Recent Poll shows a supporter of the Donald leading Biden

[dezii]

supporter for Senate race in Montana by 6 points...I post this just as a reminder to the many vocal posters here that this 'Presidential race is not a slame duck win for Biden or any Democratic presidential candidate...There are millions who support the incumbent..whether we  question their reasoning  ......can't get cocky or think it is a slam dunk for the Democrats.....
====================================

news.yahoo.com/poll-gives-democrats-bad-news-in-key-senate-race-183924844.html

[countrygirl2]

I don't take it for granted, I'm sweating bullets we don't get rid of that piece of sh**.
If he is back in, our democracy is finished.

[haapai]

I'm not sure if this is even news.  I've never even heard a suggestion that Montana's electoral votes would go to Biden or that the Republicans had any chance of losing a Senate seat there.  The only thing that might be newsworthy about this poll is that it says that Daines leads Bullock by six points according to the most recent poll when a poll two weeks ago by a different organization showed a tighter race (a two point lead for Daines.)

Missing from the article is any discussion on what the margin of error in the two polls was and how they might have differed in methodology.  I need to know stuff like that before I decide whether or not this hardening of support for an incumbent GOP Senator has any implications on how other Senate races might be trending. 

ETA:  It also matters quite a bit exactly when, as in on which dates, this poll was conducted.  A very clear "Save the Senate/distance yourself from Trump is you have to" call to action/signal went out on Thursday.  I  want to know how many of these responses were collected before that signal got sent or absorbed. 

ETA: I misread the link.  It's a bigger shift than I realized and the predicted winner has changed.  I'm embarrassed. 

[thyme4change]

I don't think Montana is going blue anytime soon.

[billisonboard]

Aug 3, 2020 17:20:23 GMT -5 haapai said:

I'm not sure if this is even news.  I've never even heard a suggestion that Montana's electoral votes would go to Biden or that the Republicans had any chance of losing a Senate seat there.  The only thing that might be newsworthy about this poll is that it says that Daines leads Bullock by six points according to the most recent poll when a poll two weeks ago by a different organization showed a tighter race (a two point lead for Daines.)

Missing from the article is any discussion on what the margin of error in the two polls was and how they might have differed in methodology.  I need to know stuff like that before I decide whether or not this hardening of support for an incumbent GOP Senator has any implications on how other Senate races might be trending. 

ETA:  It also matters quite a bit exactly when, as in on which dates, this poll was conducted.  A very clear "Save the Senate/distance yourself from Trump is you have to" call to action/signal went out on Thursday.  I  want to know how many of these responses were collected before that signal got sent or absorbed. 

Puplic Policy Poll, 9-10 July, +/-2.8, couldn't find methodology
The Montana Emerson College poll was conducted July 31-August 2, 2020. The sample consisted of likely registered voters, n=584, with a Credibility Interval (CI) similar to a poll’s margin of error (MOE) of +/-4 percentage points. The data sets were weighted by gender, age, education, and ethnicity based on 2016 voter turnout modeling. It is important to remember that subsets based on gender, age, party breakdown, ethnicity, and region carry with them higher margins of error, as the sample size is reduced. Data was collected using a list of 30,000 mobile numbers provided by Aristotle Inc. A SMS-text to web was sent July 31 to 15,000 numbers and on August 1 to an additional 15,000 numbers. Responses were open until noon August 2.,

[haapai]

How the f** does n=584 turn into a MOE or +/- 4 percentage points.  That's a tiny sample.  How did it possibly get such a small MOE?  <br>

[billisonboard]

Aug 3, 2020 18:27:26 GMT -5 haapai said:

How the f** does n=584 turn into a MOE or +/- 4 percentage points.  That's a tiny sample.  How did it possibly get such a small MOE? 

Universe size

[haapai]

Aug 3, 2020 18:33:14 GMT -5 billisonboard said:

Aug 3, 2020 18:27:26 GMT -5 haapai said:

How the f** does n=584 turn into a MOE or +/- 4 percentage points.  That's a tiny sample.  How did it possibly get such a small MOE? 

Universe size

huh?  Universe size.  Is that a jab at the population of Montana or a term that I should remember from the class on polling that I took almost 35 years ago? 

I did pretty badly in that class and just about the only things that I think that I remember from it were A.) Ignore almost any poll with a three-figure n and B.) weighted polling is not to be trusted either.  (That was back in the eighties when random polling was much more feasible and weighted polling was just a cheap way of pretending that you had done something that resembled random sampling when almost everyone that was publishing weighted polling had done nothing of the sort.)

I need some technical tips regarding how to think of the one poll that we have any methodology on. 

[The Walk of the Penguin Mich]

Aug 3, 2020 18:18:35 GMT -5 billisonboard said:

Aug 3, 2020 17:20:23 GMT -5 haapai said:

I'm not sure if this is even news.  I've never even heard a suggestion that Montana's electoral votes would go to Biden or that the Republicans had any chance of losing a Senate seat there.  The only thing that might be newsworthy about this poll is that it says that Daines leads Bullock by six points according to the most recent poll when a poll two weeks ago by a different organization showed a tighter race (a two point lead for Daines.)

Missing from the article is any discussion on what the margin of error in the two polls was and how they might have differed in methodology.  I need to know stuff like that before I decide whether or not this hardening of support for an incumbent GOP Senator has any implications on how other Senate races might be trending. 

ETA:  It also matters quite a bit exactly when, as in on which dates, this poll was conducted.  A very clear "Save the Senate/distance yourself from Trump is you have to" call to action/signal went out on Thursday.  I  want to know how many of these responses were collected before that signal got sent or absorbed. 

Puplic Policy Poll, 9-10 July, +/-2.8, couldn't find methodology
The Montana Emerson College poll was conducted July 31-August 2, 2020. The sample consisted of likely registered voters, n=584, with a Credibility Interval (CI) similar to a poll’s margin of error (MOE) of +/-4 percentage points. The data sets were weighted by gender, age, education, and ethnicity based on 2016 voter turnout modeling. It is important to remember that subsets based on gender, age, party breakdown, ethnicity, and region carry with them higher margins of error, as the sample size is reduced. Data was collected using a list of 30,000 mobile numbers provided by Aristotle Inc. A SMS-text to web was sent July 31 to 15,000 numbers and on August 1 to an additional 15,000 numbers. Responses were open until noon August 2.,

So the way I read this is that 30,000 texts were sent to cell numbers, and there was a response rate of 584?  A less than 2% response rate?

[billisonboard]

Aug 3, 2020 19:26:22 GMT -5 The Walk of the Penguin Mich said:

Aug 3, 2020 18:18:35 GMT -5 billisonboard said:

Puplic Policy Poll, 9-10 July, +/-2.8, couldn't find methodology
The Montana Emerson College poll was conducted July 31-August 2, 2020. The sample consisted of likely registered voters, n=584, with a Credibility Interval (CI) similar to a poll’s margin of error (MOE) of +/-4 percentage points. The data sets were weighted by gender, age, education, and ethnicity based on 2016 voter turnout modeling. It is important to remember that subsets based on gender, age, party breakdown, ethnicity, and region carry with them higher margins of error, as the sample size is reduced. Data was collected using a list of 30,000 mobile numbers provided by Aristotle Inc. A SMS-text to web was sent July 31 to 15,000 numbers and on August 1 to an additional 15,000 numbers. Responses were open until noon August 2.,

So the way I read this is that 30,000 texts were sent to cell numbers, and there was a response rate of 584?  A less than 2% response rate?

That is how I read it also.

[billisonboard]

Aug 3, 2020 19:09:04 GMT -5 haapai said:

Aug 3, 2020 18:33:14 GMT -5 billisonboard said:

Universe size

huh?  Universe size.  Is that a jab at the population of Montana or a term that I should remember from the class on polling that I took almost 35 years ago? 

I did pretty badly in that class and just about the only things that I think that I remember from it were A.) Ignore almost any poll with a three-figure n and B.) weighted polling is not to be trusted either.  (That was back in the eighties when random polling was much more feasible and weighted polling was just a cheap way of pretending that you had done something that resembled random sampling when almost everyone that was publishing weighted polling had done nothing of the sort.)

I need some technical tips regarding how to think of the one poll that we have any methodology on. 

I took my polling class about that long ago also. Universe is a term I remember from the class. It is the total number of the group you are surveying (e.g. registered voters in Montana). Margin of error is based on a ratio of surveyed/universe. A national survey of less than 1,000 i agree is suspect. Montana less so.

I question the randomness of respondents with the method used in the Emerson poll. The fact that i could not find methodology for the other makes me disinterested in it.

[djAdvocate]

Aug 3, 2020 17:20:23 GMT -5 haapai said:

I'm not sure if this is even news.  I've never even heard a suggestion that Montana's electoral votes would go to Biden or that the Republicans had any chance of losing a Senate seat there.  The only thing that might be newsworthy about this poll is that it says that Daines leads Bullock by six points according to the most recent poll when a poll two weeks ago by a different organization showed a tighter race (a two point lead for Daines.)

Missing from the article is any discussion on what the margin of error in the two polls was and how they might have differed in methodology.  I need to know stuff like that before I decide whether or not this hardening of support for an incumbent GOP Senator has any implications on how other Senate races might be trending. 

ETA:  It also matters quite a bit exactly when, as in on which dates, this poll was conducted.  A very clear "Save the Senate/distance yourself from Trump is you have to" call to action/signal went out on Thursday.  I  want to know how many of these responses were collected before that signal got sent or absorbed. 

well, hear it now, then.  I have three things to say:

1) polling has been thin in this state.  this is only the second head to head poll I have seen.

2) Emerson has a noted Republican lean.  now, it might turn out that their modeling is better than the other pollsters.  in the last election, for example, Trafalgar (which is not a very good pollster on average) had fantastic modeling, and predicted that Trump would win in Michigan and PA.  they were the only pollster that did so.  so, I am reluctant to call Emerson an outlier, particularly in light of the thin polling in the state.

3) election betting odds has responded to this poll favorably, moving 6.1% in the last day.  that is a big jump.  it should also be noted that there was some spillover in Iowa, as well.  they are now rating that race as slightly favoring Republicans.

having said that, the Emerson poll in the presidential race reflects the previous PPP poll within 2%, and Trump is very unlikely to lose the state.  so, no, this is not really in play in the GE, even though the Senate race is competitive.

[djAdvocate]

Aug 3, 2020 19:48:52 GMT -5 billisonboard said:

Aug 3, 2020 19:09:04 GMT -5 haapai said:

huh?  Universe size.  Is that a jab at the population of Montana or a term that I should remember from the class on polling that I took almost 35 years ago? 

I did pretty badly in that class and just about the only things that I think that I remember from it were A.) Ignore almost any poll with a three-figure n and B.) weighted polling is not to be trusted either.  (That was back in the eighties when random polling was much more feasible and weighted polling was just a cheap way of pretending that you had done something that resembled random sampling when almost everyone that was publishing weighted polling had done nothing of the sort.)

I need some technical tips regarding how to think of the one poll that we have any methodology on. 

I took my polling class about that long ago also. Universe is a term I remember from the class. It is the total number of the group you are surveying (e.g. registered voters in Montana). Margin of error is based on a ratio of surveyed/universe. A national survey of less than 1,000 i agree is suspect. Montana less so.

I question the randomness of respondents with the method used in the Emerson poll. The fact that i could not find methodology for the other makes me disinterested in it.

Emerson is highly rated.  they have an A- rating at 538, which puts them very near the top tier.  the survey in question has an MOE of 4%, due to low sample size.  however, it is statistically significant in that the last poll in MT showed the Democrat leading.  there is no overlap in the 95% certain zone, which indicates that attitudes are changing in MT toward the Democrat in the race.

this was never going to be an easy one for Democrats to win.  but it is one of two "outside chances" that are worth pursuing, imo.  the other one is Iowa.

[haapai]

Aug 3, 2020 19:29:24 GMT -5 billisonboard said:

Aug 3, 2020 19:26:22 GMT -5 The Walk of the Penguin Mich said:

So the way I read this is that 30,000 texts were sent to cell numbers, and there was a response rate of 584?  A less than 2% response rate?

That is how I read it also.

My God!  How in the hell is something with that kind of non-response rate worth crap?  Please point me toward something that tells me how in the name of all that is holy that you can reliably take response rates like that and weigh them and weight them and not be scared out of your mind that some sort of systemic bias isn't about to bite you in the butt.

 

My education on polling stopped in an era when something much closer to random sampling by randomly dialing telephone numbers was still possible.  Though we could see that non-response rates, or at least behavior that should have been tallied as non-responses, was on the rise, and seriously threatening the statistical validity of the samples that we took, we never imagined that any kind of fix was possible. 

What kind of math, statistics, or amnesia has occurred in the last three decades that makes this kind of response rate something that any pollster can deal with? 

[The Walk of the Penguin Mich]

Aug 3, 2020 22:13:52 GMT -5 haapai said:

Aug 3, 2020 19:29:24 GMT -5 billisonboard said:

That is how I read it also.

My God!  How in the hell is something with that kind of non-response rate worth crap?  Please point me toward something that tells me how in the name of all that is holy that you can reliably take response rates like that and weigh them and weight them and not be scared out of your mind that some sort of systemic bias isn't about to bite you in the butt.

 

My education on polling stopped in an era when something much closer to random sampling by randomly dialing telephone numbers was still possible.  Though we could see that non-response rates, or at least behavior that should have been tallied as non-responses, was on the rise, and seriously threatening the statistical validity of the samples that we took, we never imagined that any kind of fix was possible. 

What kind of math, statistics, or amnesia has occurred in the last three decades that makes this kind of response rate something that any pollster can deal with? 

I don’t know, it just seems to me that this would be an incredibly skewed sample of people answering.  I know that I have gotten a bunch of texts from local candidates and I am wondering how they got my phone number.  Other than it having a KY area code (the KY texts stopped a few years ago), the only people who get this number are my friends and my bank.  

[djAdvocate]

Aug 3, 2020 18:27:26 GMT -5 haapai said:

How the f** does n=584 turn into a MOE or +/- 4 percentage points.  That's a tiny sample.  How did it possibly get such a small MOE? 

I am so glad you asked!

www.surveysystem.com/sscalc.htm

statistics is weird.  you only need about 300 to get a reliable sample, even when you are surveying MILLIONS OF PEOPLE.

I love polling.  it is so interesting.

[djAdvocate]

Aug 3, 2020 22:42:11 GMT -5 The Walk of the Penguin Mich said:

Aug 3, 2020 22:13:52 GMT -5 haapai said:

My God!  How in the hell is something with that kind of non-response rate worth crap?  Please point me toward something that tells me how in the name of all that is holy that you can reliably take response rates like that and weigh them and weight them and not be scared out of your mind that some sort of systemic bias isn't about to bite you in the butt.

 

My education on polling stopped in an era when something much closer to random sampling by randomly dialing telephone numbers was still possible.  Though we could see that non-response rates, or at least behavior that should have been tallied as non-responses, was on the rise, and seriously threatening the statistical validity of the samples that we took, we never imagined that any kind of fix was possible. 

What kind of math, statistics, or amnesia has occurred in the last three decades that makes this kind of response rate something that any pollster can deal with? 

I don’t know, it just seems to me that this would be an incredibly skewed sample of people answering.  I know that I have gotten a bunch of texts from local candidates and I am wondering how they got my phone number.  Other than it having a KY area code (the KY texts stopped a few years ago), the only people who get this number are my friends and my bank.  

do you want me to explain why that is incredibly unlikely? or did you intend that remark as a rhetorical one?

[billisonboard]

Aug 3, 2020 23:02:57 GMT -5 djAdvocate said:

Aug 3, 2020 22:42:11 GMT -5 The Walk of the Penguin Mich said:

I don’t know, it just seems to me that this would be an incredibly skewed sample of people answering.  I know that I have gotten a bunch of texts from local candidates and I am wondering how they got my phone number.  Other than it having a KY area code (the KY texts stopped a few years ago), the only people who get this number are my friends and my bank.  

do you want me to explain why that is incredibly unlikely? or did you intend that remark as a rhetorical one?

I would like it explained.

[djAdvocate]

Aug 3, 2020 23:54:48 GMT -5 billisonboard said:

Aug 3, 2020 23:02:57 GMT -5 djAdvocate said:

do you want me to explain why that is incredibly unlikely? or did you intend that remark as a rhetorical one?

I would like it explained.

first of all, there are a lot of things that influence surveying, but refusal to be surveyed is not one of them.

there are Americans that LIKE being surveyed. and they will give honest opinion. I know that a lot of Republicans don't trust polling, and won't offer their opinion- but there are just as many that are hardheaded, and opinionated, and think they know it all, and will voice those opinions honestly.

where polls can go wildly wrong is bad modeling or a failure to get sufficiently randomized samples. for example, Rasmussen does this thing where they don't use cell phones- only landlines. and they only call once, since they do last minute polling. this slants their results because people who sit near the landline phone tend to follow certain demographics that favor Republicans. this is just one example of where the failure to get a sufficiently random sample will produce poor results.

pollsters have a lot of experience with this stuff, and their methodologies for gathering data are well practiced, and well vetted.  and the best pollsters have the best methods and the best results.  that is why Rasmussen is often IGNORED by poll aggregators.  they simply slant the results too much.  and this is also why 538 tends to underweight polls that are done in a shotgun way like Morning Consult (which is a daily poll).  they simply are not that reliable in data gathering in a randomized way.

it is like any scientific thing.  the better data you get, the better and more predictive the result.

so, yes, a completely random sample of 300 people can precisely predict the result for millions of voters.  I know it sounds utterly crazy, but it is a fact.  the key is to get SUFFICIENTLY random samples, and that is not always easy, but pollsters have techniques for doing it, some of which are closely guarded.  the proof is in the pudding.

i still don't entirely understand this stuff, but the math doesn't lie, and it is generally quite precise, considering how capricious voters can be.  i can't blame people for not trusting it, because of that.  it is tough to trust something you don't understand.

[djAdvocate]

regarding the Emerson poll, i think it is probably accurate in Montana.
i think that the Democrat has lost ground there.
it might not be 8%, but it is significant.
the DNC is going to have to step in if they want to remain competitive.
Montana is not an easy state for Democrats to win.

my reason for thinking it is probably accurate is that they did a national poll the same day- and it shows Biden +7%. that poll, IN MY OPINION, could not be off by more than 2%. if the same methodology was used for data gathering as in MT, it would stand to reason that that poll was ALSO correct within 2%.

of course, there is always the possibility that their data is just wildly off. the LA Times poll showing Trump leading by 3% on election day was do to a sampling error. that poll was pretty consistently off by 5% because of how that poll was conducted. i will explain why, for those who haven't heard. it is a classic case of poor methodology.

most surveys are done randomly for a reason. however, the Times only did random sampling ONCE, to create the initial set. then they called back the same 300+ people every week throughout the campaign. one of their folks surveyed was black. because he represented a group that was undersampled, his vote counted 10x what other surveyors did. and, unfortunately for the LAT, he voted for Trump. what that meant is that according to their survey, 1/3 of the black population would vote for trump, which was an oversample of about 5x. and since they never RE-randomized the sample, the only time that poll was accurate was when that one black Trump voter was not surveyed (absent).

so yeah, stuff like that can go wrong. it often does. but you hope that if you poll, repoll, and poll some more, that eventually you can get a picture of what is going on, despite the random error now and again.

[billisonboard]

Aug 4, 2020 1:18:36 GMT -5 djAdvocate said:

... so, yes, a completely random sample of 300 people can precisely predict the result for millions of voters.  I know it sounds utterly crazy, but it is a fact.  the key is to get SUFFICIENTLY random samples, and that is not always easy, but pollsters have techniques for doing it, some of which are closely guarded.  the proof is in the pudding.

i still don't entirely understand this stuff, but the math doesn't lie, and it is generally quite precise, considering how capricious voters can be.  i can't blame people for not trusting it, because of that.  it is tough to trust something you don't understand.

I understand and accept the results of random sampling. Here is the process that they say they used:

Data was collected using a list of 30,000 mobile numbers provided by Aristotle Inc. A SMS-text to web was sent July 31 to 15,000 numbers and on August 1 to an additional 15,000 numbers. Responses were open until noon August 2.,

That sounds to me like they sent a text message out saying "go to our website and fill out a survey". Out of 30,000 texts sent, just under 500 responses were used to calculate findings (unknown if there were unused/unusable responses]. I question that as being  a sufficiently random sample.

[billisonboard]

Aug 4, 2020 1:41:03 GMT -5 djAdvocate said:

regarding the Emerson poll, i think it is probably accurate in Montana.
i think that the Democrat has lost ground there.
it might not be 8%, but it is significant.
the DNC is going to have to step in if they want to remain competitive.
Montana is not an easy state for Democrats to win.

my reason for thinking it is probably accurate is that they did a national poll the same day- ...

Do you have a link to that national poll? I looked and couldn't find it.

[haapai]

I misread the polling results when I initially read the link.  Somehow, I thought it was a four-point swing in the poll numbers.  Seems now that it was eight points and the predicted outcome definitely changed.  So it actually was news. 

[djAdvocate]

Aug 4, 2020 10:03:34 GMT -5 billisonboard said:

Aug 4, 2020 1:18:36 GMT -5 djAdvocate said:

... so, yes, a completely random sample of 300 people can precisely predict the result for millions of voters.  I know it sounds utterly crazy, but it is a fact.  the key is to get SUFFICIENTLY random samples, and that is not always easy, but pollsters have techniques for doing it, some of which are closely guarded.  the proof is in the pudding.

i still don't entirely understand this stuff, but the math doesn't lie, and it is generally quite precise, considering how capricious voters can be.  i can't blame people for not trusting it, because of that.  it is tough to trust something you don't understand.

I understand and accept the results of random sampling. Here is the process that they say they used:

Data was collected using a list of 30,000 mobile numbers provided by Aristotle Inc. A SMS-text to web was sent July 31 to 15,000 numbers and on August 1 to an additional 15,000 numbers. Responses were open until noon August 2.,

That sounds to me like they sent a text message out saying "go to our website and fill out a survey". Out of 30,000 texts sent, just under 500 responses were used to calculate findings (unknown if there were unused/unusable responses]. I question that as being  a sufficiently random sample.

maybe.  or maybe 10,000 responded and 500 met the sample of sufficiently random.

[djAdvocate]

Aug 4, 2020 10:12:18 GMT -5 billisonboard said:

Aug 4, 2020 1:41:03 GMT -5 djAdvocate said:

regarding the Emerson poll, i think it is probably accurate in Montana.
i think that the Democrat has lost ground there.
it might not be 8%, but it is significant.
the DNC is going to have to step in if they want to remain competitive.
Montana is not an easy state for Democrats to win.

my reason for thinking it is probably accurate is that they did a national poll the same day- ...

Do you have a link to that national poll? I looked and couldn't find it.

sure.  it actually took me quite a bit of time to find it myself.

emersonpolling.reportablenews.com/pr/july-national-poll-biden-maintains-lead-in-presidential-race-majority-support-nationwide-mask-mandate-in-public-spaces

[billisonboard]

Aug 4, 2020 11:32:08 GMT -5 djAdvocate said:

Aug 4, 2020 10:12:18 GMT -5 billisonboard said:

Do you have a link to that national poll? I looked and couldn't find it.

sure.  it actually took me quite a bit of time to find it myself.

emersonpolling.reportablenews.com/pr/july-national-poll-biden-maintains-lead-in-presidential-race-majority-support-nationwide-mask-mandate-in-public-spaces

Thank you. Now

Aug 4, 2020 1:41:03 GMT -5 djAdvocate said:

...
my reason for thinking it is probably accurate is that they did a national poll the same day- and it shows Biden +7%. that poll, IN MY OPINION, could not be off by more than 2%. if the same methodology was used for data gathering as in MT, it would stand to reason that that poll was ALSO correct within 2%.
...

From the link:

Data was collected using an Interactive Voice Response (IVR) system of landlines (n=532) and an online panel provided by MTurk (n=432).

Different methodology.

[djAdvocate]

well, damn. there goes that theory!
thanks for looking into that bills.

conclusion: there is no indication that the Montana poll is accurate.
they used different methodology than a more verifiable national poll.

[bean29]

So, In WI, and several other states it looks like the Republicans will succeed in putting Kanye West on the ballot as an independent. I wonder what that will do to the polling.

[bean29]

It looks like they may have been a few minutes late getting to the WI Election Commission Building to turn in paperwork. They expect it will go to court. They show the woman walking through the door 18 seconds after 5 (5 pm was the deadline).

[thyme4change]

Aug 5, 2020 4:03:26 GMT -5 bean29 said:

So, In WI, and several other states it looks like the Republicans will succeed in putting Kanye West on the ballot as an independent. I wonder what that will do to the polling.

Are they thinking it will drain off the POC vote away from Biden? Or maybe young people?