
Microeconomics — Game Theory: Mixed Strategies, Repetition and Credibility
Microeconomics — An Equilibrium Is a Mutual Best Response, Holding the Other’s Choice Fixed
The Cournot equilibrium of Chapter 8 was already a Nash equilibrium. This chapter generalises the tools of game theory: when there is no equilibrium in pure strategies (mixed strategies), when choices happen in sequence (backward induction), and when the same game is repeated (discount factors). First, note that optimisation by one person and equilibrium among several are different concepts. In optimisation the environment is fixed; in equilibrium each player’s choice is the others’ environment.
1. Nash equilibrium
A Nash equilibrium is a combination of strategies in which no player has an incentive to change strategy, taking the others’ strategies as given. In the prisoner’s dilemma, (confess, confess) is a Nash equilibrium, yet it is worse for both than staying silent. Nash equilibrium is a concept of stability, not of efficiency.
2. Mixed strategies: probabilities that make the rival indifferent
The following game has no equilibrium in pure strategies (each cell shows the row player’s and the column player’s payoffs).
| Column: Left | Column: Right | |
|---|---|---|
| Row: Up | (2, 1) | (0, 3) |
| Row: Down | (1, 2) | (3, 0) |
From any cell someone wants to deviate. At (Up, Left) Column switches to Right; at (Up, Right) Row switches to Down; at (Down, Right) Column switches to Left; at (Down, Left) Row switches to Up.
In a mixed-strategy equilibrium, each player chooses probabilities that make the other player indifferent between their two pure strategies. If the other were not indifferent, they would play only one pure strategy, and then my mixing would no longer be a best response.
- If Column plays Left with probability , Row’s expected payoff is from Up and from Down. They are equal when .
- If Row plays Up with probability , Column’s expected payoff is from Left and from Right. They are equal when .
The equilibrium is (Row: Up with 1/2, Column: Left with 3/4); Row’s expected payoff is and Column’s is . A common mistake is to assume 50% by default or to compute the probabilities from one’s own payoffs. How often a penalty taker shoots left or right is determined not by the taker’s preferences but by the goalkeeper’s ability to save each side.
3. Sequential games and backward induction
When choices are made in sequence, draw the game as a tree and solve it from the end. An entrant decides whether to enter a market held by an incumbent; if it enters, the incumbent decides whether to start a price war or accommodate.
| Entrant's choice | Incumbent's response | Payoffs |
|---|---|---|
| Stay out | — | (0, 100) |
| Enter | Accommodate | (30, 50) |
| Enter | Price war | (−10, 20) |
The incumbent threatens: “If you enter, I will start a price war.” If the entrant believes it, it stays out, and (stay out, price war) is a Nash equilibrium. But once the entrant has actually entered, the incumbent is better off accommodating (50) than fighting (20). The threat is not credible.
Solving by backward induction, the incumbent accommodates at the last stage, and the entrant, knowing this, enters (30 > 0). An equilibrium that is a best response in every subgame is a subgame-perfect equilibrium; it filters out Nash equilibria that rely on non-credible threats.
The way to make a threat credible is commitment. If the incumbent builds large capacity in advance so that, after entry, flooding the market is actually more profitable than accommodating (say, price war 60 > accommodate 50), the threat becomes a genuine best response and entry is deterred. This is why irreversible sunk investments have strategic value — the same logic as the Stackelberg leader in Chapter 8.
4. Repeated games
A game in which players defect when it is played once can sustain cooperation when it is repeated. Suppose cooperating yields 400 won per period, deviating once yields 700 won in that period, and thereafter the rival punishes forever so that each period yields 50 won. With discount factor :
The numerator is the extra 300 won gained by deviating once; the denominator is the 650-won difference between the deviation payoff and the punishment payoff. If the future matters enough, cooperation is an equilibrium. The collusion condition in Chapter 8 is an application of this formula.
Two caveats. First, in a finitely repeated game whose end is known, players defect in the last period for certain, so they also defect in the one before, and backward induction unravels cooperation from the first period. Cooperation needs an uncertain end. Second, in infinitely repeated games a great many outcomes can be equilibria, not just cooperation (the folk theorem). Repetition makes cooperation possible, not guaranteed.
| Statement | Condition for credibility | Tool |
|---|---|---|
| I will retaliate | Retaliation must be a best response at that point | Subgame-perfect equilibrium |
| I will start a price war | The war path must be an equilibrium | Discount factor, observability of deviation |
| I will block entry | Options must be tied by irreversible investment | Sunk capacity, long-term contracts |
Check your understanding
In the table above, when Column plays Left with probability 3/4, what are Row’s expected payoffs from its two strategies? Up gives and Down — equal. Row is indifferent, so mixing 1/2–1/2 is also a best response. If Column played Left with probability 0.9, Up would yield 1.8 and Down 1.2, so Row would play only Up, and Column would then switch to Right, breaking the equilibrium.
References
- Robert Gibbons, Game Theory for Applied Economists, ch. 1–2
- Avinash Dixit and Susan Skeath, Games of Strategy, ch. 6–7, 10
- Hal Varian, Intermediate Microeconomics, ch. 29–30
- MIT OpenCourseWare, 14.12 Economic Applications of Game Theory
Oiyo
Editorial DeskThe OIYO editorial desk researches money, law, lifestyle, and self-understanding topics against primary sources and public statistics. Every piece carries source notes and is reviewed on a regular cycle for accuracy and usefulness.