4820–7 Strategic commitments Geir B. Asheim Strategic commitments Introduction Short- and long-term commitments 4820–7 Short-run – Price competition Short-run – Quantity competition Geir B. Asheim Long-run – Capacity competition Department of Economics, University of Oslo ECON4820 Spring 2010 Last modified: 2010.03.02 Why model repeated strategic commitments? 4820–7 Strategic commitments Geir B. Asheim Problem with repeated games No possibility for strategic commitments No change in fundamentals Introduction Markov perfect equilibrium Outline Short-run – Price competition Short-run – Quantity competition Long-run – Capacity competition Problem with two-stage competition The second stage is a static game No response to actions in the second stage Need for models with Dynamic rivalry Competition over a longer time horizon Actions change fundamentals Possible to make strategic commitments Markov strategies simplify the analysis State-dependent strategies – depending on payoff-relevant history 4820–7 Strategic commitments Geir B. Asheim Introduction Markov perfect equilibrium Outline Short-run – Price competition Short-run – Quantity competition Long-run – Capacity competition Definition (State) The state at two nodes in a game tree is the same if and only if the subgames defined by the nodes are identical. Example (An infinitely repeated game) The state is the same at all nodes. Example (A finitely repeated game) The state is the same at all nodes at a given time. Example (A two-stage game) The state in stage 2 dep. 1-to-1 on 1’s choice of K in stage 1. Definition (Markov strategy) A Markov strategy determines the same action in the same state Markov-perfect equilibrium 4820–7 Strategic commitments Geir B. Asheim Definition (Markov-perfect equilibrium – 1) A Markov-perfect equilibrium is a subgame-perfect equilibrium where players use Markov strategies. Introduction Markov perfect equilibrium Outline Short-run – Price competition Short-run – Quantity competition Long-run – Capacity competition Definition (Markov-perfect equilibrium – 2) A Markov-perfect equilibrium is a strategy profile where, in all subgames, each player uses his best Markov strategy, given the Markov strategies of his opponent. Result (The definitions are equivalent) A player’s set of best responses contains a Markov strategy if the opponents use Markov strategies. Comparison of . . . 4820–7 Strategic commitments Geir B. Asheim Introduction Markov perfect equilibrium Outline repeated games No strategic commitments only one state (no state variable) Tactical competition is studied ‘Bootstrapping’ Short-run – Price competition Short-run – Quantity competition Long-run – Capacity competition dynamic games studied by means of Markov strategies Strategic commitments change state variables Tactical competition are not studied Reduced form Π-function where profit depends on the state Outline 4820–7 Strategic commitments Geir B. Asheim Introduction Markov perfect equilibrium Outline Short-run – Price competition Short-run – Quantity competition Long-run – Capacity competition Short-run commitments Price competition Maskin, Tirole, A theory of dynamic oligopoly, II: Price competition, kinked demand curves, and Edgeworth cycles, Ecma 56 (1988) 571–599 Quantity competition Maskin, Tirole, A theory of dynamic oligopoly, III: Cournot competition, EER 31 (1987) 947–968 Long-run commitments Capacity competition Fudenberg, Tirole, Capital as commitment: Strategic investment in continuous time, J Econ Theory 31 (1983) 227–256 Short-run commitments Price competition 4820–7 Strategic commitments Geir B. Asheim Motivation: Foundation for “kinked-demand curve” Menu costs explain that price can be short-run commitments Introduction Short-run – Price competition Markovperfect equilibrium Examples Short-run – Quantity competition Long-run – Capacity competition Model: Firms alternate at setting prices ⎧ ⎪ ⎨ D(pi ) 1 Di (pi , pj ) = D(pi ) 2 ⎪ ⎩ 0 if pi < pj if pi = pj if pi > pj Per period profit: Πi (pi , pj ) = (pi − c)Di (pi , pj ) Intertemporal profit at time t: ∞ s=0 δ s Πi (p i,t+s , pj,t+s ) A Markov-perfect equilibrium (R1 , R2 ) satisfies: 4820–7 Strategic commitments Geir B. Asheim Introduction Short-run – Price competition Markovperfect equilibrium Examples Short-run – Quantity competition Long-run – Capacity competition There exists functions V1 , W1 , V2 and W2 such that 1 V1 (p2 ) = maxp Π (p, p2 ) + δW1 (p) Discounted profit given best choice now and the players follow (R1 , R2 ) later. R1 (p2 ) = arg maxp Π1 (p, p2 ) + δW1 (p) The reaction function specifies a best choice. W1 (p1 ) = Π1 (p1 , R2 (p1 )) + δV1 (R2 (p1 )) Discounted profit when the opponent acts and the players follow (R1 , R2 ). and likewise for V2 and W2 . R1 and R2 may specify mixed actions; if so, W1 and W2 are found by taking expectations. Example 1: Kinked demand curve Example 2: Edgeworth cycles 4820–7 Strategic commitments Stage game: D(p) = 3 − p and c1 = c2 = c = 1 p(x) = Geir B. Asheim Introduction Short-run – Price competition Markovperfect equilibrium Examples Short-run – Quantity competition Long-run – Capacity competition Πi (x) = x+3 3 where x ∈ {0, 1, 2, 3, 4, 5, 6} ⎧ x(6−x) ⎪ ⎨ 9 x(6−x) ⎪ 18 ⎩ 0 if pi = p(x) < pj if pi = p(x) = pj if pi = p(x) > pj Ex. 1: p p(6) p(5) p(4) p(3) p(2) p(1) p(0) R(p) p(3) p(3) p(3) p(3) p(1) p(3) p(3) p(1) Ex. 2: p p(6) p(5) p(4) p(3) p(2) p(1) p(0) p(5) R(p) p(4) p(4) p(3) p(2) p(1) p(0) p(0) Firms react to defend market shares, not to punish (as in repeated games) Short-run commitments Quantity competition 4820–7 Strategic commitments Geir B. Asheim Introduction Short-run – Price competition Short-run – Quantity competition Markovperfect equilibrium Properties Long-run – Capacity competition Motivation: Dynamic reactions Firms alternate at being Stackelberg leaders Production lags (time between change in production plans and change in output) explain that quantity can be short-run commitments Model: Firms alternate at choosing quantity Per period profit: Πi (qi , qj ) Πij < 0 — Increased quantity is aggressive Πiii < 0 — Second-order condition Πiij < 0 — Strategic substitutes (e.g., Cournot competition) Intertemporal profit at time t: ∞ s=0 δ s Πi (q i,t+s , qj,t+s ) A Markov-perfect equilibrium (R1 , R2 ) satisfies: 4820–7 Strategic commitments Geir B. Asheim Introduction Short-run – Price competition Short-run – Quantity competition Markovperfect equilibrium Properties Long-run – Capacity competition There exists functions V1 , W1 , V2 and W2 such that 1 V1 (q2 ) = maxq Π (q, q2 ) + δW1 (q) Discounted profit given best choice now and the players follow (R1 , R2 ) later. R1 (q2 ) = arg maxq Π1 (q, q2 ) + δW1 (q) The reaction function specifies a best choice. W1 (q1 ) = Π1 (q1 , R2 (q1 )) + δV1 (R2 (q1 )) Discounted profit when the opponent acts and the players follow (R1 , R2 ). and likewise for V2 and W2 . Properties 4820–7 Strategic commitments Geir B. Asheim Introduction Short-run – Price competition Short-run – Quantity competition Markovperfect equilibrium Properties Long-run – Capacity competition The dynamic reaction curves are downward sloping if Πiij < 0. The dynamic reaction curves coincide with the static response curves if δ = 0 The dynamic reaction curves are outside the static response curves if δ > 0 Long-run commitments Capacity competition 4820–7 Strategic commitments Geir B. Asheim If capital does not depreciate and cannot be sold, then capital accumulation is a long-run commitment Introduction Short-run – Price competition If capital stocks are strategic substitutes, then the firm will engage in an investment race The firm that is ahead will preserve its lead, trying to limit the other’s mobility Short-run – Quantity competition Long-run – Capacity competition Model Are capital stocks strategic substitutes? Yes, if capital stocks determine capacity, and the firms compete in prices in the short-run Model 4820–7 Strategic commitments Geir B. Asheim Introduction Short-run – Price competition Short-run – Quantity competition Long-run – Capacity competition Continuous time and an infinite horizon Ki (t): i’s capital at time t. K1 (0) ≥ 0. K2 (0) = 0. Ii (t) = K̇i (t) ≡ dKi (t) dt ∈ [0, Ī ] Per period profit: Πi (Ki , Kj ) Πij < 0 — Increased capacity is aggressive Πiii < 0 — Second-order condition Πiij < 0 — Strategic substitutes (e.g., Cournot competition) Model Intertemporal at time t: ∞ −rt profit (Πi (Ki (t), Kj (t)) − I (t))dt s=0 e Consider the case where r → 0