Ross Perot and the Multinomial Distribution

Abstract

The authors present a graphical multivariate approach to the analysis and reporting of political poll results when more than two candidates command a significant percentage of the vote. The authors argue that joint statistical inference is a more meaningful methodology in this situation; the current margin of error reporting is potentially inaccurate and may lead to inappropriate conclusions. The authors' methodology is applied to the results of four major national polls reported the week before the 1992 presidential election, with surprising implications.