How should we score athletes and candidates: geometric scoring rules
share
Individual rankings are often aggregated using scoring rules: each position in each ranking brings a certain score; the total sum of scores determines the aggregate ranking. We study whether scoring rules can be robust to adding or deleting particular candidates, as occurs with spoilers in political elections and with athletes in sports due to doping allegations. In general the result is negative, but weaker robustness criteria pin down a one-parameter family of geometric scoring rules with the scores 0, 1, 1+p, 1+p+p^2,.... These weaker criteria are independence from deleting unanimous winner (e.g., doping allegations) and independence from deleting unanimous loser (e.g., spoiler candidates). This family generalises three central rules: the Borda rule, the plurality rule and the antiplurality rule. For illustration we use recent events in biathlon; our results give simple instruments to design scoring rules for a wide range of applications.
READ FULL TEXT