1. ECO 340: Micro Theory Intro to Game Theory with applications to I/O Sami Dakhlia U. of Southern Mississippi microprof@gmail.com

2. Game Theory What is a game? Some examples Mixed Strategy Nash Equilibrium Subgame-Perfect Nash Equilibrium Applications to Industrial Organization –Entry Deterrence –Oligopolies

3. What is a game? In its so-called “normal form”, a game consists of three elements: –A set of players –A set of strategies for each player –A payoff function for every possible outcome

4. Some examples The prisoner’s dilemma ConfessDon’t confess Confess(-10,-10)(0,-11) Don’t Confess (-11,0)(-1,-1) Player 1 Player 2 The Nash Eq. (NE) is (confess, confess), regardless of guilt. This, you will note, is not Pareto efficient. Question: how does the Mafia deal with this dilemma?

5. Some examples Matching Pennies HeadsTails Heads(-1,+1)(+1,-1) Tails(+1,-1)(-1,+1) Player 1 Player 2 There is no NE in pure strategies. The unique NE is a so-called mixed strategy NE, where players randomize. Given these particular payoffs each player’s optimal strategy is to just flip the coin for a 50% probability of playing Heads.

6. Subgame-perfect NE The centipede game This is a sequential (or dynamic) game. The normal form no longer fully captures the game, which is why it is now represented in its full arborescence. This is known as the extensive form of the game. Given our assumptions of full rationality and common knowledge of rationality, it is appropriate to use backwards induction to solve the game. The (sad) prediction is that Player 1 will Accept the first offer and walk away with $1. AAAAAA RRRRRR AA R 11112222 (10 0,0)(10 2,0)(10 4,0)(10 6,0)(0,10 1 )(0,10 3 )(0,10 5 )(0,10 7 ) (5x10 6,5x10 6 ) R

7. Applications to I/O Entry-Deterrence Oligopolies

8. Entry-Deterrence Suppose that Lowe’s is the sole supermarket in a particular regional market and that Home Depot has plans to build a competing store. Of course, Lowe’s does not like this idea one bit and promises to be a most aggressive competitor should Home Depot proceed with their plan, indeed so aggressive that Home Depot’s investment would likely yield insufficient returns for a positive profit. The game is thus as follows: Home Depot can choose to enter the market, after which Lowe’s can choose whether to follow up on their threat or whether to simply share the market.

9. Entry-Deterrence To find the NE, it is sufficient to look at the game’s normal form: Home Depot Enter (q) Don’t Enter (1-q) Lowe’s Fight (p) (2,-1)(10,0) Don’t Fight (1-p) (4,4)(10,0) Clearly, there are two NE in pure strategies: (DF,E) and (F,DE). Since Nash Equilibria generically come in odd numbers, an equilibrium in mixed strategies is lurking here as well.

10. Entry-Deterrence Lowe’s will fight for sure (p=1) if the expected payoff of doing so exceeds the expected payoff of not fighting: 2q + 10(1-q) > 4 q + 10(1-q), that is, if q<0, which is not possible. On the other hand, Lowe’s is indifferent between F and DF when q=0. Home Depot will enter if -p+4(1-p) ≥ 0p+0(1-p), that is, p≤4/5. Please plot the Best-Response correspondences in (p,q) space. As you can see, there is actually a whole continuum of mixed strategy Nash equilibria. So, of all these equilibria, which one is most likely? We can answer this by recalling that our game is a sequential one and that we can thus use the concept of subgame-perfection to narrow down our selection. Go ahead and draw the extensive form of the game and use backwards induction. The only SPNE is (DF,E).

11. Oligopolies Please recall the canonical model of profit maximizing monopoly: A firm produces output Q under constant marginal cost (MC) c. Consumer behavior is described by the inverse demand function p=a- bQ. The objective function is ∏ = Revenue - Cost = pQ-cQ and Q is the firm’s choice variable. The first-order condition for an optimum is d∏/dQ = 0. Read this as marginal profit must be equal to 0. By the same token, marginal revenue must be equal to marginal cost: MR=MC. Since Revenue = pQ = (a-bQ)Q = aQ-bQ 2, MR = a-2bQ. Since Cost = cQ, MC = c. MR=MC implies a-2bQ=c, that is, Q = (a-c)/2b Price, finally, is p = a-bQ = a-b(a-c)/2b = (a+c)/2 Please draw the graph in (Q,p) space.

12. Oligopolies In the case of a Cournot Duopoly, the model allows for a second firm producing the same good. The first firm produces quantity Q 1 and the second firm produces quantity Q 2. Thus, total output Q=Q 1 +Q 2. Both firms operate under constant marginal cost (MC) c 1 =c 2 =c. Consumer behavior continues to be described by the inverse demand function p=a-bQ. The objective function of firm 1 is ∏ 1 = pQ 1 -cQ 1 and Q 1 is the firm’s choice variable. Since Revenue = pQ 1 = (a-b(Q 1 +Q 2 ))Q 1 = aQ 1 -bQ 1 2 -bQ 1 Q 2, MR 1 = a-2bQ 1 -bQ 2. MR 1 =MC 1 implies a-2bQ 1 -bQ 2 =c, that is, Q 1 = (a-c)/2b - 0.5Q 2

13. Oligopolies Repeat the same exercise for firm 2 and you should obtain Q 2 = (a-c)/2b - 0.5Q 1 Please keep in mind that this is a game, where the set of players is {Firm 1, Firm 2 }, the set of strategies for Firm i is  + (Q i  [0,  )) and that the payoff vectors are given by the profit functions. Please plot the Best-Response correspondences in (Q 1,Q 2 ) space, while keeping in mind that firms cannot produce negative outputs. The intersection of the two curves defines the Cournot-Nash equilibrium.

14. Oligopolies You can determine this equilibrium algebraically, by simultaneously solving both equations Q 1 = (a-c)/2b - 0.5Q 2 and Q 2 = (a-c)/2b - 0.5Q 1 If all goes well, you should get Q1=Q2=(a-c)/3b, thus Q=(2(a- c))/3b and p=(1/3)a+(2/3)c. As an exercise, please also calculate the firms’ profits at the Cournot NE. Show that this is indeed a NE in the sense that no firm has an incentive to produce more or less. Start with a numerical example if it makes your life easier (say a=100, b=1, c=20).

15. Oligopolies Our final application is the Stackelberg Duopoly, an extension of the Cournot Duopoly to a dynamic two-period framework. I.e., the two firms choose their strategies (Q i ) sequentially, with Firm 1 moving first (the leader) and Firm 2 moving last (the follower). Using backwards induction, we first solve the problem for Firm 2. Using the same approach as in the Cournot game, we find that Q 2 = (a-c)/2b - 0.5Q 1

16. Oligopolies Given common knowledge of payoffs and common knowledge of rationality, Firm 1 correctly anticipates Firm2’s best response and can embed it into its own profit maximization: ∏ 1 = pQ 1 -cQ 1 = (a-c-bQ)Q 1 = (a-c-b(Q 1 +Q 2 ))Q 1 = (a-c-b(Q 1 + (a-c)/2b - 0.5Q 1 ))Q 1 = (a-c- bQ 1 - (a-c)/2 + 0.5bQ 1 )Q 1 = ((a-c)/2 - 0.5bQ 1 )Q 1 The first-order condition for an optimum is thus ((a-c)/2 - 0.5bQ 1 ) -0.5bQ 1 =0, that is: Q 1 =(a-c)/2b We can now plug this result back into Firm 2 ’s best response to get Q 2 =(a-c)/4b

17. Oligopolies As an exercise, please calculate the firms’ profits. And be sure to understand the paradox of first-mover advantage! (The follower has the luxury of waiting and observing the leader’s strategy and can then optimize; nevertheless, the leader makes twice as much profit. How come?) Can you think of other “games” where the first mover has an advantage? Can you think of games where it is better to move last? Please remember to email me at microprof@gmail.com with any questions.microprof@gmail.com