Week 5: Screening Games

How the uninformed side of a market can force information out of the informed side

20 March 2028

Open the slides

Signaling puts the burden on the informed party to reveal itself; screening flips it — the uninformed party designs a mechanism that separates types by offering a menu only some would rationally choose. An insurer that offers a low-premium, high-deductible plan alongside a high-premium, low-deductible one is screening for risk without ever asking anyone’s risk category directly.

Assignment 1 is due this week: having spent three weeks learning to read signals, you now build one, and defend which of its claims would survive being screened.

Outline#

• separating equilibria via self-selection, not disclosure
• designing a menu that only reveals type if people choose honestly
• what a screening test would need to look like to actually work on a dating profile, and why most “verification” features don’t
• what a working menu still costs the honest type, even when it works

One thing this framework won’t explain#

A menu that separates types is, by this model’s own standard, a success — the honest type chose one option, the dishonest type chose another, and the mechanism did its job. What the model has no term for is the honest type’s cost of proving it: the low-risk driver still pays the deductible. The person who’s actually trustworthy still has to sit through whatever the menu asks of them, and the model treats that cost as fully justified by the information it produces, without ever asking whether it was fair to make anyone pay it.

That’s the same shape as Weeks 1, 3, and 4’s gaps, from a different angle: the model can certify that a mechanism works and still say nothing about what it costs the people it works on. Week 12 is where all four of these come back at once.

Further reading#

• Rothschild, M., & Stiglitz, J. (1976). “Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information.” Quarterly Journal of Economics, 90(4), 629–649. The paper this week’s central example is drawn from — the two-plan insurance menu above is Rothschild and Stiglitz’s own model, not an analogy built on top of it.