and large firms will have incentives to discriminate prices across countries. But in the former case welfare effects are less relevant. To the extent that PD will induce firms to invest more, allowing ‘small’ firms to engage in PD may foster competition. Sensible to use a safe harbour: allow PD for firms below a certain market share or if not dominant (not the current EU policy).

160

PD as monopolisation device PD may be used by an incumbent to exclude rivals. For instance, discriminatory pricing may exclude entrants if buyers are uncoordinated and there are economies of scale (see Figure below) Also, when only some buyers can be contested, price discrimination makes it less costly for the incumbent to exclude (it does not need to lose profits on inframarginal buyers) But an obligation to dominant firms not to discriminate (transparent pricing) may have adverse effects (helps a dominant firm to solve the commitment problem). 161

Discriminatory pricing facilitates exclusion

pm

cI

A

pm

B=π m C D Buyer 1

cI

A B=π m C D Buyer 2

The incumbent can set p=cI to one buyer, p=pm to the other, and it excludes the entrant from both markets. If entrant can match, equilibrium price will be below cI. (The incumbent makes losses on one buyer but recovers on the other. Entrant needs both buyers, so it should162

Which policy implications? Price discrimination: difficult to arrive at clear-cut policy recommendations (see more on rebates) If no dominance, probably best not intervene (but in the EU, efficiency is not the only objective of competition law) If dominance, rule of reason: price discrimination implies more aggressive pricing (each customer can be contested) - if market structure unaffected, good - if exclusion can arise, bad 163

Abusive practices (Introduction) a. Exclusive dealing b. Price discrimination c. Rebates d. Predatory pricing e. Excessive pricing f. Margin squeeze g. Tying h. Interoperability

164

Substitutability among exclusionary practices Note that exclusion can be achieved through different practices which are to some extent substitutable. For instance: If I owns both essential input and is seller downstream: - Refusal to deal (or license) Cost cI - Tying UI and DI sales - Price squeeze - Degrade interoperability

UI

Cost c0

a DI

Cost cE

E

p

165

Substitute practices, cont’d Suppose I wants to exclude E. Different practices are possible, including: - Exclusive dealing - Price discrimination - Rebates (quantity discounts) - Predatory pricing Policy implication: a form-based approach does not make sense.

I

B1

E

B2

166