Welcome, Guest. Please login or register.
Did you miss your activation email?
March 18, 2019, 08:34:37 pm
HomePredMockPollEVCalcAFEWIKIHelpLogin Register
News: Please delete your old personal messages.

+  Atlas Forum
|-+  Election Archive
| |-+  Election Archive
| | |-+  All Archived Boards
| | | |-+  2016 U.S. Presidential General Election Polls (Moderators: AndrewTX, Likely Voter)
| | | | |-+  Atlas Google Consumer Survey polls megathread
« previous next »
Pages: 1 2 3 4 [5] Print
Author Topic: Atlas Google Consumer Survey polls megathread  (Read 7155 times)
Moderate Pennsylvanian
Rookie
*
Posts: 41


View Profile Show only this user's posts in this thread
« Reply #100 on: October 26, 2016, 03:18:19 pm »

Unfortunately, the coupon offer seems to be over.

I will probably do another poll this week or next.  Any suggestions?  I prefer doing polls for rarely-polled states.  I'm thinking maybe Nebraska.  If the sample size is large enough, we might be able to guestimate NE-02.

I'm unable to reply to messages. I'll suggest Mississippi if you run another poll.
Logged
/
darthebearnc
YaBB God
*****
Posts: 7,454
View Profile Show only this user's posts in this thread
« Reply #101 on: October 26, 2016, 03:19:57 pm »

Someone please do as many polls of Utah as possible. It is by far the most interesting state at this point.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #102 on: October 26, 2016, 03:27:25 pm »

Unfortunately, the coupon offer seems to be over.

I will probably do another poll this week or next.  Any suggestions?  I prefer doing polls for rarely-polled states.  I'm thinking maybe Nebraska.  If the sample size is large enough, we might be able to guestimate NE-02.

I'm unable to reply to messages. I'll suggest Mississippi if you run another poll.

That's a possibility. 

As I said in my PM to those who have conducted polls, I'm leaning toward Maine or Nebraska, but would consider 3-6 EV states that we haven't polled - so Idaho, South Dakota, Utah and Alaska are out.  If I did Maine, I'd strive for a big enough sample size to try to break out ME-2 to the extent possible - which it might not be, as not all IP addresses have towns associated with them.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #103 on: October 26, 2016, 06:20:29 pm »

I just put a Maine poll into the field.  Assuming Google agrees to the wording, it will read:
Maine Registered Voters:
If the presidential election were held today, who would you vote for?

Donald Trump of the Republican Party
Hillary Clinton of the Democratic Party
Gary Johnson of the Libertarian Party
Jill Stein of the Green Party
I am not likely to vote in Maine

with all but the not likely to vote choice randomized. 

There will be at least 833 respondents, which might give us 250 usable ME-02 respondents.
Logged
Lok
lok1999
YaBB God
*****
Posts: 3,250
Australia


Political Matrix
E: -1.06, S: -3.02

View Profile Show only this user's posts in this thread
« Reply #104 on: October 26, 2016, 06:25:31 pm »

I just put a Maine poll into the field.  Assuming Google agrees to the wording, it will read:
Maine Registered Voters:
If the presidential election were held today, who would you vote for?

Donald Trump of the Republican Party
Hillary Clinton of the Democratic Party
Gary Johnson of the Libertarian Party
Jill Stein of the Green Party
I am not likely to vote in Maine

with all but the not likely to vote choice randomized. 

There will be at least 833 respondents, which might give us 250 usable ME-02 respondents.
My god, how much did that one cost you?
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #105 on: October 26, 2016, 06:25:50 pm »

I just put a Maine poll into the field.  Assuming Google agrees to the wording, it will read:
Maine Registered Voters:
If the presidential election were held today, who would you vote for?

Donald Trump of the Republican Party
Hillary Clinton of the Democratic Party
Gary Johnson of the Libertarian Party
Jill Stein of the Green Party
I am not likely to vote in Maine

with all but the not likely to vote choice randomized.  

There will be at least 833 respondents, which might give us 250 usable ME-02 respondents.
My god, how much did that one cost you?

Too much.  It's 15 cents per respondent.  Do the math.  It's likely the end of my polling budget for the cycle.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #106 on: October 30, 2016, 08:05:56 am »

cinyc/Google Surveys Maine

Statewide (LV): Clinton +14
Clinton 50%
Trump 36%
Johnson 8%
Stein 5%
October 26-30; 535 probably Likely Voters with Demographic info

Raw: Clinton +12
Clinton 48%
Trump 36%
Johnson 10%
Stein 7%
October 26-30; 671 probably Likely Voters

ME-01 (LV): Clinton +17
Clinton 53%
Trump 36%   
Johnson 7%   
Stein 5%
250 probably likely voters with Demographic and Location Info

Raw: Clinton +15
Clinton 51%
Trump 36%
Johnson 7%   
Stein 6%
323 probably likely voters with location info

ME-02 (LV): Clinton +6
Clinton 46%
Trump  40%
Johnson 9%   
Stein 6%
176 probably likely voters with demographic and location info

Raw: Clinton +3
Clinton 43%
Trump 40%   
Johnson 10%
Stein 7%
238 probably likely voters with location information.
-----------------------

More to come, including a link to my weighting spreadsheet when I put it into Google spreadsheets.  Overall, it's similar to UNH.  I should have had results yesterday, but the last 20 or so responses trickled in very slowly.   Most responses were 10/27-10/29.
« Last Edit: October 30, 2016, 08:14:30 am by cinyc »Logged
dspNY
YaBB God
*****
Posts: 2,593
United States


View Profile Show only this user's posts in this thread
« Reply #107 on: October 30, 2016, 08:07:50 am »

cinyc/Google Surveys Maine

Statewide (LV): Clinton +14
Clinton 50%
Trump 36%
Johnson 8%
Stein 5%
October 26-30; 535 probably Likely Voters with Demographic info

Raw: Clinton +12
Clinton 48%
Trump 36%
Johnson 10%
Stein 7%
October 26-30; 671 probably Likely Voters

ME-01 (LV): Clinton +17
Clinton 53%
Trump 36%   
Johnson 7%   
Stein 5%
250 probably likely voters with Demographic and Location Info

Raw: Clinton +15
Clinton 51%
Trump 36%
Johnson 7%   
Stein 6%
323 probably likely voters with location info

ME-02 (LV): Clinton +6
Clinton 46%
Trump  40%
Johnson 9%   
Stein 6%
176 probably likely voters with demographic and location info

Raw: Clinton +3
Clinton 43%
Trump 40%   
Johnson 10%
Stein 7%
238 probably likely voters with location information.
-----------------------

More to come, including a link to my weighting spreadsheet when I put it into Google spreadsheets.  Overall, it's similar to UNH.  I should have had results yesterday, but the last 20 or so responses trickled in very slowly.   Most responses were 10/27-10/29.

This looks like it's professionally done, nice job with it!
Logged
Castro
Castro2020
YaBB God
*****
Posts: 10,779
United States


View Profile Show only this user's posts in this thread
« Reply #108 on: October 30, 2016, 08:08:56 am »

TBH that looks a whole lot more believable than most of the polling we've seen out of Maine.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #109 on: October 30, 2016, 08:14:10 am »

The reason the CDs don't add up to the statewide numbers is that there were 109 LVs without location info that voted Clinton +25 (weighted).  They mainly came from Google's Mobile App.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #110 on: October 30, 2016, 03:58:23 pm »

Some UNWEIGHTED Crosstabs from Maine:

GenderPCT TrumpPCT ClintonPCT JohnsonPCT Stein
Male39.60%41.61%11.74%7.05%
Female28.74%58.62%7.66%   4.98%
Unknown43.75%38.39%8.04%9.82%
Total36.07%47.69%9.54%6.71%
            
AgePCT Trump   PCT ClintonPCT JohnsonPCT Stein
18-24     35.71%41.43%12.86%10.00%
25-34     27.96%48.39%13.98%9.68%
35-44     29.03%47.31%17.20%6.45%
45-54     49.44%41.57%4.49%4.49%
55-64     36.84%51.58%6.32%5.26%
65+        33.33%59.38%4.17%3.13%
Unknown39.26%43.70%8.89%8.15%

            
GeographyPCT TrumpPCT ClintonPCT JohnsonPCT Stein
Urban37.21%50.00%4.65%8.14%
Suburban37.77%47.21%9.87%5.15%
Rural   34.66%47.44%10.51%7.39%
         
            
IncomePCT TrumpPCT ClintonPCT Johnson   PCT Stein
$0-$24,99922.22%29.63%25.93%22.22%
$25,000-$49,99936.71%48.95%8.41%5.93%
$50,000-$74,99941.77%40.51%8.86%8.86%
$75,000-$99,9999.09%72.73%9.09%9.09%
$100,000-$149,999   28.57%57.14%14.29%0.00%
$150,000+50.00%33.33%16.67%0.00%
Unknown27.27%54.55%18.18%0.00%
            
            
Parental StatusPCT TrumpPCT ClintonPCT JohnsonPCT Stein
Parent47.20%42.40%6.40%4.00%
Non-parent33.66%53.14%8.25%4.95%
Unknown33.33%43.62%12.76%10.29%
-----------------------------
Age and gender unknowns are not included in the weighted LV numbers.  Note that they are generally more Trump-favorable than the knowns.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #111 on: October 30, 2016, 04:27:23 pm »

Here's a chart of how the weighted voted came in over time.  Margin is Trump minus Clinton, so Clinton lead if it was negative.  This is cumulative, i.e. the new votes for the hour were added to the prior votes in the chart.  Note that Trump was actually up for a short period of time, and ME-02 was competitive over much of the poll.  The votes that came in on Saturday changed the competitiveness of ME-02.  (Click for a larger view)

Img
« Last Edit: October 30, 2016, 05:06:59 pm by cinyc »Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #112 on: October 30, 2016, 04:56:54 pm »

Atlas Google Consumer Surveys so far:
StateSponsorMarginWeighted toDates
UtahamdcpusTrump +6Likely VotersOct 12-13
South DakotacinycTrump +17Registered VotersOct 15-17
TennesseeExtremeRepublicanTrump+33Registered VotersOct 18-20
Idahoanthony1691Trump +23UnweightedOct 18-21
Georgiarafta_raftaTrump +11Unweighted?
UtahSpeed of SoundTrump +3Likely VotersOct 19-22
AlaskaModerate PennsylvanianTrump+5Likely VotersOct 19-22
WashingtonAlconClinton +10Likely VotersOct 21-23
MainecinycClinton +14Likely VotersOct 26-30
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #113 on: October 31, 2016, 02:00:38 am »

I created a new weighting template, which is much more user-friendly than the old one.  With Jimrtex's help, I also calculated separate coefficients for every category in every state, including the 45-54 and 55-64 categories.  Men and women in each category have different coeffiicients, too, instead of assuming the male/female percentage of every age subgroup breaks out according to the overall percentage in all groups.

I also added a feature to weight those without demographic information - if you choose to do so.  You can weight them 0 to not include them, 1 as average or any number you want.  Believe it or not, this sometimes greatly affects the overall results.

The major difference from the old template is that when you input the raw results data, you should now input the NUMBER of raw votes for each candidate in each category, not the percentage.  After inputting that information and the correct coefficients for your state (included in the second sheet of the spreadsheet), the spreadsheet should calculate everything else automatically.

The new template is here.

I'm finding some links to other Google Surveys commissioned by non-Atlas posters on 538.  Should we discuss them here or in a separate thread?
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #114 on: November 06, 2016, 10:52:24 am »

cinyc/Google Survey New Mexico Poll

LV Weighted: Clinton +15
Clinton 45%
Trump 30%
Johnson 17%
Stein 5%
McMullin 4%

330 probably LV with imputed age and sex info; November 4-6

Raw: Clinton +9
Clinton 42%
Trump 33%
Johnson 17%
Stein 4%
McMullin 4%

391 probably LV; November 4-6

If you weight those without demographic info evenly (weight=1), Clinton's lead shrinks to 10.  They were pro-Trump, unlike those with demo info.

More to come.
-----------------------------
So one of these polls is not like the others, showing the race close.  Who will be wrong?  Likely me.  But we'll have to wait until election night.

All polling was done in English-only, which is a limitation with general Google Surveys (you can do a bilingual Android Mobile App survey, but those skew too young to give reliable results).
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,673
Sweden


View Profile Show only this user's posts in this thread
« Reply #115 on: November 06, 2016, 10:56:58 am »

It actually looks more "conventional wisdom" than other polls Cheesy
Logged
Speed of Sound
LiberalPA
YaBB God
*****
Posts: 14,198
United States


View Profile Show only this user's posts in this thread
« Reply #116 on: November 06, 2016, 11:07:05 am »

Interesting numbers, cinyc, especially with how close it started. Thanks for much for running this one, especially so close to the election!
Logged
BeTo In DiSaRrAy
Fubart Solman
YaBB God
*****
Posts: 4,646
United States


View Profile Show only this user's posts in this thread
« Reply #117 on: November 06, 2016, 11:11:37 am »

Interesting that McMullin does so well in NM. I wouldn't've expected that.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #118 on: November 06, 2016, 11:28:10 am »

Interesting numbers, cinyc, especially with how close it started. Thanks for much for running this one, especially so close to the election!

I'm going to do a graph of the results over time, if I have time later today.  One reason it was likely close in the beginning is that Google Surveys usually starts with respondents from its mobile app, and they broke more for Trump than those from other sources.  Usually, they're more Democratic-leaning - but not in New Mexico. 

As you know, I also didn't start weighting the results until about 100 respondents were in, but once I didn't, I noticed that it really was a double-digit Clinton lead at that time because they were significantly under sampling women, who were much more pro-Clinton than men.
Logged
cinyc
YaBB God
*****
Posts: 11,617


View Profile Show only this user's posts in this thread
« Reply #119 on: November 09, 2016, 09:16:29 pm »

Atlas Google Consumer Surveys so far:
StateSponsorMarginWeighted toDatesActual Results
UtahamdcpusTrump +6Likely VotersOct 12-13Nope.  It's Trump +19 so far, but Salt Lake County is partially out.
South DakotacinycTrump +17Registered VotersOct 15-17Nope.  Trump +30
TennesseeExtremeRepublicanTrump+33Registered VotersOct 18-20Close.  Trump +26
Idahoanthony1691Trump +23UnweightedOct 18-21Understated  It's Trump +31
Georgiarafta_raftaTrump +11Unweighted?Overstated.  Trump +5
UtahSpeed of SoundTrump +3Likely VotersOct 19-22Nope.  It's Trump +19 so far, but Salt Lake County is partially out.
AlaskaModerate PennsylvanianTrump+5Likely VotersOct 19-22Nope.  Trump +15
WashingtonAlconClinton +10Likely VotersOct 21-23Nope.  Clinton +18
MainecinycClinton +14Likely VotersOct 26-30Lol.  Clinton +3
New MexicocinycClinton +15Likely VotersNov 4-6Overstated.  Clinton +8.  Unweighted and weighting those with no demo info was closer than most NM polls
Logged
Pages: 1 2 3 4 [5] Print 
« previous next »
Jump to:  


Login with username, password and session length
Logout


Terms of Service - DMCA Agent and Policy - Privacy Policy and Cookies
 

Powered by SMF 1.1.21 | SMF © 2015, Simple Machines