Proper Scoring Rules for Fractiles
J L Cervera*, J Munoz** · 1996
Abstract In this paper we consider the problem of how an expert should be paid when he/she states the information about the distribution of a random quantity in terms of fractiles, and an observation is subsequently taken. The reward, as a function of the stated fractiles and the actual observation, should encourage the expert to be honest. We introduce a family of such functions, discuss their properties and consider some extensions. The ideas are illustrated with an example about the number of delegates to Valencia-5.