techniques Wojciech Dorabialski, Polish Competition Authority (UOKiK) TAIEX Workshop on Detecting Cartels, Tirana 2014 Contents
• Reactive vs proactive cartel detection techniques
• Pros and cons of proactive techniques • Economics of collusion – conditions for tacit collusion • Structural tests • Behavioral tests • Other techniques Two types of detection techniques
Reactive techniques Proactive techniques
• Signals from whistleblowers • Market news monitoring
• Signals from trade partners
• Atitrust activity monitoring and consumers • Market studies
• Signals from competitors
• Intelligence gathering • Market screening using • Leniency programs structural indicators • Reward programs (Korea, UK) • Market screening using behavioral indicators • Econometric models • Manipulation tests • + (Inspections & dawn raids) Pros and cons of proactive techniques
Cons Pros
• More resource-intensive than • Reactive techniques won’t
reactive techniques work if no proactive techniques are employed • Do not provide hard evidence • Provide enough evidence to sufficient to bring a case to justify a formal investigation/ the court dawn raid • Data limitations • Data becoming more easily • Not very popular, no „best accessible practice”, no guides/blueprints • Growing interest, also from the private enforcement community Economics of collusion p.1
• Most economic detection techniques follow from a simple
model of tacit collusion • Repeated oligopoly model (Bertrand or Cournot type) • Collusive outcome with joint monopoly profit is a (non- unique) equilibrium • It must be possible to punish firms that cheat • The punishment is a reversion to competitive prices • Collusive equilibrium condition: profit from collusion must exceed profits from „cheating” Economics of collusion p.2 • The collusive equilibrium condition aka cartel stability condition looks like this: d m m 1 o • Πd – Πm is the difference between profit from deviation and cartel- member profits • Πm – Πo is the difference between cartel-member profits and competitive equilibrium profits • δ is the time-discount factor • The fact that stability condition is satisfied for a particular industry only means that collusion (tacit or overt) is possible as an equilibrium outcome Economics of collusion p.3
• In addition to the cartel stabilty condition, there are two
more requirements that an effective overt catel must fulfill: • 1. An agreement on the exact course of action (in practise, there is no such thing as a commonly known optimal strategy) • 2. Parties to the agreement must be able to monitor each-other’s behavor Economics of collusion p.4 • It follows from all 3 conditions that cartels are more stable (or more likely to form) if: • There are few firms in the industry • The products are homogenous • Demand is predictable (stable or steadily increasing) • The market is mature, there are no innovations • The market is transparent (firm’s strategic decisions easy to observe/detect) • Barriers to entry are high • Spare production capacity is maintained • There is symmetry in costs/production capacity • Buyers are dispersed (no buyer power) Structural indicators • Based on the implications from cartel stability conditions, we can look for indicators that will point to industries, where cartels are more likely to form/survive • The indicators can then be used in an econometric model to estimate likelihood of cartel across industries • One example is Grout&Sonderegger (2005) study for the UK Office of Fair Trading • The dependent variable was whether a cartel has been detected or not • Logit, ordered logit and OLS model estimation was run across 3-digit industry classification, with a few indicators as explanatory variables (industry concentration ratio, R&D expenditure) • The main identification problem – we are identifying the sectors that have similar characteristics to those, where cartels have been detected (we are not really detecting new cartels) Grout & Sonderegger (2005) selected results Industries where cartels were Industries with high estimated detected probability of cartel detection
• Basic chemicals • Telecommunications
• Water transport • Airplane & spacecraft production • Pharmaceuticals • Consulting & audit • Pipes • Flour • Food processing • Cargo services • Non-alcoholic beverages • Travel agency services • Rubber • Book publishing • Plastic • Rail engines and carriages • Pesticides Indices based on structural indicators p.1
• None of the collusion indicators/factors/markers are
necessary for collusion • Other indicators are also possible, based on observations regarding real cartels • We can assign weights/importance to various indicators or create a single aggregate „collusion likelihood index” • Main problem: aggregation method more or less arbitrary Indices based on structural indicators p.2
• The NMa (Dutch Competition Authority) was
probably the first to create such index, called „The Competition Index” • Petit (2012) uses the index to identify 30 industries „prone to anticompetitive behaviour”, the list partly coincides with the lists in G&S • The Romanian Competition Authority created a similar index called AICP (Aggregate Index of Competitive Pressure), the list differs significantly from G&S Behavioral indicators • Instead of looking at structural factors that facilitate the emergence and stabilize cartels, we can ask: how do cartels behave? • If we have a „suspicious” industry and we can observe the behavior of firms in that industry, we may be able to detect cartel activity • Harrington (2006) identifies 2 types of behavioral „collusive markers”: • Price markers • Quantity markers • One must keep in mind, that cartels are easiest to detect during their formation or collapse • Note that the markers/indicators are usually assessed in relations to some benchmark: another geographical market, a similar product market or past performance on the same market Behavioral indicators – price markers • A higher list (or regular) price and reduced variation in prices across customers • A series of steady price increases is preceded by steep price declines • Price rises and imports decline • Firms' prices are strongly positively correlated • A high degree of uniformity across firms in product price and other dimensions including the prices for ancillary services • Low price variance • Price is subject to regime switches Behavioral indicators – quantity markers
• Market shares are highly stable over time
• There is a subset of firms for which each firm's share of total supply for that subset of firms is highly stable over time • A firm's market share is negatively correlated over time Behavioral indicators – „super plus” factors • Marshall & Marx (2012) propose a list of 6 „super plus” behavioral factors • In the opinion of the authors, they indicate a very high likelihood (near certainty) that a cartel is present in the industry • Sharing sensitive information (the kind that firms normally hide from competitors) • Transactions between competitors (inter-firm transfers) at non-market prices • Within-firm shift in incentives (e.g. price before volume) • Dominant-firm conduct by non-dominant firms (e.g. refusals to deal to foreclose a market) • Combo evidence: prices and profits high and increasing when a. at least some firms restrict production, b. market shares stable and firms have excess capacity • A well-specified econometric model performs poorly (does not predict price movement) over the suspected cartelized period • Problem: the factors difficult to observe to industry outsider Other techniques
• In case of bid-rigging in public procurement, structural and
behavioral indicators can be combined and applied systematically • Korean Fair Trading Commission uses the above approach in its Bid-Rigging Automated Analysis System • Econometric techniques can be used when a suspect industry and a suspect cartel period is identified • Data manipulation tests, which identify strange pricing/bidding patters can be applied, e.g. Benford analysis Literature
• Grout P. and S. Sonderegger „Predicting cartels”, OFT Economic
Discussion Paper, March 2005 • Petit L. „The Economic Detection Instruments of the Netherlands Competition Authority: The Competition Index”, NMa Working Papers, January 2012 • Paun R. and P. Prisecaru „An Economic Instrument to Evaluate the (Pro) Competitive Nature of Industries: The Aggregate Index of Competitive Pressure”, Romanian Competition Journal, December 2013 • Harrington J.E. „Behavioral Screening and the Detection of Cartels”, 2006 EU Competition Law and Policy Workshop/Proceedings • Marshall R.C. and L.M. Marx „The Economics of Collusion”, MIT Press, April 2012