When informed parties (high-quality sellers, productive workers) can take costly actions observable to uninformed parties (buyers, employers), they can signal their type by choosing separating actions that low-type parties don't mimic. Equilibrium requires the separating action to be incentive-compatible: beneficial for high-types, expensive for low-types. Education signals worker productivity; warranties signal product quality. Excessive signaling (relative to information value) is socially wasteful.
Your prerequisite on information asymmetry introduced the adverse selection problem: when sellers know quality and buyers don't, the market can unravel. Low-quality goods drive out high-quality goods because buyers, unable to distinguish them, are only willing to pay the average price. The result is a market that produces too little high-quality output or collapses entirely — the lemons problem. Signaling is a market-based response to this failure. Rather than waiting for an outside authority to certify quality, high-quality sellers can take an observable, costly action that credibly communicates their type.
The key insight is that the signal must be differentially costly: cheap to take for high types, expensive to take for low types. If both types could afford the signal equally, it would not separate them — any low type would mimic the high type and collect the premium price. The separating equilibrium exists when the cost structure creates a natural wedge. Michael Spence's education model is the canonical example: suppose college education does not increase worker productivity at all. A high-productivity worker can still use a college degree as a signal if completing college is less costly for them (in time, effort, or forgone wages) than for a low-productivity worker. Employers, observing the degree, rationally infer high productivity and pay the premium. Low-productivity workers don't attend college because the wage gain does not justify their higher cost of completing it.
The incentive compatibility conditions formalize this logic. A separating equilibrium requires: (1) the high type prefers to signal over not signaling given the wage premium it earns; (2) the low type prefers not to mimic the high type given the cost of doing so. If condition (2) is violated — if mimicking is too cheap — the equilibrium collapses into a pooling equilibrium where everyone signals and the signal conveys no information. If condition (1) is violated — if signaling is not worth the cost even for high types — no one signals. Other real-world signals include product warranties (costly for low-quality firms that expect many claims), conspicuous consumption (costly for those who cannot sustain high spending), and credentialing in professions.
The social welfare implications are subtle. In the education example, if degrees are pure signals and do not raise productivity, then the entire cost of education is a social waste — it is spent on sorting workers who were already sorted by ability, not on creating new human capital. The resources devoted to signaling (tuition, years of study) are consumed without generating the underlying productivity gains a naive observer might assume. This does not mean signaling always wastes resources; sometimes signals are informative and productive simultaneously. But the analysis reveals that when the private return to a signal exceeds its social return (because it merely redistributes a fixed wage premium rather than creating value), markets tend to over-invest in signaling. The optimal signal from a social standpoint would be as thin a wedge as needed to achieve separation — not thicker.
No topics depend on this one yet.