Statistically efficient demand allocation for general matching markets

Determine a statistically efficient vector of edge-sampling demands for general stable matching markets, rather than merely realizing a prescribed demand vector.

Background

The paper derives an information-load lower bound that incorporates both historical evidence and the endpoint-capacity constraints of matching schedules. On the restricted priority-star market class, Priority-Search achieves the resulting first-order information-load constant, but that achievability result does not extend to arbitrary matching markets. For general markets, the unresolved issue is how to select statistically efficient edge-sampling demands that satisfy the information constraints and matching capacities, rather than simply scheduling a demand vector supplied in advance.