Jonathan Wand

Jonathan Wand, Jasjeet S. Sekhon, and Walter R. Mebane, Jr.

Proceedings of the American Statistical Society

We apply a method of multinomial outlier detection we previously developed (Mebane, Sekhon, and Wand, 2001) to the 2000 Canadian Federal Election. The model is generalized to allow the choice set of voters to differ by election riding. We find substantially fewer outliers in the Canadian election, relative to the analyses of the 2000 U.S. Presidential election. (Mebane et al., 2001; Wand, Shotts, Sekhon, Mebane, Herron, and Brady, 2001). Moreover, the explanations for the outliers found in the Canadian election are not plausibly associated with electoral malfeasance.