Posted on 09/17/2012 1:51:18 PM PDT by Leto
A simple rebalance model anyone can use If anyone wants to rebalance polls that are of interest to them, here is a simple model that will work as long as the poll provides a intra-party voting preference. You can cut and paste the formulas or copy to a google spreadsheet. You simple fill in the data matrix on the left and the outcome is the matrix on the right.I have left the "=" sign off of the cells on the right so just add them and you are good to go. This is not mysterious and I think polling would become better and more honest if it was widely used. The example is an ARG Ohio poll that had the sample 42D, 32R, and 26I and O+3%. Since Ohio only went for Obama D+7 over R we can check what the outcome would be if the poll had 38D, 35R, and 32I and the outcome is projected to be Romney+5.33%. UPDATE 9/17: I would also suggest putting this helpful aid in your toolbox Very often a poll will "overlook" or "neglect" to document the Party ID of the poll sample. If any intra-party breakdown is available one can use the referenced calculator to determine the ID numbers the poll uses. It can be any intra-party breakdown not just for voting. It can be for job approval or any other question asked in the poll. If it doesn't produce a commonsense answer there is a disconnect in the poll that should make it's internal consistency suspect.
The thread is here: http://www.freerepublic.com/focus/f-news/2931353/posts
I cam across this site today, he is running a more sophisticated version of the spreadsheet and is updating it daily. I am VERY jealous of this guy and his work.
The site seems to be updated daily. I would recommend that people bookmark this site and check it daily, especially when the Obamedia is pushing their 'this poll says the race is over' narrative.
Actually I do think this race is over and the Fascist in the WH is toast.
http://polls2012.blogspot.com/
Political opinion polls published by the establishment news media have become the modern equivalent of the political cartoon.
Taking his latest numbers, 50.73 Romney, 45.78% Obama, 3.49% Undecided. If one assumes that Romney gets 2/3rd of the undecideds that would make it a 53%-47% Romney.
I find this number interesting since it tracks the Univ of Colorado economic prediction so closely.
““Based on our forecasting model, it becomes clear that the president is in electoral trouble,” said Bickers, also director of the CU in DC Internship Program.
According to their analysis, President Barack Obama will win 218 votes in the Electoral College, short of the 270 he needs. And though they chiefly focus on the Electoral College, the political scientists predict Romney will win 52.9 percent of the popular vote to Obama’s 47.1 percent, when considering only the two major political parties.”
Your post takes me back to 2008, when several Freepers made similar posts. I actually thought going into Election Day that McLame/Palin were going to win because so many “knew” the polls were skewed. I’m sure they are. I just hope they are as skewed as you think.
The stats are based of the party ID polls that Rasmussen has been doing for 10 years. The public polls ignore this and make up the numbers they need to produce the desired result.
So where does the voter fraud get counted in these polls?
We can only overcome the rampant voter fraud if we come out to vote in the millions and millions!!!
I have a sneaking suspicion that that is exactly what will happen though. I use Chick-fil-A as my guideline...
Yup, it is like Deja Vu all over again.
The big difference in 2008 is that Obama did not have a public record, now he is presiding over 14% unemployment and 30% underemployment......
Understanding the numbers involved, the statistics shall set you free from the obamedia.
2008 was an unusual year. When you combined the Bush fatigue, the disenchantment with McCain, the "historic" factor in voting for Obama, and the implosion of the economy, you ended up with a turnout model that favored the Democrats.
But, party identification with the Democrats peaked in 2008. It's been declining ever since, then shifting in favor of the Republicans -- for the first time since Rasmussen started polling back in 2004:
A lot of people thought the polls were skewed unreasonably back in 2008 toward Democrats -- and they were skewed. But, it turned out to be accurate. The party identification peaked at +10 points for the Democrats.
But now, it's +4 for the Republicans, and the poll turnout models aren't reflecting that. Back in October, 2010, it was still +3 for the Democrats, and shifted to +1 for the Republicans in November. Remember how the election turned out?
If the turnout is anything close to the party identification polls, and the independents break for Romney, then Obama doesn't have a chance.
You are quite correct in your recollection on 2008.
So...we have to take freeper condemnation of the polls with a bag of salt.
That said....Rasmussen shows Romney up nationally, and just a hair behind in the states that matter.
With about a two point bump nationally (very doable) and two or three in each of Ohio, VA and FL, Romney will be in good shape.
Remember: he is only down by 3 in Gallup as of today.
McCain was never this close.
Are the polls skewed? Maybe. But ultimately in 2008 the good ones (especially Rasmussen and the RCP average) they were quite right. The propaganda organizations did tend to under poll McCain’s final numbers. But it didn’t matter.
In any event...you are, again, right and this is all just context that we must keep in mind.
An interesting rebalance would be simply to use a 2010 turnout model. Maybe that would exaggerate the GOP turnout and the Dem depression...but still....it would be an interesting adjustment. And obviously Romney would do quite well in that set of numbers.....
Does your scheme make any assumptions about the number of GOPs voting for Obama and the number of Dems voting for Romney?
This is important. If the answer were, for example, 50% of Dems vote for Romney and 50% of GOP vote for Obama then the mix of Dem to GOP doesn’t matter. If the answer is GOP and Dem loyalty is 100% then you should be able to just add the percentages in your head.
Now neither of these extreme cases is true but the number you choose is important and needs to be declared and to be part of the calculation - even if it’s 95% party loyalty.
I don't know if Leto's calculations do so, but the blogsite he cited does make those adjustments.
“McCain was never this close”
Well that just is not factual. Many polls had McCain up as high as 9 pts during the Jeremiah Wright controversy which was right before the “financial meltdown”. It changed the dialogue, played on the only strength that Obama was doing well in the polls, the economy. He never looked back. Interestingly, Obama is up nationally by 15 pts on his “foreign policies” which is why I suspect this ME crisis will also play to his strengths in the polls. Do you believe in coincidences?
It would. I don't think the the 2012 turnout will be the same as the 2010 turnout. 2010 was also an unusual year, due to the Tea Party and ObamaCare. In Presidential election years, there's always a higher turnout, as there are lots of people that don't bother during the mid-terms.
In 2008, it was 57.8%, the highest since 1968. In 2010, it was 37.8% -- the highest since 1994 (when the Republicans took control of both the Senate and the House).
Do these published polls from the major pollsters use ‘weighting factors’? If so do you know the factor used? What would be the impact of such a factor on the published results if used?
“Does your scheme make any assumptions about the number of GOPs voting for Obama and the number of Dems voting for Romney?”
In my personal spreadsheet I do.
In polls where they give the internals they will say have 90% of dems voting for O, 4% Romney and 5% undecided. they reverse for Romney. The turnout model is then applied to the % for each party and the independents.
It looks like the blog works with the raw numbers for the polls.
Good. You did it right. That's a factor that is easily overlooked. To your credit, you didn't.
Actually the turnout number for 2010 was only +1.4% R IIRC, the current ras party affiliation number is +4.3 R, so polls using a +8% D models could be off by 12%.
WHen you look at the spread sheet is is interesting how close the ‘rebalanced’/normalized numbers are...
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.