A statistical tie is a question of the margin of error.
It's a dumb term -- just because you're not 95% sure it isn't a tie doesn't mean it should be treated as likely 50/50.
But that is roughly how probability works if if one can't determine the difference between a 57-43 chance and a 43-57 chance. At this stage one can't make so easy a distinction. A few days from Election Day we will see far more states that look as if they are in the margin of error clearly show which side of 50-50.
That...what? The fact that preferences are more fluid now, and likely voters more concrete, has nothing to do with this statistical calculation.
An exact tie looks much like a 50-50 chance. Statistical ties close to even, especially if they are poll numbers bouncing around exact ties, look like 50-50 chances. When the numbers quit bouncing, then one has something other than a 50-50 chance.
Again, just because we can't be more than 95% sure there isn't a tie, doesn't mean a 94% chance should be assumed to be a tie. "Statistical tie" is a dumb, dumb term.
In the end there are no statistical ties except absolutely-even counts. Dubya won Florida by 537 votes in 2000, and we define that as a victory for Dubya because such is how the law is set for determining who wins the election. Barack Obama ended up with a few more than 28K votes more than John McCain in Indiana, and by definition that was a win for Barack Obama even if it was a statistical tie (a 0.03% difference).
I don't even understand how you're apply MoE to what's supposed to be a
population, not a
sample, and cannot realistically be a representative sample if you're treating it as a population (e.g., if you claim there are improperly uncounted ballots or something.)
