Controls
Presets
Parameters
Quality
Camera
Model checks
The science
The market. Everyone has wealth W = 100 and may lose L with probability p. A share λ are high-risk (pH), the rest low-risk (pL), and the insurer cannot tell them apart. A contract (premium q, cover a) leaves consumption W - q without an accident and W - L - q + a with one, valued with CRRA utility. Competition forces every contract to break even (Rothschild and Stiglitz, Quarterly Journal of Economics 1976).
Screening. With full information both types would buy full cover at their fair price p L (Mossin's theorem). With hidden types the high-risk type would pretend to be low-risk, so the low-risk contract must be rationed: the cover aL at the fair price pL aL is cut until the high-risk type is exactly indifferent (the incentive constraint binds), while the high-risk type keeps full cover. The low-risk type pays for the information problem with a lower certainty equivalent.
Existence. The separating pair is an equilibrium only if no pooling contract priced at the average fair price pbar = λpH + (1-λ)pL is preferred by the low-risk type. If the share of high-risk people is small, pbar is close to pL and such a deviation exists; then there is no pure-strategy equilibrium at all. The threshold λ* is found numerically. A single pooled full-cover price is also fragile: when pbar exceeds the most the low-risk type will pay, they leave, the average risk rises and the market unravels (the adverse-selection spiral).
Simplifications. Two types, one accident size, one period, no moral hazard (cover does not change behaviour, which the original topic mixed in and the existing health-insurance moral-hazard lab covers), no administrative costs, competitive zero-profit insurers and the Rothschild-Stiglitz pure-strategy notion (Wilson, Riley and Miyazaki define other equilibria). The default numbers are illustrative. Checks: both incentive constraints; Mossin full cover by numerical maximisation; zero profit by Monte Carlo (4.5 sigma); the best pooling deviation by derivative bisection versus a 6001-point grid; the low-risk willingness to pay in closed form; and the sign change of the separating-versus-pooling gap at λ*.