Game Theory · GT-23
The uninformed party designs a menu of options that forces the informed party to reveal their hidden type by which one they pick.
The mirror image of signaling: instead of the informed party voluntarily sending a costly signal, the uninformed party designs a set of contract options (a 'menu') such that different hidden types will self-select into different choices, revealing private information indirectly through their own optimizing behavior.
Formalized primarily by Michael Spence (screening's twin concept to his own 1973 signaling model) and significantly developed in insurance-market contexts by Michael Rothschild and Joseph Stiglitz's 1976 paper on adverse selection.
The Mechanism
A menu that sorts customers by their own hidden preferences
The insurer never learns any individual customer's true risk directly — but by offering exactly these two contracts, low-risk and high-risk customers separate themselves into different choices purely by acting in their own self-interest, revealing the very information the insurer couldn't observe directly.
01 · SELF-SELECTION DOES THE WORK, NOT DIRECT INQUIRY
The uninformed party never has to ask, or trust an answer
Because customers can't be trusted to honestly self-report their true risk type (they'd have every incentive to claim low risk regardless), a well-designed menu instead makes truthful self-sorting the customers' own best response — each type ends up revealing itself through its choice, not through a claim that could be a lie.
02 · THE MENU MUST SATISFY 'INCENTIVE COMPATIBILITY'
Each type must genuinely prefer its intended option
A screening menu only works if each type is actually better off picking the option designed for it than mimicking another type's choice — this constraint (formally called incentive compatibility in mechanism design) is the central mathematical challenge in designing a working screening contract, and getting it wrong lets the wrong types pick the wrong options.
03 · IT'S EVERYWHERE ONCE YOU LOOK FOR IT
Tiered pricing, deductibles, and volume discounts are all screening
Airlines offering economy vs. business fares, software companies offering free vs. paid tiers, and insurers offering different deductible levels are all running screening menus — deliberately designed option sets that sort customers by hidden willingness-to-pay, risk tolerance, or usage intensity, without ever directly asking or verifying.
Where It Fails / Inversion
Where it fails / inversion
A badly designed menu (one that fails incentive compatibility) lets the wrong types mimic each other, defeating the entire purpose — Rothschild and Stiglitz's 1976 analysis went further and showed something more troubling: in some insurance-market conditions, no stable screening equilibrium exists at all, because any menu that works can be undercut by a competitor offering a slightly better deal that attracts only the low-risk types, unraveling the whole market.
How To Use It
Worked example · SaaS pricing tiers
A software company offering a free tier, a mid-tier, and an enterprise tier isn't guessing at customer segments arbitrarily — it's running a screening mechanism. Price-sensitive individual users self-select into the free tier; growing teams that need collaboration features self-select into mid-tier; large enterprises needing SSO, compliance, and dedicated support self-select into the top tier — each genuinely revealing its type through the choice itself, without the company ever needing to directly verify company size or budget.
How to use it
When you can't directly verify a counterpart's true type (their risk, their budget, their real priority) but need to act differently depending on it, design a menu of options where each type's own self-interest leads it to the choice you'd want it to make — check carefully that each type actually prefers its intended option to mimicking another, or the whole mechanism collapses.
See Also