I’d like to thank Kevin Drum of Political Animal for clearing up a bit of the nonsense around the phrase statistical tie
.
Essentially: if in a two-candidate race you have people’s preferences within the margin of error — say, 48% to 45% with a 4% MoE — there’s a fair number of folks who’ll call that a statistical tie, with the implication that the apparent lead is actually meaningless. This is wrong, because the whole idea of a margin of error is meaningless without the companion notion of a confidence interval.
OK, so many of you probably already know this, and most of the rest probably don’t care. But if you want the numerical details for the implied confidence in a statement like X is winning
, go see Kevin’s post.