Presentation on theme: "Www.ccp.uea.ac.uk Tacit versus overt collusion. Firm size asymmetries and numbers: overview of current CCP research Stephen Davies (drawing on research."— Presentation transcript:
www.ccp.uea.ac.uk Tacit versus overt collusion. Firm size asymmetries and numbers: overview of current CCP research Stephen Davies (drawing on research by Davies, De and Olczak) CCP 5 th Annual Conference June 19th 2009
Motivation Tacit collusion and cartels: different ways of ‘skinning the cat’? But how far are they direct substitutes for the firms concerned? Relevant to a variety of issues: - Theory: does one model fit all? - Empirical: estimating the ‘degree of market power’ in structural models – when is ‘market power’ evidence of a cartel? -Policy – what’s the counterfactual for cartels (e.g. in calculating overcharge); what evidence to look for in coordinated effects mergers etc.? -My own interest is in the (differing) roles for firm numbers and asymmetries within the cartel/tacitly collusive group Purpose of this presentation: overview some of our own work in CCP and future plans, in three areas: –Merger control –Cartels –Experimental work
Collusion: firm numbers and asymmetries S-C-P (e.g. Scherer/Ross): collusion more likely the fewer the number and the ‘more similar’ they are Confirmed in theory of simple repeated games (e.g. Ivaldi et al 2003) Cautionary note: measuring symmetry by market shares may be empirically expedient, but potentially a poor indicator of the underlying asymmetries – which are what really matter (costs, capacity, differentiation)
The oligopoly triangle A useful descriptive device for comparative purposes & to focus the mind Not all referees agree! When comparing small number oligopoly markets, plot S1 against S2:
S2 (%) S1 (%) S1=100-S2 S1=S2 S1=100-2S2 100 MON DUO TRI 50 33 A
S2 (%) S1 (%) S1=100-S2 S1=S2 S1=100-2S2 100 MON DUO TRI 50 A 33 B C D S1=50-S2 S1=100-3S2
Interpreting the triangle segments Regions in ascending size of fringe (F): A: literal triopoly (zero F) or very small fringe S2>S3+F B: Fringe + S3 larger than S2 C: Fringe larger than S2 (sufficient) D: Fringe larger than S1 (sufficient)
EC Coordinated Effects Mergers – Davies, Olczak, Coles (2008) One policy area where competition authorities are required to assess likelihood of tacit collusion = coordinated effects merger decisions Objective of research: to identify market structures where EC’s merger decisions imply that tacit collusion anticipated to occur if merger allowed Sample: 62 mergers (386 markets) from 1990-2004 where collective dominance (tacit collusion) a serious concern. In these, the EC intervened for: collective dominance (CD) in 29 markets single dominance (SD) in 89 markets.
Markets where interventions occurred – choice between CD and SD
EXHIBIT 1: EC’s implicit ‘model’ of tacit collusion
Experimental literature on tacit collusion Huck et al (2004) meta-analysis 19 previous studies, Cournot experiments, 2-5 symmetric firms: no communication, fixed groups interacting repeatedly, homogeneous product, usually linear model. Measuring collusion (inversely) by ratio of experimental ‘industry’ output to analytical Cournot-Nash outcome, they find, across these studies, statistically significant inverse correlation between N and collusion. But, on average, it is only in two firm markets that actual output < the Cournot level. Their own experiments corroborate the meta analysis – collusion sometimes occurs when there are only 2 firms, but never in with N>3. Even in 3 firm markets, the average outcome is close to the Nash equilibrium. Hence: “Two are few and four are many” Literature on asymmetries much thinner. Huck et al report only two in the Cournot setting: Mason et al (1992) - price lower where firms have asymmetric, rather than symmetric, costs.
EXHIBIT 2 Fonseca & Normann (2008) Bertrand-Edgeworth (Compte et al): 4 alternative market structures. N=2 or 3; capacity symmetric or asymmetric. Find: p C >p D >p A >p B Reductions in firm numbers lead to increased prices, both in the symmetric and unsymmetric cases i.e. p C >p A and p D >p B Asymmetry leads to reduced prices, holding numbers constant at either 2 or 3, i.e. p C >p D and p A >p B
Cartels: numbers of participants (surveys) Hay & Kelley Fraas & Greer PosnerLevenstein & Suslow De Mean 7.2516.729.1n.a.6.1 Median 786-1085 N<10 79%60%64%63%85%
Typology of EC cartels “Tacit-collusive compatible” – consistent with structures in EC coordinated effects mergers: broadly symmetric concentrated duopolies “Dominant leader” – largest firm accounting for 50%+ of cartels sales, and largest rival typically much smaller (say <20%) “Unconcentrated” – neither of largest firms accounting for much more than 40% or 30%, and usually much less. This type is fairly heterogeneous, including five or six cartels which might be categorised as triopoly or quadropoly, but the other twenty entailing very significant fringes.
Where next? Why are large asymmetries and N possible in cartels but not tacit collusion? And does one model really explain tacit collusion and cartels? Obviously communication & maybe ringleaders matter – ongoing work in experimental? What are the sources of asymmetries in cartels? De: is it really asymmetry in the senses described in theory? Do cartels only form when tacit collusion unsustainable? More empirical work investigating cartels before and after (Olczak also looking at theoretically.) Ditto substitutability of collusion & mergers. Mergers: are our findings peculiar to EC? Other jurisdictions? Do structural remedies remove dominance concerns? Can a remedy imposed to prevent SD make CD more likely and vice versa? (Davies and Olczak 2008b) Is tacit collusion necessarily worse for welfare than single dominance? – NO, not if tacit collusion is unstable (Olczak 2009)