Originated by A. Michael Spence (Job Market Signaling, 1973), with a standard treatment by Martin J. Osborne (An Introduction to Game Theory, 2004).
Explains how one party conveys hidden quality through an action too costly for a weaker party to copy, and how the other side reads such signals.
A two-year warranty tells buyers about quality only while it costs a weak maker more than it earns
Buyers cannot see which commercial dishwasher maker builds reliable machines, so they pay extra for a warranty. Set what the warranty costs each maker in expected repairs and the premium buyers pay: below the line only the reliable maker gains from offering it and the warranty separates the two; above it the weak maker copies it and the signal means nothing. Illustrative.
What the warranty tells buyersPooling: copying pays for the unreliable maker too, so a warranty no longer tells buyers which machine is reliable.Separating: only the reliable maker gains from the warranty, so buyers can read it as a sign of quality.
An example
A young maker of commercial dishwashers was losing restaurant-chain bids to older brands. Its machines failed far less often, about one in twenty-five over two years against roughly one in four for the cheapest rivals, but every sales deck said reliable, and buyers had learned to ignore the word.
The chief operating officer looked for something a weak rival could not afford to copy. A two-year parts and labor warranty would cost the company about $15 a machine in expected repairs, given its failure rate. For a rival with its competitors' record, the same promise would cost about $90. Buyers told the sales team they would pay around $50 more for a machine with that cover. So the warranty earned the company $35 a machine and would lose a weak rival $40. The company launched it with no exclusions, and two chains' procurement teams added a two-year full warranty to their bid requirements. The weakest rival stopped bidding for those contracts within a year.
Why it works
When one side knows something the other cannot check, claims are cheap and therefore worthless: every seller says its product is good. What carries information is an action that costs a low-quality sender more than a high-quality one, so much more that only the high-quality sender finds it worth taking. Then the action sorts senders into types, and the receiver can read quality from the choice alone.
The signal need not be useful in itself. A warranty that a reliable maker rarely pays out on still works, because its value lies in what it would cost someone else. The same logic reads incoming signals: ask what the signal costs to send for each type of sender. If a weak sender could send it cheaply, it tells the receiver nothing, however impressive it looks.
The thinking behind it
A. Michael Spence A. Michael Spence, a Nobel laureate economist, set out signaling theory in Job Market Signaling (Quarterly Journal of Economics, 1973). Employers cannot observe a candidate's productivity, so candidates reveal it through a costly action, such as education, that is cheaper for more productive people. The signal then separates the types even if it adds nothing to productivity, which is the condition the figure on this page tests for a warranty.
Martin J. Osborne Martin J. Osborne, an economist at the University of Toronto, gives a standard treatment in An Introduction to Game Theory (Oxford University Press, 2004), an undergraduate textbook that covers strategic and extensive games, repeated games, and games of incomplete information. Signaling games sit among the last: one player knows a fact the other cannot see, chooses an action the other can see, and the other reads the action to infer the fact.
How to run it
- Name the hidden fact. State what one side knows and the other cannot see: quality, commitment, financial strength, or skill.
- List the signals available. Write down every action the informed side could take that the other side can observe and verify.
- Price each signal for each type. Estimate what it costs a strong sender and a weak sender to send it, in money, risk, or time.
- Check the gap. Keep only signals whose cost to a weak sender exceeds what the signal would earn it.
- Choose the cheapest signal that separates. Among those, pick the one that costs the strong sender least.
- Make it verifiable. Remove the exclusions and fine print a weak sender could hide behind, or the cost gap disappears.
- Read incoming signals the same way. When a counterpart signals, ask what it would have cost a weaker version of them to send the same thing.
Where it goes wrong
The first failure is a signal that is costly for everyone. A glossy certification that both strong and weak suppliers can buy proves only that they paid for it. Spot it by pricing the signal for a weak sender; if they can afford it, find another.
The second is a signal that weak senders can fake at low cost: a warranty full of exclusions, a guarantee from a company too small to honor it. Buyers learn quickly, and the signal stops working for everyone. Keep it simple and enforceable. The third is overpaying. Once the signal separates, spending more on it buys nothing; trim it to the least that still sorts the types.
When to use something else
- Game Theory. When both sides' information is shared and the question is how each will respond to the other's moves.
- Moral Hazard Analysis. When the hidden thing is what someone will do after the deal, more than what they already are.
- Skin in the Game. When the aim is making someone bear the cost of their own claims, inside a relationship that will continue.
Where it sits in SOLVE eX
Go to the source
- Job Market Signaling, Quarterly Journal of Economics, 1973
- An Introduction to Game Theory, Oxford University Press, 2004