Microeconomics › Section J · Monopolistic competition and oligopoly › Topic J.4
Repeated and sequential games
One-shot dilemmas reward treachery; but most real rivals meet again, and most real moves happen in order. Repetition lets yesterday's behaviour be punished tomorrow, and sequence lets earlier moves shape later ones - two features that decide whether cooperation survives and whether threats are believed.
The shadow of the future
The prisoners' dilemma punished cooperation because the game ended: confess, part, done. Play the same game repeatedly with no known end and a new force appears - tomorrow. A player who cheats this round can be punished the next; one who cooperates can be rewarded. The future casts a shadow over the present, and under that shadow, cooperation that collapsed in the one-shot game can become self-enforcing: no contract, no cartel meeting, just each firm understanding that treachery buys one good round and a bad forever after. This is the standard story of how long-lasting oligopolies avoid price wars without signing anything illegal.
Tit-for-tat
How should a repeated player actually behave? In influential computer tournaments run by political scientist Robert Axelrod in the 1980s, the strongest performer was one of the simplest strategies ever submitted: tit-for-tat. Open by cooperating; thereafter copy whatever the opponent did last round. Its success came from four properties: it is nice (never cheats first), retaliatory (punishes cheating promptly), forgiving (returns to cooperation the moment the opponent does), and clear (the opponent can see exactly how their moves map to consequences). The lesson travels beyond machines: predictable, proportionate responses sustain cooperation better than either naivety or vengeance.
Triggers and the value of the future
Punishment strategies come in strengths. Tit-for-tat punishes for one round. The grim trigger punishes forever: cooperate until the opponent cheats once, then never again. Both raise the price of cheating by sacrificing future gains - and whether that price deters depends on how much players value the future. Someone indifferent to tomorrow cannot be threatened with tomorrow. Economists capture this with a discount factor: how much a future payoff is worth today. The general result: the more the players value future payoffs - and the longer the game plausibly lasts - the easier cooperation is to sustain.
Worked example: the arithmetic of staying honest
Two firms collude implicitly: cooperation pays each 40 million dollars a year forever; cheating secretly earns 70 million this year, then triggers a price war worth only 10 million a year thereafter. Let d be the value placed on a dollar next year compared with today. 1. Value of cooperating forever: 40/(1 - d). 2. Value of cheating: 70 + 10d/(1 - d). 3. Cooperation wins when 40/(1 - d) exceeds 70 + 10d/(1 - d), which simplifies to 60d > 30, that is d > 0.5. 4. Check: at d = 0.4, cheating (76.7) beats cooperating (66.7); at d = 0.6, cooperating (100) beats cheating (85). 5. Read it: as long as a future dollar is worth more than half a present one, the ongoing relationship outvalues the one-time treachery. Assumptions: payoffs as stated, grim trigger carried out, no known end.
Unravelling: the known last round
One clause in that result does heavy lifting: no known end date. Give the game a definite final round and cooperation unwinds, by the same backward logic everywhere in this topic. In the last round there is no tomorrow to fear, so each player cheats - the one-shot dilemma again. But then the second-to-last round is now effectively the last round in which behaviour matters, so each cheats there too; and the reasoning rolls backward until cheating starts from round one. Real approximations appear where relationships visibly end: end-of-term markets, a firm's announced exit, a departing employee's last week. In theory, cooperation needs a future that is genuinely open-ended.
Sequential games and game trees
In a sequential game, players move in turn, and each sees what earlier movers did - entry decisions, price announcements, contract offers. The picture is a game TREE: dots for decision points, branches for choices, payoffs at the ends. The solution method is backward induction: start at the final decision points and find each mover's best choice there; then ask what the mover just before would do knowing that; and so on to the root. Backward induction is the discipline of taking the LAST decision seriously first - and it exposes which earlier moves, threats and promises are real.
Worked example: deterring entry with a threat
An incumbent monopoly earns 100 million dollars, undisturbed. A rival considers entering; the incumbent publicly threatens a price war. If entry happens: fighting hurts both - the incumbent earns 30 million and the rival loses 10 million; accommodating shares the market - incumbent 60 million, rival 25 million. 1. Backward induction starts at the last move: once the rival has entered, does the incumbent fight or accommodate? 60 beats 30: accommodate. 2. The rival, foreseeing this, compares entering (25 million) with staying out (0): it enters. 3. The threat failed because it was not credible - a threat the threatener would not carry out is ignored by a rational rival. 4. The incumbent's real remedy is changing the payoffs of step 1 - commitment, the next card. 5. Contrast: had fighting been genuinely profitable, the same threat would deter entry without ever being used. Assumptions: payoffs as stated; both players rational and informed.
Commitment: binding your own hands
How can the incumbent make its threat credible? By changing its own future payoffs in advance - by commitment. Build excess capacity that makes fighting cheap; sign long-term contracts whose penalties make backing down costly; invest in a public reputation and stake it on the promise. Each move deliberately limits the firm's own options, and precisely because of that, it changes the rival's: the threat stops being cheap talk because carrying it out is now the firm's own best move. Commitment devices are everywhere once you look: non-refundable deposits, scorched-earth provisions, ships burned on arrival. Self-restriction, used strategically, is power - the paradox that makes sequential games surprising.
Who moves first?
Sequence creates advantages - sometimes for the early mover, sometimes for the late one. First-mover advantage: the firm that builds the plant first or captures network users first can set the standard, lock up inputs, and shape the game others must play - the commitment logic working in its favour. Second-mover advantage: the firm that watches can learn from the pioneer's mistakes at the pioneer's expense, entering with a better product once the market is proven. Which advantage dominates depends on the game: when commitment and network effects dominate, move first; when learning and adaptation dominate, move second. Neither is a law - both are answers to the structure of the specific tree.
Common mistakes to avoid
First, believing threats just because they are made - backward induction asks whether the threatener would ACTUALLY carry them out. Second, ignoring the discount factor: the value of the future, not good intentions, sustains cooperation (the arithmetic held only for \(d > 0.5\)). Third, expecting tit-for-tat to be the last word - it is a benchmark that performed well in a specific tournament, not a universal prescription. Fourth, assuming repetition always helps: with a known end date, repetition unravels to one-shot behaviour. Fifth, treating first-mover advantage as automatic - in some games watching wins.
Key takeaways
Repetition puts the future in the payoff: cheating this round invites punishment in the next, and Axelrod's 1980s tournaments crowned tit-for-tat - nice, retaliatory, forgiving, clear - as a simple way cooperation defends itself. The arithmetic: with cooperation worth $40m a year and cheating $70m once plus $10m after grim-trigger punishment, cooperation wins whenever a future dollar is worth more than half a present one (\(d > 0.5\)). Known end dates unravel cooperation backward, since the final round is a one-shot dilemma. Sequential games read as trees and solve by backward induction: the incumbent's price-war threat fails because once entry happens, accommodating ($60m) beats fighting ($30m) - so the rival enters, and only credible commitments (excess capacity, contracts, reputation) can deter. First or second can be the better seat; the tree decides.
Flashcards
Flashcard
The same game played again, with no known end
Test your understanding
Question 1 of 5
In the cooperation example, cooperating pays $40m a year forever, while cheating pays $70m once and then $10m a year after punishment. When does cooperation win?
Question 2 of 5
Why does a known final round destroy cooperation in a repeated prisoners' dilemma?
Question 3 of 5
An incumbent threatens a price war if a rival enters, but fighting would leave it with $30m while accommodating leaves $60m. What does backward induction predict?
Question 4 of 5
How can a firm make an otherwise empty threat credible?
Question 5 of 5
What made tit-for-tat successful in Axelrod's tournaments?
