4820–9 Information Geir B. Asheim Information and strategic behavior Introduction Static competition 4820–9 Dynamic competition Geir B. Asheim Department of Economics, University of Oslo ECON4820 Spring 2010 Last modified: 2010.03.16 Asymmetric information . . . 4820–9 Information Geir B. Asheim Introduction Static competition Dynamic competition . . . gives opportunity for signaling and forming a reputation Such firm behavior is strategic if it changes the competitors’ beliefs about the private information of the firm. Outline (1) Static competition under asymmetric information (a) Bayesian equilibrium (b) A simple model of price competition (2) Dynamic competition under asymmetric information (a) Perfect Bayesian equilibrium (b) The Milgrom-Roberts (Ecma 1982) model of limit pricing (c) Multi-market reputation Equilibrium concept for static games of asymmetric information 4820–9 Information Geir B. Asheim Introduction Static competition Bayesian equilibrium Price competition Dynamic competition Definition (Bayesian equilibrium) A Bayesian equilibrium is a profile of type-contingent strategies {ai∗ (ti )}ni=1 such that each player maximizes his expected utility contigent on his type and taking the other players’ type-contingent strategies as given: ai = ai∗ (ti ) maximizes pi (t−i |ti )Πi (a1∗ (t1 ), . . . , ai , . . . , an∗ (tn ), t1 , . . . , ti , . . . , tn ) t−i Example: Firm 2 does not know the cost of firm 1. But firm 2 has a probability distribution over what costs firm 1 may have. And firm 1 knows what probability distribution firm 2 has. How are these interactive beliefs formed? A simple model of price competition (1) 4820–9 Information Geir B. Asheim Structure: (1) Prices set (2) Demand determines quantities Firms 1 and 2 have constant unit cost c1 and c2 and set prices p1 and p2 simultaneously. Sales are given by Introduction Static competition Bayesian equilibrium Price competition Dynamic competition Di (pi , pj ) = a − bpi + dpj which is the demand function for firm i. max(pi − ci )Di (pi , pj ) pi Firm 1’s unit cost c1 is c L with probability x and c H with probability 1 − x, where c L < c H . Write c e ≡ xc L + (1 − x)c H . Firm 2’s unit cost c2 is c. Firm 1 knows its own cost; firm 2 believes that firm 1’s cost is c L with probability x and c H with probability 1 − x. A simple model of price competition (2) 4820–9 Information Geir B. Asheim Introduction Static competition Bayesian equilibrium Price competition Dynamic competition Results: The low cost type of firm 1 charges a higher price than it would have done if information were symmetric. Why? The high cost type of firm 1 charges a lower price than it would have done if information were symmetric. Why? Suppose firm 1 could report verifiable information about c1 costlessly before the firms compete. Would such information be spread if firm 1 wants to accommodate entry? Would such information be spread if firm 1 wants to deter entry? Equilibrium concept for dynamic games of asymmetric information 4820–9 Information Geir B. Asheim Introduction Static competition Dynamic competition Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation Definition (Perfect Bayesian equilibrium) A Perfect Bayesian equilibrium is defined by the following two properties: (1) Given beliefs and the strategies of the other players, each player’s strategy must satisfy sequential rationality. (2) Beliefs are obtained from strategies and observed actions using Bayes’ rule whenever possible. Problems: (i) Many equilibria (How to coordinate? How to predict?) (ii) How are interactive belief about player types formed? Limit pricing 4820–9 Information Geir B. Asheim Introduction Static competition Dynamic competition Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation Can an incumbent firm deter entry by setting a low pre-entry price? (Limit pricing, Bain, 1949) Only if a low price before entry signals a low price after entry. Before: A large K is an aggressive commitment which has the unintentional affect of leading to a low price. Now: A low p signals a harsh post-entry environment. Milgrom-Roberts (Emca 1982) model of limit pricing 4820–9 Information Geir B. Asheim Introduction Static competition Dynamic competition Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation Model: Two periods Two firms Structure: t =1 Firm 1 is the incumbent, has private information about own cost, and sets price (p1 ) in period 1. t =2 Firm 2 observes p1 and makes an entry decision. After entry, firm 2 observes firm 1’s true cost before the firms compete in period 2. Assumptions 4820–9 Information Geir B. Asheim x: Probability that firm 1’s cost = L 1 − x: Probability that firm 1’s cost = H Introduction Static competition M1t (p1 ): Monopoly profit for firm 1, given p1 and cost = t Dynamic competition M1t = max M1t (p1 ) Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation t = arg max M t (p ) pm 1 1 D1t : Duopoly profit for firm 1, given cost = t D2t : Duopoly profit for firm 2, given cost = t D2H > 0 > D2L M1t > D1t , t = L, H Separating equilibrium 4820–9 Information Geir B. Asheim Introduction Static competition Dynamic competition Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation Result The incumbent manipulates his price, but the entrant is not fooled. He learns the incumbent’s cost perfectly. Entry occurs exactly when it would have occurred under symmetric info. Result Even though the incumbent doesn’t fool the entrant, he engages in limit pricing: the low-cost type would be mistaken for the high-cost type if it did not sacrifice short-run profits to signal. Result Social welfare is higher than under symmetric information. 2nd period welfare is not affected as entry is not affected. 1st period welfare is increased because the low-cost type reduces its price. Pooling equilibrium 4820–9 Information Geir B. Asheim Introduction Static competition Dynamic competition Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation No pooling equilibrium if xD2L + (1 − x)D2H > 0. Assume xD2L + (1 − x)D2H < 0. Result The incumbent manipulates the price without revealing information. The entrant does not enter when appropriate. Result The low-cost type charges its monopoly price. The high-cost engages in limit pricing to deter entry. Result Ambiguous welfare consequences. Higher in 1st period as type H sets lower price. Lower in 2nd period due to less entry. Multi-market reputation Resolving the “chain store paradox” 4820–9 Information Geir B. Asheim Introduction Static competition A firm that is established in many markets will wish to fight entrants aggressively (even at a loss in each market) to improve its reputation as a tough competitor in other markets. This will deter further entry. Dynamic competition Perfect Bayesian equilibrium Limit pricing Separating equilibrium Pooling equilibrium Multi-market reputation Not a subgame-perfect equilibrium, since the incumbent will not fight in the ultimate market, and hence, not in the penultimate market etc. (Chain store paradox) The chain store paradox is resolved by introducing asymmetric information (Kreps & Wilson, Milgrom & Roberts, JET 1982)