NYT LIVE POLL THREAD:

[Zaybay]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

[wbrocks67]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Wait what, is that true? Any pollster using 2014 as a baseline in this era is doing a huge disservice.

[Zaybay]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Wait what, is that true? Any pollster using 2014 as a baseline in this era is doing a huge disservice.

check their methology page, all of these numbers have been using 2014 as a base.
https://www.nytimes.com/2018/09/06/upshot/live-poll-method.html

You must be logged in to read this quote.

So, basically all of these numbers of gender, age, demographics, and whatnot are based on who turned out in 2014, which explains why all of these polls seem to be more R-friendly.

[bilaps]

democrats again trying hard to dismiss a poll where they don't like the result

nothing new really..

it's by far the best poll of any given congressional district. it's not a gold standard cause you don't have a gold standard in a district like tx 23 or any house district tbh

and i would always take an individual poll of a district even with 300 lv rather than generic national ballot of 1000 rv

[Zaybay]

democrats again trying hard to dismiss a poll where they don't like the result

nothing new really..

it's by far the best poll of any given congressional district. it's not a gold standard cause you don't have a gold standard in a district like tx 23 or any house district tbh

and i would always take an individual poll of a district even with 300 lv rather than generic national ballot of 1000 rv

1. I have been a vocal critic of this poll the entire time, Im not jumping ship due to poor results.

2. We do actually have a golden standard for house races, Monmouth

3. This poll dosnt really do LV.....its still an RV poll. So your getting a poll of around 500 RV voters, which is the worst combo possible.

[bilaps]

democrats again trying hard to dismiss a poll where they don't like the result

nothing new really..

it's by far the best poll of any given congressional district. it's not a gold standard cause you don't have a gold standard in a district like tx 23 or any house district tbh

and i would always take an individual poll of a district even with 300 lv rather than generic national ballot of 1000 rv

1. I have been a vocal critic of this poll the entire time, Im not jumping ship due to poor results.

2. We do actually have a golden standard for house races, Monmouth

3. This poll dosnt really do LV.....its still an RV poll. So your getting a poll of around 500 RV voters, which is the worst combo possible.

you would agree though that this poll is better than any other poll in any of these districts? and certainly more meaningful than generic national ballot

[Holy Unifying Centrist]

300 lv has a margin of error of 8%

1000 lv has a margin of error of 3%

[bilaps]

300 lv has a margin of error of 8%

1000 lv has a margin of error of 3%

1000 national lv is much more irrelevant for house district than 300lv from the district

[Zaybay]

democrats again trying hard to dismiss a poll where they don't like the result

nothing new really..

it's by far the best poll of any given congressional district. it's not a gold standard cause you don't have a gold standard in a district like tx 23 or any house district tbh

and i would always take an individual poll of a district even with 300 lv rather than generic national ballot of 1000 rv

1. I have been a vocal critic of this poll the entire time, Im not jumping ship due to poor results.

2. We do actually have a golden standard for house races, Monmouth

3. This poll dosnt really do LV.....its still an RV poll. So your getting a poll of around 500 RV voters, which is the worst combo possible.

you would agree though that this poll is better than any other poll in any of these districts? and certainly more meaningful than generic national ballot

Well, no, because we have gotten Monmouth polling and other top pollsters in some of these districts. And in the case of GCB vs house polling, I would say it depends. In districts like WV-03 and TX-23, which have special factors, House polling is better. But in most of the CA, IL, PA, NJ, and such races, GCB gives a better picture.

[bilaps]

Well we have 2 monmouth polls from wv-3 and ca-48 from june or july, would love to see them poll wv-3 again.

[Zaybay]

Well we have 2 monmouth polls from wv-3 and ca-48 from june or july, would love to see them poll wv-3 again.

I just want more polls from them period.

[Devils30]

The TX-23 answers on the Mueller investigation, approval and generic ballot are a big red flag as well.

[Skye]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Freaking Siena is a mediocre pollster?

I've seen it all in this place.

[GeorgiaModerate]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Freaking Siena is a mediocre pollster?

I've seen it all in this place.

Oh, there are worse takes than that around here.  Another poster (not Zaybay) recently called NBC/Marist "liberal propaganda polls".

[Holy Unifying Centrist]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Freaking Siena is a mediocre pollster?

I've seen it all in this place.

Zaybay is a gigantic D hack. He is a world class contortonanist who tries to twist anything as good for dems, or discredit anything that isn't good for dems.

[Virginiá]

If anyone seriously believes Ted Cruz is up 7 and Hurd is up 9 in Texas-23 in this current year, you need your head examined. Your first clue should be Trump's approval rating in a district Hillary Clinton of all people won by 4 points at a point in time where Trump is polling in the high-30's, low-40's nationally.

The problem I have accepting such a result is that Hurd barely won in 2014 - a Republican wave year, and in 2016, he also almost lost to his challenger, although TX's swing against Trump could have hurt him too, but that also means he should be even more vulnerable now.

On the other hand, it's been in print for a while now that both parties seem to think Hurd is, for the time being, polling ahead, so maybe it turns against him in the final weeks, like waves tend to do, or it doesn't. TX-23 being a Latino-heavy district gives me pause, as they aren't always a reliable base for Democrats (in both turnout and support).

My opinion is that even if Hurd holds on this year, he has a good chance of being knocked off in 2020.

[Gass3268]

If anyone seriously believes Ted Cruz is up 7 and Hurd is up 9 in Texas-23 in this current year, you need your head examined. Your first clue should be Trump's approval rating in a district Hillary Clinton of all people won by 4 points at a point in time where Trump is polling in the high-30's, low-40's nationally.

The problem I have accepting such a result is that Hurd barely won in 2014 - a Republican wave year, and in 2016, he also almost lost to his challenger, although TX's swing against Trump could have hurt him too, but that also means he should be even more vulnerable now.

On the other hand, it's been in print for a while now that both parties seem to think Hurd is, for the time being, polling ahead, so maybe it turns against him in the final weeks, like waves tend to do, or it doesn't. TX-23 being a Latino-heavy district gives me pause, as they aren't always a reliable base for Democrats (in both turnout and support).

My opinion is that even if Hurd holds on this year, he has a good chance of being knocked off in 2020.

CA-21 is another one where Democrats would have better odds in 2020.

[Zaybay]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Freaking Siena is a mediocre pollster?

I've seen it all in this place.

Zaybay is a gigantic D hack. He is a world class contortonanist who tries to twist anything as good for dems, or discredit anything that isn't good for dems.

Sure pal, if you want to blindly believe polls without diving in to see the gears underneath, go right ahead.

Call me what you want, that doesnt change the quality of the polls.

Its odd, everyone seems focused on a choice of words "mediocre", and not my reasons for using them(small sample size, using 2014 voting numbers, still using RV)
 

[Holy Unifying Centrist]

If anyone seriously believes Ted Cruz is up 7 and Hurd is up 9 in Texas-23 in this current year, you need your head examined. Your first clue should be Trump's approval rating in a district Hillary Clinton of all people won by 4 points at a point in time where Trump is polling in the high-30's, low-40's nationally.

The problem I have accepting such a result is that Hurd barely won in 2014 - a Republican wave year, and in 2016, he also almost lost to his challenger, although TX's swing against Trump could have hurt him too, but that also means he should be even more vulnerable now.

On the other hand, it's been in print for a while now that both parties seem to think Hurd is, for the time being, polling ahead, so maybe it turns against him in the final weeks, like waves tend to do, or it doesn't. TX-23 being a Latino-heavy district gives me pause, as they aren't always a reliable base for Democrats (in both turnout and support).

My opinion is that even if Hurd holds on this year, he has a good chance of being knocked off in 2020.

Hurd's favorable rating has gotten a lot higher since 2016. A lot more people know and respect him now.


Not gonna bother responding to Zaybay, the user who cried when he realized black people can actually vote for Larry Hogan and aren't locked into voting dem.

[Zaybay]

If anyone seriously believes Ted Cruz is up 7 and Hurd is up 9 in Texas-23 in this current year, you need your head examined. Your first clue should be Trump's approval rating in a district Hillary Clinton of all people won by 4 points at a point in time where Trump is polling in the high-30's, low-40's nationally.

The problem I have accepting such a result is that Hurd barely won in 2014 - a Republican wave year, and in 2016, he also almost lost to his challenger, although TX's swing against Trump could have hurt him too, but that also means he should be even more vulnerable now.

On the other hand, it's been in print for a while now that both parties seem to think Hurd is, for the time being, polling ahead, so maybe it turns against him in the final weeks, like waves tend to do, or it doesn't. TX-23 being a Latino-heavy district gives me pause, as they aren't always a reliable base for Democrats (in both turnout and support).

My opinion is that even if Hurd holds on this year, he has a good chance of being knocked off in 2020.

CA-21 is another one where Democrats would have better odds in 2020.

Looking at the districts that should be D but arent, CA-21, TX-23, the FL seats, all of them share one thing, its a majority Hispanic seat with a white population that is heavily R. I can see these seats being competitive in 2020, but in 2018, unless Hispanic turnout increases, these seats will likely stay GOP until 2020.

[GeorgiaModerate]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Freaking Siena is a mediocre pollster?

I've seen it all in this place.

Zaybay is a gigantic D hack. He is a world class contortonanist who tries to twist anything as good for dems, or discredit anything that isn't good for dems.

Sure pal, if you want to blindly believe polls without diving in to see the gears underneath, go right ahead.

Call me what you want, that doesnt change the quality of the polls.

Its odd, everyone seems focused on a choice of words "mediocre", and not my reasons for using them(small sample size, using 2014 voting numbers, still using RV)
 

I think people are taking issue with your blanket characterization of Siena as a mediocre pollster.  Criticizing these particular polls for their methodology is reasonable.

[GeorgiaModerate]

A sound take on TX-23 from Nate Cohn:

https://twitter.com/Nate_Cohn/status/1039308049423458305

[Florida Man for Crime]

Zaybay, you are misunderstanding their methodology.

They are not "using 2014 as a baseline."

Take a look at the wikipedia article on stratified sampling to learn what they mean when they say "stratified by turnout in vote." This means explicitly that they are making sure that they call both people who voted in 2014 and people who did not vote in 2014. Stratification makes sure you include some part of a subpopulation in your sample. So this is the opposite of what you are saying.

https://en.wikipedia.org/wiki/Stratified_sampling

You also are apparently not understanding what a turnout model in a voter file is, and how that is different from a likely voter model typically used by a poll.


Now, go look at the "Our turnout model" section of the results page for, for example, MN-03.

This lists "Our poll under different turnout scenarios."

One scenario they list is "Our estimate," which in the case of MN-03 projects a turnout of about 328k, and under. This is the turnout scenario they are using for the default topline result that they report. Indeed, this has Phillips +9, which not coincidentally matches their topline result.

By contrast, another turnout scenario they list is "The types of people who voted in 2014" This scenario projects a turnout of about 268k, which not coincidentally is very close to the 269,361 people who voted in the 2014 congressional election in MN-03 (some slight deviation being allowable as a result of changes in registration/demographics since 2014).

So you are just spreading blatant falsehoods and misleading people by repeating the claim that the poll is based upon 2014-like turnout. It explicitly is not, and they indeed directly show how their results would differ if they assumed 2014-like turnout. If you want to talk about what the result would look like if 2014 were used as a baseline, fine, but you should be referring to the alternative "The types of people who voted in 2014" turnout scenario, not to the turnout scenario that they are actually using for their topline.


I suspect that underlying this, you may not be understanding what they are doing in making their turnout model. From here:

You must be logged in to read this quote.

You may be misinterpreting "vote-history-based model of turnout in the 2014 midterm election" to mean that they are only including/calling people who voted in 2014. That is incorrect. What they are doing is making an individual-level probability model of turnout, predicting 2014 turnout (y variable in a regression) using data in the voter file such as known records of whether the individual voted in other elections and also demographic characteristics such as age as explanatory variables (x variables in a regression). If you are familiar with logit/probit models, you can think of it as though they are using one of those, though technically they are probably using some slightly different but related techniques.

Now, what this means is that each voter is assigned a probability, including all voters who did not vote in 2014. Indeed, many voters who did not vote in 2014 but who have voted in other elections will have quite high scores and have a very high chance of being called and included in the poll.

Next, the model is adjusted in various ways having nothing to do with 2014 and everything to do with post-Trump elections. I won't walk through the details of those, but for the moment will assume you can read and understand that part. After these modifications, the model is projecting an electorate that is not the same as in 2014 and is similar to what we have seen in post-Trump elections/special elections. And in addition, even after those modifications, they give greater weight to voters who say they will vote, even if they have a low individual turnout score.

This is all a very different process from screening out people who did not vote in 2014 from the call list, which I think is what most people would understand by using 2014 as a baseline for the poll. Many voters who did not vote in 2014 are in fact included in the poll, and overall turnout projections differ significantly from 2014.

In conclusion, if you want to say that they are using something as a baseline, it is patently and explicitly incorrect to say that they are using 2014 as a baseline. If you want to oversimplify it to the point of saying that they are using some previous election as a baseline, it would be much more correct to say that they are using the Virginia 2007 and Congressional Special elections like in OH-12 etc as a baseline.

This may turn out to be correct or incorrect (personally I think it is about the best projection of turnout one can make with the available data), but that is what their default topline turnout scenario is reflective of, not 2014.

[Zaybay]

Perhaps this is what happens when you put faith in a mediocre pollster that has a small sample size, while using 2014 as the baseline for the demographics used. You get buggy numbers.

Freaking Siena is a mediocre pollster?

I've seen it all in this place.

Zaybay is a gigantic D hack. He is a world class contortonanist who tries to twist anything as good for dems, or discredit anything that isn't good for dems.

Sure pal, if you want to blindly believe polls without diving in to see the gears underneath, go right ahead.

Call me what you want, that doesnt change the quality of the polls.

Its odd, everyone seems focused on a choice of words "mediocre", and not my reasons for using them(small sample size, using 2014 voting numbers, still using RV)
 

I think people are taking issue with your blanket characterization of Siena as a mediocre pollster.  Criticizing these particular polls for their methodology is reasonable.

It was just word choice, literally any other word could go there, Im not writing a college essay with these posts, you know. I actually like Siena when they switch to a LV model, because they stop using terrible methology. Ive already made my opinion on Siena's current methology, anyway.

[Zaybay]

snip, though I must say, very well written!

Alright, a bit long, but you still misunderstand what I mean.

I will say it as clearly as possible, so we are all on the same page.

My problem with these polls is threefold
1. The polls have a rather low sample of 500 voters
2. The use of RV, even though its after Labor day, though I do applaud that they keep it in one of the small tabs.
3. And this is the one everyone is getting confused on, the prerequisites they are calling are based on 2014. Basically, according to their methology, if the turnout was 34% R, 26% D, 10% 18-27, 30%40-65, these are the numbers they use. Thats my problem with it, these numbers arent reliable in the current scenario, and other pollsters have adapted fine by using other polls, likely voters, that kind of stuff. Siena has not, which is why, when they switch from this system to likely voters, they become a much better pollster all around.

I hope that clears everything up.