Has anyone considered that perhaps, given that we have 50 polls quoting 95% confidence intervals, and that we consequently expect 2-3 outliers, the TN results might be one of them, or are we having too much fun making prejudiced statements about likely voters in the state?
The "confidence interval" is a random cutoff point low down on a bell curve, originally agreed on for scientific experiments for which a bell distribution of probabilities is a certainty. In such a bell distribution, a spot just within the "margin of error" is not perceptively more likely than a spot just outside it, both are infinitely more likely than a complete, how-the-hell-did-they-come-up-with-
that outlier, and both are infinitely less likely than a result very near the true result.
Also, these are political "polls" taking unscientific subsamples of an actual, non-laboratory population, then unscientifically reweighted to at least somewhat match the pollster's notion of what a scientific subsample of the population, were one available, would look like. They are not scientific experiments.
And we're comparing polls taken
now with election results in six weeks, anyhow - you can never know how much of the error is due to voters changing their minds or lying and how much is due to your bad polling.
Hence why pollsters' results don't actually match up with a Bell Curve distribution, and hence why, while it's a good shorthand remainder of the fact that even the best pollsters get it wrong sometimes, MoE lingo should not be taken literally.
Meh. I just need to get that off my chest once or twice every election season. Nothing personal. Anyways, as Ernest pointed out, the few GE polls we've had of the state so far, old as they are, mostly agree with this poll.