OFT Roundtable on Conjectural Variations Introduction and Coordinated Effects Adrian Majumdar Adrian.Majumdar@rbbecon.com Contributors Adrian Majumdar Benoît Durand Chris Doyle Alan Crawford • We are indebted to Greg Shaffer for his invaluable comments and contributions to Chapters 2 and 5 and to Glen Weyl for his detailed and insightful comments on Chapter 4. • We have benefited substantially from comments from Matthew Bennett, Katie Curry, Stephen Davies, Peter Davis, Amelia Fletcher, Sonia Jaffe, KaiUwe Kuhn, Boryana Miteva, João Santos Silva, Ian Small, Frank Verboven, Chris Walters and from participants at an OFT Roundtable and an OFT seminar to discuss this report. 2 Outline of the report / Terms of reference Introduction – What are CVs and how is their use justified? A backward-looking approach – A forward-looking approach – Can empirical techniques use CV models to tell us anything about how firms competed in the past? Can CVs be used by policy makers, particularly with respect to unilateral effects analysis in merger control in pricing pressure tests? A broader perspective on firms’ reactions – In the spirit of CVs, what are the implications of modelling how firms react to changes in their rivals’ actions for coordinated effects? – In light of this, to what extent should the Airtours criteria be revised…e.g. to include “parallel accommodating conduct” as per the US guidelines? 3 1. What are conjectural variations? 4 1. What are Conjectural Variations? Conjectural variations capture the expectation a firm has about how its rivals will react when it changes its quantity or price. Impact of conjecture on the intensity of competition – Benchmark is zero (standard approach in one-shot simultaneous move game) – Negative conjectures intensify competition (i.e. a reduction of output by one firm is conjectured to be met with an increase in output of other firms). – Positive conjectures dampen competition (i.e. a reduction in output of one firm is conjectured to lead to a reduction in output of other firms) Example: quantity setting duopolists: CV = -1, 0, 1 => perfect competition, Cournot, perfect collusion 5 2. Selective summary of the literature on CVs 6 2. A brief summary of the literature Theoretically awkward (firms cannot react in a one-shot game) Short cut to a more complex game with responses by rivals? Yes – CV use of parameters allows replication of outcomes… but multiple equilibria Consistent conjectures narrow down equilibria? Evolutionary stable? Reachable with boundedly rational agents? Possibly – but might not have equilibrium at all… and may still fail to capture underlying game CV parameters… useful concept for applied work – compare market outcomes against CV=0 benchmark… “as if” firms behaved with those conjectures. Where do conjectures come from? Short-cut to modelling outcome is silent on cause of outcome 7 3. A broader perspective on CV models 8 3. A broader perspective: layout A broader perspective: – In the spirit of CV’s, what are the implications of modelling how firms react to changes in their rivals’ actions for coordinated effects? Summarising the literature (selectively chosen to meet OFT ToR) 1. Continuous punishment strategies 2. Speed of punishment 3. Alternating move models can give rise to stable, high-price equilibria Policy issues: – In light of the literature, to what extent should the so-called Airtours criteria be revised? – Does the CV literature motivate “parallel accommodating conduct” (mooted in the US horizontal merger guidelines) as a coordinated effects theory of harm in merger cases? 9 3. A broader perspective: should the Airtours criteria be revised? – Coordinated effects (Merger Assessment Guidelines) : 1. firms need to be able to reach and monitor the terms of coordination; 2. coordination needs to be internally sustainable among the coordinating group; in particular there must be adequate deterrents to ensure there is no incentive to deviate from the agreement; and 3. coordination needs to be externally sustainable, in that there is little likelihood of coordination being undermined by competition from outside the coordinating group. Similarly, the Airtours criteria: 1. Alignment 2. Internal stability (monitoring and punishment) 3. External stability The report focuses on criteria 1 and 2 10 3. A broader perspective: textbook treatment of coordination • Textbook treatment: threat of ‘punishment’ must outweigh gain from ‘cheating’ ‘Cheating’ Phase ‘Punishment’ Phase Payoff Gain to cheating Impact of punishment Detection Lag Payoff under coordination 11 3. A broader perspective: responding to rival’s actions Continuous punishment strategies – Does the “punishment fit the crime”? (Friedman and Samuelson (1994)) – – arguably more “realistic” to assume that firms react more to greater deviations from an agreement, so that the size of the punishment increases with the magnitude of the deviation. Lu and Wright (2010): continuous punishment strategies in the form of pricematching punishments – if firm 1 lowers its price, then firm 2 will match that price (within certain bounds) – Better scope for alignment with simple punishments (easier to communicate)? – Less harmful collusive outcomes (softer punishment)? – Coordination may be less likely to entirely break down than previously thought, as following a deviation firms may still carry on coordinating, simply to a lesser extent 12 3. A broader perspective: responding to rival’s actions Speed of punishment – Instantaneous price matching – Deviation profit ruled out by assumption? – More effective punishment strategies and therefore more stable coordinated outcomes – Multiple (high price) equilibria with instant price-matching – monopoly outcome focal only with symmetry (Hviid and Shaffer, 1999). Sufficient asymmetry gives rise to competitive outcome. – Kinked-demand curve theories (with implicit pre-play communication, Bhaskar, 1988). If firm i undercuts j, j can undercut i. The pre-play stops when no-one undercuts. Then sales are made at those prices. Symmetry => monopoly; asymmetry => Stackelberg. 13 3. A broader perspective: responding to rival’s actions Alternating move models and supra-competitive prices – Firms commit to a certain action and rivals may be able to respond while the firm is still committed. Implications for competitive effects if this is how firm’s behave? – Markov strategies – respond only to rival’s last action, not history of actions. “Cheating” matters only if occurs in last period. – Maskin and Tirole (1988b) show that stable high price equilibria may emerge from alternating move games: “... the strategies in the supergame literature typically have a firm reacting not only to other firms but to what it did itself. By contrast, a Markov strategy has a firm condition its action only on the other firms' behaviour. Thus, in a price war, a firm cuts its price not to punish its competitor (which would involve keeping track of its own past behaviour as well as that of the competitor) but simply to regain market share. It strikes us that these straightforward Markov reactions often resemble the informal concept of reaction stressed in the traditional I.O. discussion of business behaviour (e.g., the kinked demand story) more closely than do their supergame counterparts". – I rationally anticipate my attempts to gain market share will be met aggressively by you (without us coordinating at all). – Extensions involve multiple equilibria. Edgeworth cycles too – depends on discount factor. 14 3. A broader perspective: what is the competitive outcome? – – Concept 1: What is the competitive level? One-shot outcome? – Even absent conjectures there is a distinction between textbook “collusion” and collusion recognised by Authorities and courts. – Textbook: one penny above the one-shot price sustained in a repeated game is collusion! The most consumer friendly equilibrium of all possible equilibria? – But multiple equilibria are common – How does the system end up at a high-price equilibrium? – Shocks? Not anti-competitive – but is this a competitive outcome? – Focal points – how common are they? – Communication? May be anti-competitive. 15 3. A broader perspective: what is the competitive outcome? Concept 2: do accommodating conjectures imply coordination? – If firms take into account the responses of their rivals, they may well set higher prices than if they took a short term view and ignored these reactions – But this does not itself imply that there is any form of anti-competitive behaviour, or what Authorities might reasonably consider to be coordination in practice – – Why should firms be expected to behave as if they did not take into account their rivals’ responses – is this a realistic benchmark? – A firm having a conjecture as to how its rivals will respond is not anti-competitive in and of itself – A key question: how were those conjectures formed? Arguably there is a distinction between the outcome ultimately obtained, and the manner in which it was arrived at – Simply being aware of rival responses and rationally anticipating them (not coordination?) – Seeking to shape rivals’ reactions to make them accommodating (coordination?) 16 3. A broader perspective: what is the competitive outcome? Concept 3: impact of a merger in the presence of accommodating conjectures – Conjectures remain constant? “Unilateral effect”? See Jaffe and Weyl (2011) – Conjectures change to become more accommodating? “Coordinated effect?” – But why do they change? – – New focal point arises (not necessarily “coordinated”) – Communication among firms is more likely / more effective at moving system to high-price equilibria (coordinated) Is there something in between a coordinated effect and a unilateral effect… or is that distinction too simplistic? – Joe Farrell: mergers can be harmful without neatly falling into the standard categories 17 3. A broader perspective: should the Airtours criteria be revised? Criterion 1: Alignment is under-researched – but important. – How to commit to price matching? Contractual? Pre-play communication? Need for symmetry for the most harmful outcomes. – Multiple equilibria plague this literature! Symmetry matters to appeal to Pareto dominance of monopoly outcome? Criterion 2: Current criteria already address monitoring and punishment… – Current approach encompasses both continuous and instantaneous punishments – Coordination can – in theory at least – be supported where “the punishment fits the crime”. – Re-affirms the current approach that places substantial importance on monitoring (pre-requirement of rapid reaction). Criterion 3: External stability. Not discussed – but important. 18 3. A broader perspective: where does ‘parallel accommodating conduct’ fit in? US Merger guidelines, coordinated effects section “Parallel accommodating conduct includes situations in which each rival’s response to competitive moves made by others is individually rational, and not motivated by retaliation or deterrence nor intended to sustain an agreed-upon market outcome, but nevertheless emboldens price increases and weakens competitive incentives to reduce prices or offer customers better terms.” Link to theory? – Focal point equilibria with Markov strategies (Maskin and Tirole, 1988b)? – Accommodating conjectures? – Can we measure / predict changes in conjectures caused by merger if we do not understand where the conjectures come from in the first place? – When does a merger cause or strengthen this parallel accommodating conduct? We do not deny the possibility of a harmful merger that is not clearly a unilateral or coordinated effects case, but do not consider theory and practice to offer robust guidance on when these mergers are most likely to occur or how often such “in between” cases would fail to be caught – we would not advocate extending the Airtours criteria for the assessment of coordinated effects. 19 Locations and contact London Brussels The Connection 198 High Holborn London WC1V 7BD Telephone +44 20 7421 2410 Email: london@rbbecon.com Bastion Tower Place du Champ de Mars 5 B–1050 Brussels Telephone: +32 2 792 0000 Email: brussels@rbbecon.com The Hague Melbourne Lange Houtstraat 37-39 2511 CV Den Haag The Netherlands Telephone: +31 70 302 3060 Email: thehague@rbbecon.com Rialto South Tower, Level 27 525 Collins Street Melbourne VIC 3000 Telephone: +61 3 9935 2800 Email: melbourne@rbbecon.com Johannesburg Augusta House, Inanda Greens 54 Wierda Road West Sandton, 2196, Johannesburg Telephone: +27 11 783 1949 Email: johannesburg@rbbecon.com 20