When does search among locally correlated alternatives admit an index policy?
The answer depends on the flexibility of local experimentation: with fixed local protocols, search is indexable, while greater flexibility can destroy indexability. The results are applied to study how optimal search evolves in R&D and career mobility.
Can gradual learning help to implement outcomes that cannot be implemented when information is revealed from the start?
Yes. Gradual information can relax incentives by progressively restricting the outcomes available as the agent learns. Priors under which a social choice function is implementable are recursively characterized.
When does insurance increase health expenditure?
The interaction of adverse selection and hidden action is necessary for insurance to raise health expenditure, and whether it does so depends on how risk aversion changes with wealth. Optimal menus feature terminal pooling without requiring shape restrictions on the type distribution.