From the folk theorem of game theory, proved by Robert J. Aumann (1959) and James W. Friedman (A Non-cooperative Equilibrium for Supergames, 1971), extended by Drew Fudenberg and Eric Maskin (1986), and set out by Martin J. Osborne (An Introduction to Game Theory, 2004).
Shows how cooperation can hold between self-interested parties who expect to deal again, because the threat of future punishment outweighs a one-time gain from cheating.
A parts supplier and its customer: full quality holds while the future is worth more than one year of cutting corners
The supplier earns $10M a year making parts to full specification. Cutting corners on materials pays a bigger margin once, until the customer finds out and moves most of its volume to a second source for good. The point sets that one-time gain against the profit the future loses; above the line, cheating does not pay and quality holds. Lower the chance the relationship runs another year and watch the point fall through the line. Illustrative.
Where the relationship standsQuality holds: the business the future would lose outweighs the one good year.Quality breaks: one year of cutting corners is worth more than everything it costs later.
An example
A maker of electric motors bought its copper windings from one supplier and worried, after an industry scandal, that the supplier might one day substitute cheaper alloy in a batch. Full incoming inspection would cost $1.2 million a year.
The procurement head wrote the supplier's choice as a repeated game. Each year, making to specification earned the supplier about $10 million on the account. Cutting corners for a year might add $8 million before detection. After that the motor maker would qualify a second source and move most of the volume, leaving the supplier perhaps $4 million a year. With the contract renewing every year and both sides planning a decade ahead, she judged a 60 percent chance at least that the business continued in any given year. The future loss came to about $9 million against the $8 million gain: cooperation held, but narrowly.
She skipped inspection and widened the margin. A qualified second source took 10 percent of volume, which made the threat credible, and a three-year volume commitment raised the value of the future to the supplier.
Why it works
In a single encounter, a self-interested party with a chance to gain at the other's expense usually takes it. A one-shot analysis therefore predicts cheating everywhere, which does not match how long business relationships behave. The repeated game explains the difference: when the parties expect to deal again, each can make future cooperation depend on present behavior, and the gain from one act of cheating has to be set against all the future business it throws away.
The result turns on a comparison anyone can make: the one-time gain against the discounted value of the future lost to punishment. That comparison also says what to work on. Make the future longer and more certain, make the punishment real and quick, and keep any single act of cheating small, and cooperation holds without anyone having to trust anyone.
The thinking behind it
James W. Friedman James W. Friedman, an economist and game theorist, proved an early folk theorem in A Non-cooperative Equilibrium for Supergames (1971). Any outcome better for every player than the one-shot equilibrium can be sustained in an infinitely repeated game when players are patient enough and threaten to revert to that equilibrium after a deviation. It is the logic behind the motor maker's threat of a second source.
Robert J. Aumann Robert J. Aumann, a game theorist, gave an early formal statement of how cooperation can be sustained in infinitely repeated games in a 1959 paper, Acceptable Points in General Cooperative n-Person Games, one of the results later grouped under the name folk theorems. His work showed that repetition alone can hold together outcomes no single encounter would support. He shared the 2005 Nobel Memorial Prize in Economic Sciences for work on conflict and cooperation.
Drew Fudenberg Drew Fudenberg and Eric Maskin, in The Folk Theorem in Repeated Games with Discounting or with Incomplete Information (1986), proved that patient players, who value the future highly, can sustain almost any outcome that gives each of them more than their minimum, even when future payoffs are discounted. Martin J. Osborne sets out these results for students in An Introduction to Game Theory (2004).
How to run it
- Write the payoffs for one round. For each side, set out what it earns if both cooperate, if both defect, and if one cheats on the other.
- Pick the cooperative outcome to sustain. Name the behavior that keeps both better off, such as full specification, prompt payment, or shared information.
- Define the response to a deviation. State what happens after cheating: lost volume, a second source, an ended contract, and how fast it follows.
- Estimate how much the future counts. Judge the chance the relationship continues each round and how far ahead each side plans.
- Compare the one-time gain with the future lost. Cooperation holds when the discounted value of the punishment exceeds the gain from one round of cheating.
- Widen the margin where it is thin. Lengthen commitments, make the response credible and quick, and break large one-time temptations into smaller ones.
Where it goes wrong
The common failure is relying on a punishment the other side does not believe. A buyer threatens to switch suppliers but has never qualified another, so the threat adds nothing to the future loss. It shows up as cheating that persists after warnings. Make the response real before it is needed, for example a small share of volume with a second source.
A second failure is forgetting the end of the game. Near a known final round, such as a contract's last year or a business being sold, the future shrinks to nothing and cheating pays. Watch relationships with a visible end date most closely. A third is noise: honest mistakes read as cheating trigger punishments that destroy the cooperation they were meant to protect.
When to use something else
- Tit for Tat. When the parties need a simple, readable rule for playing the repeated game month by month.
- Nash Equilibrium. When the interaction happens once, and the question is which outcome neither side would leave on its own.
- Mechanism Design. When the work is designing the rules or contract so that honest behavior is each side's best move from the start.
Where it sits in SOLVE eX
Go to the source
- Acceptable Points in General Cooperative n-Person Games, Contributions to the Theory of Games IV, 1959
- A Non-cooperative Equilibrium for Supergames, 1971
- The Folk Theorem in Repeated Games with Discounting or with Incomplete Information, Econometrica 54, 1986
- An Introduction to Game Theory, chapter 14, 2004