Article ID: | iaor20051435 |
Country: | United States |
Volume: | 58 |
Issue: | 1 |
Start Page Number: | 46 |
End Page Number: | 54 |
Publication Date: | Feb 2004 |
Journal: | American Statistician |
Authors: | Wu Samuel S., Yang Mark C.K. |
Keywords: | decision: studies, statistics: general |
The scandal in the 2002 Winter Olympic figure skating judging forced the International Skating Union (ISU) to address the vulnerability of the current nine-judge placement rule. The ISU proposed a stricter ethics code for the judges and changes to the decision rule. Unfortunately, the new decision rule is no better than the previous rule in countering manipulation. In this article, we construct an alternative rule and show that, under reasonable assumptions, our rule is more robust against manipulation than existing rules. When other properties such as Condorcet efficiency, frequency of ties, and violation of monotonicity are compared, our rule is also more desirable. In addition, a commonly used rank sum rule is examined and shown to be even worse than the other rules. These results are applicable to any decision rule based on the relative ranks of proposals or contestants.