Is "Moral Hazard" Inefficient? The Policy Implications of a New Theory