Microeconomics › Section J · Monopolistic competition and oligopoly › Topic J.5
Game theory applications
Game theory leaves the textbook far behind: it prices advertising slots at auction, predicts when climate promises hold, explains why fairness makes people cut off their own payoffs, and asks whether pricing algorithms can learn to collude on their own. This topic goes beyond the book - into the laboratory, the auction house and the atmosphere.
Where the games go next
This topic follows game theory out of the classroom. In the laboratory, public-goods and bargaining games test whether the self-interest assumption predicts real people. At auction, game design decides whether governments and advertisers raise billions or give assets away. In the atmosphere, the prisoners' dilemma appears at planetary scale with no enforcer. And in modern markets, pricing algorithms are playing repeated games at machine speed. The tools of the last two topics travel to all of it.
Public goods games in the laboratory
A standard experiment: each participant gets tokens to divide between a private account (kept) and a shared pot (multiplied and split equally among all). Everyone earns most if all contribute; each individual earns most by contributing nothing and free-riding. It is the prisoners' dilemma with many players. The findings are a lesson in themselves: real people typically contribute about half at first - more than pure self-interest predicts - but cooperation decays in later rounds unless players can punish free-riders, which restores it. People are neither saints nor machines; institutions that make cooperation enforceable do the rest.
Worked example: the free-rider's arithmetic
Five players each hold 10 tokens. Contributions to the pot are doubled, then split equally among all five. 1. All contribute: the pot holds 5 x 10 x 2 = 100 tokens; each receives 100 / 5 = 20. 2. One player free-rides while the rest contribute 10: the pot holds 4 x 10 x 2 = 80; each receives 80 / 5 = 16. 3. Totals: contributors end with 16; the free-rider ends with the kept 10 plus 16 = 26. 4. The temptation is exact - if the others contribute, free-riding pays 26 against the 20 from contributing, a gain of 6 - yet if all free-ride, everyone ends with the original 10. 5. The multiplier means full contribution (20 each) beats universal free-riding (10 each) for everyone: the group optimum is real, and so is the individual incentive to defect. Assumptions: one-shot play, no punishment.
The ultimatum game: fairness at work
One player - the proposer - is given a sum, say $20, and offers a split. The responder accepts (both get the split) or rejects (both get nothing). Pure self-interest predicts a token offer, accepted - something beats nothing. Experiments consistently find otherwise: proposers commonly offer 40-50%, and responders frequently reject offers below about a fifth, walking away from money to punish an insult. The finding is not that people are irrational, but that payoffs include fairness: the utility of decent treatment (and of punishing its absence) sits in the payoff matrix alongside dollars - a correction to the model's assumption, not a refutation of its method.
Auctions: designing the game
An auction is a game whose rules the seller designs - and the design changes behaviour. In the English (ascending) auction, prices rise until one bidder remains; each bidder can safely stay in until the price passes their value. In the sealed-bid second-price (Vickrey) auction, the highest bidder wins but pays the SECOND-highest bid - a twist that makes bidding one's true value the best strategy, as the next card's arithmetic shows. Online advertising slots, spectrum licences and government assets are routinely sold with such designed mechanisms, and the design has raised substantial sums for treasuries - applied game theory as revenue policy.
Worked example: why truth wins a second-price auction
A bidder values a licence at $300 in a second-price auction. 1. Rival bids $240: bidding $300 wins and pays $240 - a gain of $60. Bidding $250 instead would have won at the same $240 price; bidding higher adds nothing. 2. Rival bids $280: bidding $300 wins and pays $280, gaining $20. Shading to $250 would lose a trade worth $20. 3. Rival bids $320: bidding $300 loses - correctly, since winning would mean paying $320 for a $300 licence. Overbidding would have won the auction and lost money. 4. Conclusion: whatever rivals do, bidding exactly $300 never hurts and sometimes helps - truthful bidding is weakly dominant. Assumptions: private values; the winner pays the second price.
The winner's curse
Second-price logic relies on each bidder knowing their OWN value. But some auctions sell items of common but UNCERTAIN value - oil drilling rights, broadcasting licences - worth the same to everyone, only nobody knows the number. Bids are then estimates, and the highest bid comes from the most optimistic estimator: the winner is systematically the bidder who overvalued the item. That is the winner's curse, and sophisticated bidders defend by shading bids below their estimates - bidding not what they think it is worth, but what it is worth GIVEN that they win. Auction design and experience both soften the curse.
Climate agreements: the dilemma at planetary scale
Reducing emissions is a global public good: every country enjoys the benefit of a stable climate, but each pays the full cost of its own cuts. The structure is the prisoners' dilemma with nearly two hundred players - each is tempted to free-ride, and a world of free-riders produces the worst joint outcome. The repeated-game tools explain what agreements like the Paris Agreement attempt: with no global enforcer, they work through pledges, transparency reviews, and the value of reputation among nations who will meet again. Reputation-based cooperation is real but fragile: the United States' second withdrawal from the agreement took effect in January 2026.
Bargaining: outside options and patience
Bargaining - wages, house prices, treaties - is a sequential game with two special levers. First, outside options: what each side gets if talks fail anchors what they will accept, so improving your alternative - another job offer, another buyer - is often worth more than clever tactics. Second, patience: the side less pressed for time can credibly hold out, and much of bargaining power is simply the ability to wait. Strikes and deadlines are costly demonstrations of resolve. The prescription: before the negotiation, invest in your alternatives and your clock.
Algorithms learning to collude
Online sellers increasingly set prices with algorithms that react to rivals' prices in seconds. Research findings suggest that learning algorithms can sometimes settle at elevated prices with no instruction to do so - emergent tacit collusion, resembling a repeated game in which punishment for undercutting arrives instantly and mechanically. The policy puzzle is real: antitrust law targets agreements, and no one agreed to anything - the machines simply learned. Competition authorities are studying how existing law should treat strategic behaviour with no strategist.
Common mistakes to avoid
First, reading experimental rejections of self-interest as refutations of game theory: the finding is that payoffs include fairness, which the framework can absorb - update the payoffs, not the method. Second, assuming auctions are always won by the highest valuation: with common values, the winner is often the most over-optimistic. Third, forgetting that second-price truthfulness assumes private values - with uncertain common values, shade your bid. Fourth, treating climate agreements as failed enforcement: they are reputation-based repeated games, fragile by structure, not by accident. Fifth, calling algorithmic price alignment collusion without evidence - the research is suggestive and young, and the law has not settled.
Key takeaways
Game theory's applications stretch from the lab to the planet. Public-goods experiments show people contributing about half, decaying without punishment - the free-rider's arithmetic (26 tokens against 16) drives it - and punishment restores cooperation. The ultimatum game shows fairness in payoffs: 40-50% offers and costly rejections. Second-price auctions make truthful bidding weakly dominant (the $300 bidder wins at $240); common-value auctions carry the winner's curse. Climate is the prisoners' dilemma at scale, held up by pledge, transparency and reputation. Bargaining rewards outside options and patience, and pricing algorithms suggest cooperation can be learned with no one agreeing at all.
Flashcards
Flashcard
Contribute to a shared, multiplied pot or keep tokens privately
Test your understanding
Question 1 of 5
In the five-player public goods game (10 tokens each, contributions doubled, split equally), what does a lone free-rider end with while contributors get 16?
Question 2 of 5
In the ultimatum game, what does standard experimental evidence find, in broad terms?
Question 3 of 5
Why is bidding your true value the best strategy in a second-price auction?
Question 4 of 5
What is the winner's curse in common-value auctions like oil drilling rights?
Question 5 of 5
Why are climate agreements structurally fragile in game-theory terms?
