Screening Mechanisms¶
Why It Matters¶
Direct asking about type fails (everyone claims to be high quality). Testing everyone is expensive. Screening mechanisms extract information cheaply by letting people reveal their type through choices. The key insight: different types have different costs for certain actions, so you can design choices where only the high type finds the high-commitment option worthwhile.
The Key Move¶
Design multiple options with different effort/reward bundles. Structure them so high types prefer high-commitment (high effort, high reward) and low types prefer low-commitment (low effort, low reward). The differential cost of effort creates separation. Then observe which option each person chooses - the choice reveals their type. Check: if a low type mimicked the high-type choice, would they be worse off? If yes, you've achieved separation.
Classic Application¶
Insurance contracts with deductibles. High-risk customers have higher expected claims, so they value comprehensive coverage more. Offer two contracts: low deductible (high premium) and high deductible (low premium). High-risk types choose low deductible despite high premium (they expect to use it). Low-risk types choose high deductible (they're willing to self-insure). The insurance company learns risk type from contract choice without medical testing.
Surprising Application¶
Hiring processes with unpaid "homework" assignments. Candidates who are genuinely interested and capable will invest time in a quality submission. Candidates who are just testing the market or not seriously interested won't. The assignment screens for commitment and ability simultaneously. But beware: this only works if the homework cost is higher for low types than high types. If both types find it equally burdensome, you've just added friction without gaining information.
Failure Modes¶
Creating "screens" that equally burden all types (no separation achieved). Ignoring that screening is costly for both parties (candidates hate homework, companies hate reviewing it - sometimes just make a guess). Over-screening (extracting information you don't actually use in decisions). Forgetting that screens can be gamed once they're known (candidates learn to mimic high-type signals). Screening for the wrong variable (what you screen for may not correlate with what you actually care about).
Go Deeper¶
Rothschild and Stiglitz (1976), "Equilibrium in Competitive Insurance Markets." Spence (1973), "Job Market Signaling." Salanie, "The Economics of Contracts," Chapter 2.