Coherent risk from finite models

Tomasz Kania

Coherent estimators at different sample sizes need not describe the same population risk. We prove that dyadic estimators do so exactly when duplicating every scenario leaves the e…