Posted on 09/14/2012 9:03:41 PM PDT by Leto
Polls have caused a lot of consternation among my friends here. Some say to ignore polls, but this election season I have found them to be very consistent even thought the results vary quite a bit.
The key is understanding the turnout model the poll is using. In 2008 the turnout was +7.6% dem, in 2010 it was +1.3% R. In the last Rasmussen survey 8/2012 the party affiliation is +4.3% R.
Most polls are using turnout models that vary from +6 to +10% D. This is why Obama is leading in most polls.
The latest NYT/CBS poll has Obama +3 among likely voters 49-46%. Their turnout model is +6% D close to the 2008 turnout model. 35D/39R/32I for their likely voters in this poll.
I have a spreadsheet I use to see the results if the 2010 turnout model is used or any other model.
Below is the CBS NYT poll 'normalized' to reflect both the 2010 turnout model and the current Rasmussen party id.
I think the pollsters are wrong using 2008 percentages for sampling now
but I would worry more about the last week of polling baring a big surprise..which would be trailing and not caught likely...2000 missed a bit of Bush’s DUI final days which almost cost him the enchilada
if the election were to diverge significantly from the aggregate of the final week’s polling it would be the first time since..well ..Dewey
Romney does not excite so I do not think we will attain your 2010 model...and Obamacare is not as fresh
The waxing the Dems would take in 2010 was clearly foretold by the polls..but I think we beat the polls by a couple of points turnout wise
But I do think we will beat them over 2008 Dem-GOP turnout
if 59% of vote is white we win...any less and we better have a low minority turnout of both dead or alive voters there
very good post !
CNN/ORC poll
50.4 percent Democrats and
45.4 percent Republicans
4.2 percent independents.
I find it unfortunate that too many posters on Free Republic are so quick to look for Democrat poll oversampling that such a sloppy analysis of the CNN/ORC poll, September 7-9 2012, PDF - Page #21 of 48, (Question 1/1A) by the Examiner is taken as gospel. Only 4.2% "Independents" in the CNN/ORC Likely Voter sample, really? The published data supports no such conclusion.
Demo- Indep- Repub-
Total crat endent lican
----- ----- ------ ------ -------- -------- --------
Obama,Biden, lean 52% 97% 40% 2%
Romney,Ryan, lean 46% 3% 54% 96%
Other * * * *
Neither 2% * 4% 2%
No opinion 1% * 2% *
Sampling Error +/-3.5 +/-6.0 +/-6.5 +/-6.5
The CNN/ORC question 1/1A table reproduced above has an unstated 95% confidence level (most poll do) for the (n~709) sample size of Likely Voters. The CNN/ORC table for the Total Likely Voter response has the margin of error (or confidence interval) of (+/- 3.5). This published sampling error for the Likely Voters is a rounded number as with the sample size of (n=709) the MoE is actually (+/- 3.68). Now that it is known that the CNN/ORC published MoE numbers are rounded, check the Democrat (+/- 6.0) sub-sampling error, this should be a sample size of 267. The Republican and Independent sub-sample error number of (+/-6.5) would both be derived from a sample size of 227. Since these published sample error number are rounded, they only indicate a range of sample sizes, depending on what the amount of rounding that was performed by CNN/ORC. The published sampling error number alone show that the sample contained more Democrats than either Republicans or Independents.
Rounded MoE (+/- 6.0) Rounded range 6.24 - 5.75 Sample # range 247 - 290 Rounded MoE (+/- 6.5) Rounded range 6.88 - 6.25 Sample # range 203 - 246
Given the above Likely Voter sample data the following analysis follows:
Here is how I see the political ID poll data breakdown for the CNN/ORC national poll published September 10th, 2012:
The original national CNN data, 709 Likely Voters for September 07-09, 2012:
The derived poll data presented below for the same national CNN, Likely Voters, n = 709
Rounding the below by plus or minus (0.49%) yields above published (rounded) CNN two digit poll data.
Looks as if CNN/ORC have published a Democrat [+10.36%] Likely Voter oversampled poll, with Independents comprising around 32.3% of the likely voter sample. Just for fun, at the far right of the spreadsheet data is a "what-if" the political ID breakdown was (35%R, 35%D, 30%I) - Romney on top by ~4%.
dvwjr
There have been a lot of good points raised here.
This is responds to a bunch of different posts.
The polls will be more accurate in the last week or so before the election. The pollsters will adjust their voting models so that the ‘result’ will be accurate. This will be spun as dramatic changes which usually don’t occur.
Presidential Elections are about the incumbent. As long as Romney don’t do something stupid, this election will be a referendum on OBAMA. The Obama approval ratings have been in the mid 40’s, his vote total will closely mirror his approval rating.
The dem registration has gone down in the swing states by 800k since 2010, the R registration is down 80k. This doesn’t bode well for the dems in this election.
There is also an enthusiasm gap between R’s and D’s that favors the R’s.
These things point to a turnout model that will IMO be as good or better than 2010. I think people really don’t like Obama.
State Polls tend to be less accurate than national polls, not sure why. However if Romney wins nationally by say %5 in the popular vote there is no way he loses the electoral college.
The campaigns get polls that are far more accurate than anything in the public polls. If you look at the campaigns their behavior tells you a lot about what they are ‘seeing’ in their internal polls.
The Obama campaign is going totally negative, wild charges OTOH Romney is being soft in their critique. IMO this indicates who thinks they are winning and who is desperate.
The media also understands that Obama being behind by double digits with independents means he is toast unless something changes. Note the under sampling of Indies in some polls (along with the under sampling of R’s in almost/all polls), and the lack of mention of Indies in the reporting of these polls.
Do you think that The media would report the results of independents if Obama was UP by double digits???? ;)
dvwjr great job on the CNN poll, like I said if you ‘normalize’ the turnout model the polls are consistent and show Romney with a big lead, that can move to a landslide.
IMO the money advantage for Romney over the last 30 days will lock down these numbers.
Excellent post! Thanks!
You are right. I try, and it’s sometimes difficult when the temptation is so great to beg him to double down on his awesome press conference. But regardless you’re right...
tag for later
I agree. The thrill for Obama is gone among some black voters and plenty of college students. The black voters, because that barrier has been broken, and college students, especially recent graduates, because the economy is crap and they haven't been able to find good jobs in their field, or any other, for that matter.
http://www.breitbart.com/Big-Government/2012/09/14/surprise-oversampling-dems-puts-obama-in-lead
Anyone interested in understanding the polls should check him out on a regular basis.
http://i1.cpcache.com/product/690827672/seals_removed_one_threat_rectangle_magnet.jpg
Thanks for the links. I’ll check them out!
ping for later
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.