Welcome, Guest. Please login or register.
Did you miss your activation email?
August 21, 2019, 04:09:32 am
News: 2020 Presidential Predictions (General) are now active.

  Atlas Forum
  Election Archive
  Election Archive
  2016 U.S. Presidential General Election Polls (Moderators: AndrewTX, Likely Voter)
  IBD/TIPP Daily Tracking Poll thread FINAL - Trump + 1/ Trump + 2 (search mode)
Pages: [1] Print
Author Topic: IBD/TIPP Daily Tracking Poll thread FINAL - Trump + 1/ Trump + 2  (Read 10829 times)
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« on: October 20, 2016, 06:15:07 am »

Head-to-head, 10/20::
Clinton     43.1% (-0.5%)
Trump     41.0%  (+0.4%)
Other       5.0%   (-0.5%)
Not Sure 10.8%   (+0.5%)
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #1 on: October 20, 2016, 06:48:19 am »

Best thing about them going to a tracker, like Rassy, is they essentailly diminish their impact on the various election projection models (538, upshot, etc).
Ehm, no. They will have exactly the same impact. But it will be spread out the impact on each day rather than once in a month.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #2 on: October 20, 2016, 07:06:45 am »

Best thing about them going to a tracker, like Rassy, is they essentailly diminish their impact on the various election projection models (538, upshot, etc).
Ehm, no. They will have exactly the same impact. But it will be spread out the impact on each day rather than once in a month.

Nate Silver has said in the past that trackers are weighted less than regular polls.

Just look at http://projects.fivethirtyeight.com/2016-election-forecast/national-polls/. Daily trackers have descent weights.
If you compare similar polls, for example Ipsos and Quinnipac (they have the same rating A-, approximately the same sample size and the dates) have almost the same weights, even though Ipsos is a daily tracker. Nate probably meant that they threat them as a weekly trackers.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #3 on: October 22, 2016, 06:52:41 am »
« Edited: October 22, 2016, 06:54:33 am by Happy Sad Trumpista »

do all trackers include such strange samples?

still trump +1 but....

http://www.investors.com/politics/ibd-tipp-presidential-election-poll/

clinton leading in the northeast by 9, midwest by 1, west by 6...trails south by 11 and this results into a trump lead?

other nuggets:

catholics are more anti-clinton than evangelicals, old people like clinton more those between 45 and 64, hispanics are closer than the last 2 elections.

All of that is normal and expected.


How? You get weird sub-samples, but dude, c'mon.
Small subsampbles. Poll has ~700LV, meaning that each subsample is about 100-200 LV. And this should be weighted as well. By chance can 5 old African-Americans all coming from Northeast etc. Trump's African American from LA Times is an extreme example of that. He got a 30 weight which is a methodological issue, but even in "normal" "A" pollsters some groups get 3-4 weights (young non-whites), while others (old whites) get 0.3.

This together with small samples gives weird results. Ipsos 50-nationall survey is a good example, some small states with small samples give strange result, while on average nationall lead is consistent with 538 average. Monmouth and subsamples of SUSA are another good examples. Monmouth is one of the best pollsters on average, but their small samples caused sometimes weird swings.

Upshot wrote a good article http://www.nytimes.com/2016/09/19/upshot/introducing-the-new-york-times-upshot-siena-college-poll.html

Pollsters do not publish subsamples to give some deeper insights, no. They publish it in first place for transparency.


I'm not defending this particular poll. I believe in average and think that 538/Upshot gives a fair picture of where the race stands.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #4 on: October 22, 2016, 11:33:22 am »

IBD/TIPP may have been the most accurate if you average past 3 elections but that's due to their performance in 2004 and 2008. In 2012, they had nearly +3 Republican bias.
In 2012 everyone had about 2-3 R bias. They are good, but are clearly an outlier this year/right now. No pollster is perfect in each election. Nothing strange here.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #5 on: October 24, 2016, 06:44:14 am »

Intresting. Both LA Times, IBD and probably ABC polls shows a further dive of Trump's support recently. I'll watch Gallup fav/unfav update today.

Election-is-rigged/I-won't-accept-the-results effect? It will be a really fun week Cheesy
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #6 on: October 24, 2016, 10:47:29 am »

IBD is definitely going to show a major swing towards Clinton right before the end so they can keep up their "most accurate pollster" shtick, mark my words.

2012 is not a good example because most polls showed the race as dead heat so I checked out 2008.

IBD herded hard-core in 2008. They were, on average, showing lower margin for Obama only to join the crowd toward the end.

Check out 2008 numbers for the last few weeks.

Diageo/Hotline   10/20 - 10/22   769 LV   3.5   48   43   Obama +5
Reuters/C-SPAN/Zogby   10/20 - 10/22   1206 LV   2.9   52   40   Obama +12
ABC News/Wash Post   10/19 - 10/22   1335 LV   2.5   54   43   Obama +11
IBD/TIPP              10/18 - 10/22   1060 LV   3.5   45   44   Obama +1
GWU/Battleground   10/16 - 10/22   1000 LV   3.5   49   45   Obama +4
FOX News   10/20 - 10/21   936 LV   3.0   49   40   Obama +9
CBS News/NY Times   10/19 - 10/22   771 LV   --   52   39   Obama +13


ABC News/Wash Post   10/23 - 10/26   1314 LV   2.5   52   45   Obama +7
Pew Research           10/23 - 10/26   1198 LV   3.5   53   38   Obama +15
IBD/TIPP              10/22 - 10/26   824 LV   3.5   47   44   Obama +3
GWU/Battleground   10/20 - 10/26   1000 LV   3.5   49   46   Obama +3
Newsweek                    10/22 - 10/23   882 LV   4.0   53   41   Obama +12

But somehow their average became more realistic toward the end.

Rasmussen Reports        11/1 - 11/3   3000 LV   2.0   52   46   Obama +6
Reuters/C-SPAN/Zogby  11/1 - 11/3   1201 LV   2.9   54   43   Obama +11
IBD/TIPP                   11/1 - 11/3   981 LV   3.2   52   44   Obama +8
FOX News                        11/1 - 11/2   971 LV   3.0   50   43   Obama +7
NBC News/Wall St. Jrnl   11/1 - 11/2   1011 LV   3.1   51   43   Obama +8
You are cherry-picking. Other polls during the same period:

IBD/TIPP
IBD/TIPP   11/1 - 11/3   981 LV   3.2   52   44   Obama +8
IBD/TIPP   10/22 - 10/26   824 LV   3.5   47   44   Obama +3
IBD/TIPP   10/18 - 10/22   1060 LV   3.5   45   44   Obama +1
IBD/TIPP   10/14 - 10/18   1072 LV   3.0   47   42   Obama +5

Gallup
Gallup         10/31 - 11/2   2472 LV   2.0   55   44   Obama +11
Gallup (Traditional)*   10/30 - 11/1   2503 LV   2.0   51   43   Obama +8
Gallup (Expanded)*   10/30 - 11/1   2475 LV   2.0   52   43   Obama +9
Gallup (Traditional)*   10/24 - 10/26   2446 LV   2.0   50   45   Obama +5
Gallup (Expanded)*   10/24 - 10/26   2343 LV   2.0   53   43   Obama +10
Gallup (Traditional)*   10/20 - 10/22   2399 LV   2.0   50   46   Obama +4
Gallup (Expanded)*   10/20 - 10/22   2349 LV   2.0   51   45   Obama +6
Gallup (Traditional)*   10/16 - 10/18   2590 LV   2.0   49   46   Obama +3
Gallup (Expanded)*   10/16 - 10/18   2277 LV   2.0   51   44   Obama +7


Pew
Pew Research   10/29 - 11/1   2587 LV   2.0   52   46   Obama +6
Pew Research   10/23 - 10/26   1198 LV   3.5   53   38   Obama +15
Pew Research   10/16 - 10/19   2382 LV   2.5   53   39   Obama +14

Gallup showed approximately the same trend from Obama ~+5 to ~+10
Pew showed a "hardboard herding" as you said.


IIRC, 538 calculate their rating based on last 4 weeks of elections. So it should at least take care of "final-week herding". In theory.

P.S. Right now IBD/TIPP is heavily off. No doubt.
That's why one should average.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #7 on: October 26, 2016, 05:06:43 am »

Weighting by Party ID is indefensible.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #8 on: October 28, 2016, 05:47:42 am »

2% swing towards Hillary today.

C+3.0    4-way
C+3.4    2-way

Quote
You must be logged in to read this quote.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #9 on: October 29, 2016, 07:03:13 am »

Average and move on. Both were outliers and now "converging" on average to "true value".
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #10 on: October 30, 2016, 06:14:10 am »



Clinton    Trump   Johnson   Stein      Other   Not sure
44.1%   41.6%   5.6%      2.4%      2.3%      4.0%

Two-way

Clinton   Trump   Other   Not sure
44.9%   41.3%   5.3%      8.5%
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #11 on: November 01, 2016, 06:16:08 am »

You just have to read the editorial to understand how objective the people behind this poll are.
First of all, "Just last week, Clinton appeared to be building an insurmountable lead in both the popular vote and the electoral college"?
Really guys? Just last week you were showing Trump leading Clinton. Unless of course you admit that your poll is crap and we should trust the other ones.

Second, they find that a majority of voters find Trump "honest and trustworthy". Not compared with Clinton but on his own right. There is literally not one poll to be even close to that, he usually polls as bad as her when it comes to that characteristic.

And third, they are talking about "the reopening of the FBI investigation regarding Clinton's e-mails" when nothing like that has happened, just parroting Jason Chaffetz's talking points.

Read PPP:s stupid Trump vs something-ridicolous question. Still no Red hacks bashing to be partisan that "fake" their data...
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #12 on: November 02, 2016, 06:02:45 am »

1.3% points change in 4-way and 1.0% in 2-way.

4-way
Trump 44.4% (+0.7)
Hillary 44.0% (-0.6)
Johnson 4.1%
Stein 2.2%.

In the two-way poll
Trump  44.2% (+0.3)
Clinton 44.0% (-0.7)

http://www.investors.com/politics/clinton-and-trump-are-tied-with-election-just-5-days-away-ibdtipp-poll/

Quote
You must be logged in to read this quote.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #13 on: November 06, 2016, 06:15:19 am »
« Edited: November 06, 2016, 06:17:42 am by Little Big BREXIT »

Trump Leads Clinton By 1 Point With 2 Days Left

Trump +1 in 4-way

Clinton +1,5 in 2-way

Quote
You must be logged in to read this quote.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #14 on: November 07, 2016, 07:01:54 am »

Haha, what? But good news anyway for Trump Roll Eyes


Not really.  538 uses national polls to set trend and state polls to set level of support.  Will have tiny impact

The rating of IBD/TIPP
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #15 on: November 08, 2016, 02:41:09 am »

Will the update today? Wounder if the find out that the race is at Clinton +4 Roll Eyes
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #16 on: November 08, 2016, 04:21:47 am »
« Edited: November 08, 2016, 04:26:23 am by Erich Maria Remarque »

They will very likely be 3-6% points off, but I have to give them the credit that they have balls to go with this Smiley

4-way: ('Not Sure' Allocated)
Trump   45.0%
Clinton  43.4%
Johnson  7.6%
Stein      2.0%
Other      2.0%

2-way:
Clinton 43.3%   
Trump  42.3%   
Other     5.3%
Not sure 9.0%


4-way: ('With Not Sure')
Clinton 41.2%   
Trump 42.7%   
Johnson 7.1%    (+2)!
Steing 1.9%   
Other 2.0%   
Unsure 5.0%


Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #17 on: November 09, 2016, 02:36:15 am »

They will very likely be 3-6% points off, but I have to give them the credit that they have balls to go with this Smiley

Wow! Balls of steel!
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #18 on: November 09, 2016, 05:33:58 am »

Inside MOE.  Almost all other showed an embarrassing herding.
Logged
Erich Maria Remarque
LittleBigPlanet
YaBB God
*****
Posts: 3,654
Sweden


« Reply #19 on: November 09, 2016, 09:16:00 pm »

as wrong as anyone else.....only the other direction.

Nope. The difference is that they didn't herd. That had race ~Tie during final week, and this was pretty close. Clearly, many pollster firm, thought that Clinton will surge (after 2nd letter???) and added a few points/herded to C+4-5. It is almost impossible that all polls had so small spread in their results.
Logged
Pages: [1] Print 
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

© Dave Leip's Atlas of U.S. Elections, LLC