High Point University/SurveyUSA: Tie in NC, Gardner+4 in CO, Shaheen+2 in NH

[DrScholl]

A note to krazen and his little family here, the Colorado poll only has a 6% sample of Hispanics, which is below the actually number that will vote. That makes a big difference.

[DrScholl]

Its usually a good sign when the other side starts unskewing the polls.

It's not unskewing like you and your people did in 2012. The demographic samples actually matter, that's why Gallup flubbed up so badly last cycle.

Try again.

[DrScholl]

A note to krazen and his little family here, the Colorado poll only has a 6% sample of Hispanics, which is below the actually number that will vote. That makes a big difference.

Gardner will do much better with whites than a mere +6. That makes a big difference.

He's not going to win 90% of whites. You have to realize that whites in Colorado don't vote the same as they do in the south.

[DrScholl]

A note to krazen and his little family here, the Colorado poll only has a 6% sample of Hispanics, which is below the actually number that will vote. That makes a big difference.

Gardner will do much better with whites than a mere +6. That makes a big difference.

He's not going to win 90% of whites. You have to realize that whites in Colorado don't vote the same as they do in the south.

Holy mathematical strawman!

Judging from who I was responding to, I figured that was what you believed.

[DrScholl]

Its usually a good sign when the other side starts unskewing the polls.

It's not unskewing like you and your people did in 2012. The demographic samples actually matter, that's why Gallup flubbed up so badly last cycle.

Try again.

I did not unskew in 2012. I admit to doing so in 2010, and having seen the results of that election, learned my lesson not to do so in 2012 or this election.

As others have mentioned before, SurveyUSA has had a tendency to give accurate top-line results despite weird crosstabs. How exactly is unskewing based on racial breakdowns that you happen to disagree with somehow better than unskewing based on partisan breakdowns that you happen to disagree with?

It's not unskewing, it's based on how previous polls have worked. PPP was able to get accurate results in 2012, because they used a demographic sample that included more minorities, when other pollsters expected those groups not to show up. Party ID is fluid, unlike demographics. This has been discussed before.