Game Theory · GT-21
The cost of a signal is the point — not what it superficially says.
When one party has private information, they can credibly reveal it to another party by taking a costly action that would be irrational for someone without that information to fake. The cost of the signal — not its content — is what makes it trustworthy.
Formalized by Michael Spence in his 1973 paper "Job Market Signaling" (Quarterly Journal of Economics), part of the body of work that earned him a share of the 2001 Nobel Memorial Prize in Economic Sciences alongside George Akerlof and Joseph Stiglitz.
The Mechanism
Two job candidates, one signal (a degree) — toggle the cost gap
When the cost gap between types is wide, a separating equilibrium emerges — only genuinely high-quality candidates find the credential worth its cost, so the signal is informative. Employers can infer quality from the degree itself, without ever observing ability directly.
01 · THE SINGLE-CROSSING PROPERTY
The signal must be cheaper for the truth to bother sending
A signal only works as information if it costs meaningfully less (in effort, risk, or resources) for the high-quality type to produce than for the low-quality type. If the cost were identical for both, both would send it, and the receiver would learn nothing from seeing it — the signal has to differentiate on its own cost structure, not its face value.
02 · SEPARATING vs. POOLING EQUILIBRIA
A wide gap separates; a narrow gap pools
When the cost difference between types is large, a "separating equilibrium" emerges — only the high type finds the signal worthwhile, so sending it reliably reveals quality. When the gap narrows enough that low types can also afford to fake it, the system collapses into a "pooling equilibrium," where everyone sends the same signal and it stops conveying any information at all.
03 · CONTENT IS INCIDENTAL
The signal's content can be irrelevant to the underlying quality
Spence's original insight was deliberately provocative: a college degree can function as a valid signal of employability even if almost nothing taught in the classroom is directly used on the job. What the credential proves is that the holder could pay the cost — in time, tuition, and difficulty — that a less capable or less committed candidate would have found too expensive to sustain.
Where It Fails / Inversion
The trap: credential inflation collapses separation into pooling
Signaling only works as long as the cost gap between types stays wide. When a credential becomes common enough — more people can afford the years and tuition regardless of underlying ability — the low-cost curve creeps toward the high-cost curve, the separating zone shrinks, and eventually the signal pools: everyone has the degree, so having it stops distinguishing anyone. This is the mechanism behind credential inflation, and it explains why the market response is usually to demand a costlier signal (a master's degree where a bachelor's once sufficed), not to abandon signaling altogether.
A second failure mode: mistaking a genuinely costless claim for a signal. If a candidate can simply assert a quality with zero cost or risk of being caught wrong, that's not a signal at all — it's cheap talk, and it carries none of signaling theory's information value.
How To Use It
Worked example · the peacock's tail
Biologist Amotz Zahavi's "handicap principle" applies the identical logic to evolution: a peacock's elaborate tail is metabolically expensive and makes escaping predators harder — exactly the properties that make it a credible signal of genetic fitness. A weak or diseased peacock literally cannot afford to grow and carry such a tail and survive. A healthy one can. Peahens that select for elaborate tails are, without knowing any biology, correctly reading a costly signal that cannot be cheaply faked.
The parallel to job-market signaling is exact: the tail's beauty is incidental (like a degree's specific coursework); its cost is the entire message. Both cases show the same underlying rule — you cannot design a meaningful signal by making it cheap and easy; you have to make it something only the genuinely qualified can afford.
How to use it
Before trusting any claim of quality — a resume line, a vendor's certification, a competitor's public commitment — ask what it actually cost the sender to produce, and whether someone without the claimed quality could have sent the identical signal just as easily. If the answer is yes, you're looking at cheap talk dressed up as a signal, not real information.
See Also