NYT LIVE POLL THREAD:

[Fmr. Gov. NickG]

I know it's early in MN-03, but a 100 point lead sounds awfully hard to make up.

[Fmr. Gov. NickG]

These polls aren't actually that interesting if they are all just weighted to show up as virtual ties.

[Fmr. Gov. NickG]

These polls aren't actually that interesting if they are all just weighted to show up as virtual ties.

That's a good point. It would be pretty suspicious if literally every poll showed a result within 1 or 2%.

Well, they are polling highly competitive districts so far.  It's not that surprising.

Even if every district were in fact tied, sampling error should produce a larger range of results.  If you saw six polls of the same race that were all within the 1% of each other, you would start to suspect herding, often through manipulating the weights to achieve a certain balance in the turnout model.

[Fmr. Gov. NickG]

I'm happy NYT is polling VA-07 next.  So far, all of the polls have been in districts everyone assumes are toss-ups, and a poll of Lean-R district adds some diversity.  I hope we'll get some polls of districtds like CA-22 and PA-08 as well.

[Fmr. Gov. NickG]

Seems odd to be polling Minnesota in the middle of the opening Vikings game.

[Fmr. Gov. NickG]

I think NYT suspended their MN-03 poll until they can find enough Paulsen supporter to get the race back to within one point.

[Fmr. Gov. NickG]

If you are polling from an existing file of registered voters, how do you account for people who haven't registered yet, but will prior to the election day?  Voters in every state still have about a month left to register.  And I have to imagine those voters will heavily favor Democrats.

[Fmr. Gov. NickG]

How is NYT going to get to 100 polls by election day if they are only starting one per day?

[Fmr. Gov. NickG]

All these polls are pretty much going as I would have expected, except for CA-48, where I'm surprised Rouda doesn't have a clear lead, and MN-03, which I thought would be closer.  I'd also be surprised in Hurd is up by 10 when the poll is complete, but it's way too early to assume that.

[Fmr. Gov. NickG]

The 40-point gender gap in WV-03 is completely nuts.  How do people in West Virginia stay married??  I can't imagine living with a Trump supporter.

[Fmr. Gov. NickG]

Looking further into the WV-03 crosstab, Miller is winning 74-16 among the very small sample of 18-29 year olds.  This group was significantly undersampled, so it is likely that the 21 young people they did get were substantially overweighted.

[Fmr. Gov. NickG]

This is inaccurate. More people voted "lean D" in that WV-03 poll than voted "lean R", and that was especially true before I bumped it. Most voted toss up though.

But even putting aside the head to head margin, the poll is ugly for Ojeda. Almost as ugly as MN-03 was for Paulsen. I wouldn't hold my breath waiting for Ojeda to get a last minute surge among people who approve of Trump, think Republicans should take the House, think Trump is draining the swamp, don't think black people should have the right to kneel, and think discrimination against whites is worse than it is against minorities.

The poll didn't ask about either of these things.

I think he's nitpicking the wording you used.

[Fmr. Gov. NickG]

I've always considered TX-23 Tilt R/Lean R for a while. I'm aware of the district's demographics and that it went to Clinton, but Hurd is a fairly strong incumbent.

It seems to be a common misconception about Republicans that represent heavily Hispanic districts in which the white voters are very heavily Republican that the Republican candidates are "strong candidates" or "strong incumbents."

Hurd is regarded as a "strong incumbent."

Valadao is regarded as a "strong incumbent."

What do they have in common? Not that they are actually strong incumbents, but that in their districts, Hispanics simply do not vote, ESPECIALLY in a midterm. Whereas white people DO vote. And they vote for Republicans. It is as simple as that.

You could stick Eric Paulsen in TX-23 and Will Hurd in MN-03 and then Eric Paulsen would be a "strong candidate" and Will Hurd would be a "DOA incumbent who has never had a tough race who is about to get triaged."

How do you think Blake Farenthold won TX-27? Same way. It most certainly was not that Blake Farenthold was a strong candidate.

Both Valadao and Hurd did strongly outperform their party's presidential nominee in 2016. (Though I suppose Paulsen did too.)

[Fmr. Gov. NickG]

Looking further into the WV-03 crosstab, Miller is winning 74-16 among the very small sample of 18-29 year olds.  This group was significantly undersampled, so it is likely that the 21 young people they did get were substantially overweighted.

Yeah, but in WV it is somewhat more plausible that the 18-29s would be heavily Republican than in VA (maybe not quite 74-16, but still strongly Republican).

In general though, with how difficult it is to get young people to answer a poll, the ones that they do reach tend to have disproportionate influence. So that may be something to pay attention to in other polls.

In MN-03, for example, they got a much larger sample of 18-29 year olds (67), but in that case there wasn't that much difference between their support and older voters.

It may be true that younger voters in WV are heavily Republican.  I just think pollsters should be very wary of overweighting very small subsamples that will tend to show extreme results.

[Fmr. Gov. NickG]

So Republicans are leading in every House race polled so far, except one in suburban Minneapolis. Why do we think the House is going to flip again?

Assuming that all seven completed polls so far are 100% accurate (in terms of the final two party margin), none of these districts are necessarily on the list of must takes for Democrats to win back the House. In fact, 538's current deluxe model shows that these polls fit right where we expect them to be in a 228-207 Democratic win.

If any of these races is a "must win" for Dems, it's CA-48.  Though that race might not be decided until weeks after the election, so we'll almost certainly know the national result first.

[Fmr. Gov. NickG]

I wonder why they polled so few people in VA-07 today.

[Fmr. Gov. NickG]

https://twitter.com/Nate_Cohn/status/1039627190822944770

This is exactly what I was concerned about last night in the WV-03 poll.
It's a repeat of that one young black Trump supporter who always counted as like 30 people in the LAT tracking poll in 2016.

[Fmr. Gov. NickG]

I admire the transparency of these polls.  And they do seem to be for the most part using established best practices.  

But the biggest thing I am taking away from all of them is how completely futile it is to try to poll 20-somethings over the phone.  Like, do you know any college student who will answer an unknown number in their cell phone?

[Fmr. Gov. NickG]

Because both the actual sample and target sample of 18-29 voters is so small, this probably doesn't have an enormous effect on the bottom lines of these polls.

But this demographic is consistently undersampled and unrepresentative, with terrible response rates leading to a lot of error that are essentially "false positives".

I'm not sure what the best solution to this is.  But an adequate solution CAN'T be to just give the tiny subsample you do get a lot more weight.

[Fmr. Gov. NickG]

Because both the actual sample and target sample of 18-29 voters is so small, this probably doesn't have an enormous effect on the bottom lines of these polls.

But this demographic is consistently undersampled and unrepresentative, with terrible response rates leading to a lot of error that are essentially "false positives".

I'm not sure what the best solution to this is.  But an adequate solution CAN'T be to just give the tiny subsample you do get a lot more weight.

Any weight that gives a single respondent the value of two responses is too heavy.
Solution is oversampling, even if it is a small oversample. Would greatly improve the clear issue they are having.

My concern is that any sample, regardless of how large, of young people taken over the phone is going to be unrepresentative and error-prone.

Maybe someone could go around handing out polls in a food truck?

[Fmr. Gov. NickG]

CO-06 might be the first of these polls where I will be genuinely disappointed if the Democrat doesn't have a clear lead.

[Fmr. Gov. NickG]

So far Democrats are leading 5 of the 12 polls, including 5 of 9 rated by Cook as Toss-Up, and 0 of 3 rated as Lean Republican.  (All of these 12 are rated Toss-Up by Sabato, but I do think these results may indicate he was at least too quick to move WV-03 and WI-01 there.)

[Fmr. Gov. NickG]

I would categorize three races differently:

Excellent: CO-06, MN-03

Good: MN-08, WI-01

Meh: IL-06, IL-12, KY-06, VA-07, WV-03

Bad: CA-48

Yikes: TX-23

[Fmr. Gov. NickG]

If the Mucarsel-Powell+7 result holds (IF), that would be kind of surprising to me. I'd sooner expect Hurd to be down 7 and Curbelo to be up 8, rather than the other way around.

Why?  For the most part, the 2018 election will be a referendum on Trump.  Curbelo's district is D+6 and vote for Hillary by 16 points; Hurd's is R+1 and voted for Hillary by only 3.   

[Fmr. Gov. NickG]

The overall trend I'm picking up from this poll is that Trump matters even more and individual candidates matter less than the conventional wisdom suggests.