1/34
Looks like no tags are added yet.
Repeat Game (Collusion)
- The same stage game (e.g., the prisoner's dilemma) is played over multiple periods
- Players see each other's past actions, and future payoffs are discounted by a factor β, so total payoff = Σ βᵗ·uₜ
- A higher β means the player is more patient
Backward Induction
- The way to solve a finitely repeated game: start at the last period and work backward. In the last period, there is no future, so players play the one-shot Nash equilibrium (defect).
- Knowing that, they defect in the second-to-last period too, and so on back to period 1.
- General result: if a stage game has a unique Nash equilibrium, repeating it a known, finite number of times means playing that equilibrium every period.
- Cooperation unravels
Grim Trigger Strategy
- Cooperate as long as the opponent has always cooperated
- If they defect even once, defect forever
- This only works when the game is infinite or has no known end, because then backward induction can't be applied
- The threat of losing all future cooperation keeps players in line
Condition for cooperation (repeated PD)
- Cooperation is sustainable if the payoff from cooperating forever is at least the payoff from defecting once and being punished forever after - In the class example this works out to β ≥ 1/6
Cooperation is not the unique equilibrium: defecting every period is also a Nash equilibrium
Credibility of threats
- A punishment only deters if the punisher would actually carry it out when the time comes
- Grim trigger is credible in the PD because reverting to mutual defection is itself a Nash equilibrium, so no one wants to deviate from the punishment
Unknown end date
- Real interactions are never truly infinite
- Collusion can still be rational if firms do not know when the relationship will end, for example, if the game continues each period with some probability
- With a known deadline, backward induction kills cooperation
Three requirements for sustained cooperation (recap slide)
Players must be sufficiently patient, threats must be credible, and the end date must be unknown
Collusion
- In an oligopoly, each firm's price or quantity choice affects its rivals' profits; these effects are competitive externalities
- Firms collude when they agree on prices or quantities that soften these externalities
- They can also collude on other variables like investment or advertising
Explicit vs. tacit collusion / enforcement problem
- Explicit agreements to fix prices, divide markets, or restrict output are illegal in the US and EU
- Because a collusive "contract" can't be enforced in court, cartels enforce it themselves: mainly through the threat of price wars, or in extreme cases violence (e.g., NYC garbage hauling in the 1990s)
Cartel examples
- OPEC (a legal cartel)
- De Beers diamonds
- The fined vitamins
- Graphite electrodes
- Citric acid
- French mobile telephony cartels
General sustainability condition
- With grim trigger, collusion holds if the collusive payoff stream beats deviating once and then being punished forever
- Here π^collusion is the per-period collusive profit, π^d the one-period profit from deviating, and π* the punishment (competitive) profit
- Collusion is easier when the punishment is harsh (low π*) and when collusive profits are large (high π^collusion)
Collusion in Bertrand
- Each firm gets ½π^m when colluding
- Undercutting slightly gives about π^m
- The punishment is marginal-cost pricing, so π* = 0. This gives β ≥ 1/2
Collusion in Cournot
- The condition is β ≥ (n+1)² / (n² + 6n + 1)
- Key lesson: the more firms there are, the harder collusion is to sustain, because each firm's gain from deviating grows relative to its share of the collusive profit
Partial collusion
- If β is too low for full collusion, firms can still coordinate on a less aggressive outcome, which reduces the temptation to deviate
-In Cournot, this means total output between the monopoly and Nash levels
Imperfect monitoring
- If firms can't observe rivals' actual transaction prices (as opposed to posted prices), a deviation may go unpunished, which makes collusion harder
- Sharing price information is risky: firms may lie, and communication can attract antitrust scrutiny
Danish ready-mixed Concrete Case
- Prices were negotiated privately, so price cuts were unobservable
- In 1993 the Danish competition authority began publishing firms' transaction prices to increase "transparency."
- Prices rose 15-20% within a year, because transparency made collusion easier to monitor
- The authority stopped publishing in 1996. Lesson: well-intended transparency can facilitate collusion
Meet-the-competition clause
Customers can claim a refund if a rival charges less
This turns consumers into unwitting monitors of the cartel
Most-favored-customer clause
- Past customers get a refund if the firm cuts its price later
- This raises the cost of cutting prices, so it weakens the incentive to deviate
Price wars / why grim trigger is unrealistic
- In the simple model the threat works perfectly, so deviation, and therefore a price war, never happens in equilibrium
- Observed price wars imply firms face uncertainty, such as a demand shock that is hard to distinguish from a rival cheating
- Grim trigger is also too harsh for a world with mistakes
Finite (minimal) punishment
- Instead of punishing forever, punish for T periods and then return to collusion
- The minimal punishment length is the smallest T that still deters deviation
Determinants of collusion (conclusion slide)
- Collusion is easier with frequent interactions, observable prices or quantities, no foreseen end, and few firms
Merger
- The legal consolidation of two firms into one
- The two key questions are whether the merger is profitable for the firms, and whether it improves welfare from the government's view
Horizontal merger
A merger between competitors in the same industry
It reduces the number of firms and increases market power, often raising prices
Merger profitability condition
A merger is profitable if π(n−1) ≥ 2π(n): the merged firm's profit in a market with one fewer firm must exceed the two firms' combined pre-merger profits
Duopoly → monopoly merger
With n = 2, the merger is always profitable, since ¼ > 2/9
Mergers in fragmented markets (merger paradox)
- With n ≥ 3 and no synergies, a two-firm merger is not profitable in Cournot.
- Non-merging firms capture much of the gain from reduced competition, and the merged firm loses market share.
- For example, with n = 3 the merging pair's share falls from 2/3 to 1/2. The slides don't use the name "merger paradox," but that's the standard term if it comes up
- Key lesson: mergers are more likely to be profitable in concentrated markets than in fragmented ones
Total welfare standard vs. consumer surplus standard
- The two criteria a regulator can use
- Under the total welfare standard, CS + PS must rise after the merger
- Under the consumer surplus standard, only CS must rise
- US and EU authorities typically focus on consumer surplus
Superior Propane (Canada, 1998)
- The merger substantially lessened competition, causing C$6M/year in deadweight loss
- But it created C$29M/year in cost efficiencies, so the Tribunal approved it under the total surplus standard
- The Court of Appeal later required more weight on consumer surplus
- The case illustrates the difference between the two standards
Clayton Act Section 7 (1914)
- The US law prohibiting mergers that may substantially lessen competition or tend to create a monopoly
- Enforced by the FTC and the DOJ Antitrust Division
Consumer surplus in the model
- With P = a − q, CS = ½q². A merger helps consumers only if it raises total output
- Without synergies, fewer firms means less output, so a regulator would block every merger under either standard
Efficiency defenses / welfare-enhancing mergers
- A merger can be good for society if it brings scale economies, production synergies, greater innovation capacity, or access to scarce resources
Synergies and the threshold x̂
- If merging lowers the merged firm's cost from c to c − x, the merger is profitable when x > x̂
- x̂ is the minimum cost reduction needed to compensate the merging firms for the market share they lose to outsiders
- x̂ can be negative with n = 2 (profitable even with slightly negative synergies) and is positive with n ≥ 3
Consumer-surplus threshold x′
- Under the CS standard, the regulator allows the merger only if x ≥ x′ = (a − c)/(n + 1)
- Key lesson: x′ > x̂, so for synergies between x̂ and x′ a merger is profitable for the firms but harmful to consumers
- Cost savings alone don't guarantee consumers benefit
Vertical merger
- A merger between firms at different stages of the supply chain, such as a manufacturer and a retailer
- It often increases efficiency, for example by lowering transaction costs
- Its main risk is foreclosure: limiting rivals' access to inputs or markets
Double marginalization
- When an upstream and a downstream monopolist each add their own markup, the final price ends up higher than a single integrated monopolist would charge
- Profits rise too, from 3/16 to 4/16 of (a − bc)²/b, so both the firms and consumers gain. The slides' phrase: "The only thing worse than one monopoly is two monopolies."
- This is a big reason vertical mergers are blocked less often than horizontal ones