I have Romney ahead in Ohio. The polls have Obama ahead. Who is right/wrong???

[Will]

Ohio is producing an interesting dilemma for me.

While the national polls, and to a lesser degree polls in hugely important swing states like Florida and Virginia, are beginning to produce top-line numbers that fall in line with what I have been showing (Romney surging into a lead) one key state, THE key state by most accounts, has been a different story:

Ohio

On Friday October 12th my tracker saw Romney finally take the lead, albeit a minuscule one at 0.2%.

In the RCP average Obama led by 1.3%

On Friday October 19th, Romney moved to a 1.4% lead in my Ohio tracker.



Obama led by by 2.5%

So, Romney lead getting larger for me, Obama lead getting larger across the polls in RCP's average.

As of today, RCP has Obama +2.1%. 5 polls with an Obama lead, 3 with a tie, NONE with Romney ahead.

Using the 7 most recent of these polls which publish crosstabs, I have Romney ahead by 2.2% as of today (next update on Friday).

Why is this happening?

Well, because the sampling in these polls is generally, on average, showing a Democratic turnout advantage VERY similar to 2008. In fact, my results based on 2008 voter ID show Obama +3%. My results based on other election years, even the 2006 Democratic beatdown mid-term elections, have Romney ahead.

So.....

I can't know for sure what the actual electorate will look like. But I can do a little more research to illuminate Ohio.

Most 2008 polls are unavailable, but I found 2 from around this time in 2008 that were still up for Ohio: Survey USA and Marist:

SurveyUSA Election Poll #14750

http://www.realclearpolitics.com/RCP..._OH_102908.pdf

They showed Obama up by 8% and 9% respectively with INDEPENDENTS.

CNN's Exit Polling from election day showed Obama winning Ohio independents by, you guessed it, 8%:

Local Exit Polls - Election Center 2008 - Elections & Politics from CNN.com

The 7 most recent Ohio polls in my tracker show and average Romney lead of 49.3% to 37.6% among independents, an 11.7% lead.

So while I cannot know for sure whether the polls, which seem to be estimating turnout almost equivalent to 2008, at Democrats +8%, or my model, which is estimating a much closer Democrats +1.2% (Republicans won in 2010 by 1%, by 4% in 2004, and lost by 3% in 2006) has the most correct prediction of actual turnout, I DO know the following:

*Polls in 2008 seemed to have accurately predicted the independent vote.

*Polls in 2012 are showing a nearly 20% swing among independents, from Obama ~ +8% to Romney ~ +11%

*If I mess with the turnout model, the worst case-scenario under which Romney can win is Democrats +4% in turnout, down 4% for them from 2008, but up 5% from their 1% loss in 2010.


I cannot wait to see who is right.

(I hope it's me. )

[Giraffe (He/Him)]

Once this is decided, are you going to go back and look at your data and see what was what? Just outta curiosity?

[Will]

Originally Posted by Thomas

Once this is decided, are you going to go back and look at your data and see what was what? Just outta curiosity?

YES!!!

Especially Exit Poll Date by party ID vs. final results in my aggregators, as well as what turnout actually looked like.

For my national tracker, and individual trackers, I have been using a model that is simply the average turnout from 2004-2010 elections.

I did this nationally first, because the result it gave me made sense to me, somewhere between 2008 and 2010.

I just kept the same method for each individual state, even when my guess as to what turnout will be differed. I wanted uniformity, and I wanted to remove the human element (me) from the equation. So whether I like the resulting make-up or not, my official party ID numbers for each state are based on average party ID from 2004-2010.

I am still trying to decide whether or not I will make a final prediction independent of what my models show as a result. I will of course relay what the models say as well.

[mrvette]

Bluntly said, I am torn between early voting here at the local library, or waiting to vote at some community center on election day....


honestly, I WANT a physical confrontation with the 'black activists' types....

panthers, whoever they call themselves......

bet I pack more than they do......try me, bastards.....

[Taurus]

They do, but as of late they have dropped their weighting of Dems to 4 or 5 points. The problem is that the early polls from several weeks ago that were so badly skewed makes the average look off.

[Joecooool]

Election Forecasts - FiveThirtyEight Blog - NYTimes.com

[Taurus]

Originally Posted by Joecooool

Election Forecasts - FiveThirtyEight Blog - NYTimes.com

NY Times, really? How about the poll that was the most accurate in 2008 & 2010;

2012 Electoral College Scoreboard - Rasmussen Reports™

Probably closer to reality.

[Sea Six]

Originally Posted by mrvette

Bluntly said, I am torn between early voting here at the local library, or waiting to vote at some community center on election day....


honestly, I WANT a physical confrontation with the 'black activists' types....

panthers, whoever they call themselves......

bet I pack more than they do......try me, bastards.....

It's idiotic statements like this that eliminates any and all possibility of using self-defense laws as a legal defense in a shooting incident as you have described.

It also makes law-abiding CCW permit holders look bad in the eyes of the public.

And last, but certainly not least... posting such tripe in an online public forum ensures that your premeditated intentions not only can, but will be used against you in a court of law; likely, in a first degree murder trial.

Govern yourself accordingly.

[Will]

Originally Posted by Joecooool

Election Forecasts - FiveThirtyEight Blog - NYTimes.com

He was VERY accurate in 2008, but only because the polls, averaged together, were very accurate. The RCP average was only tenths of a point off of the actual result, Obama +7.2%.

He simply uses polls' top-line results, not their internal data.

Also, his model uses outside factors like economic data from the individual states.

His model is basically a hybrid of an academic election prediction models, and a poll aggregation model.

In other words, very subject to being wrong, just as basically about 1/2 of all existing models will be proven wrong, because the academic models are pretty split about who will win.

I do like his site though. Sometimes he does some very good analysis. And he is a statistical genius. We'll see whether he's right or wrong this time. That, and we'll see if he continues to show an Obama win up through Nov. 5th.

[Giraffe (He/Him)]

honestly, I WANT a physical confrontation with the 'black activists' types..

You're what, 68 years old? You're gonna have to set your walker down, find your glasses and grab a seat before you can upholster your gun.

Gene, really?

[Fasglas]

Originally Posted by Sea Six

Govern yourself accordingly.

Tyler?

[Giraffe (He/Him)]

It also makes law-abiding CCW permit holders look bad in the eyes of the public.

In other words; Get off my side.

[Joecooool]

Originally Posted by Will

He was VERY accurate in 2008, but only because the polls, averaged together, were very accurate. The RCP average was only tenths of a point off of the actual result, Obama +7.2%.

He simply uses polls' top-line results, not their internal data.

Also, his model uses outside factors like economic data from the individual states.

His model is basically a hybrid of an academic election prediction models, and a poll aggregation model.

In other words, very subject to being wrong, just as basically about 1/2 of all existing models will be proven wrong, because the academic models are pretty split about who will win.

I do like his site though. Sometimes he does some very good analysis. And he is a statistical genius. We'll see whether he's right or wrong this time. That, and we'll see if he continues to show an Obama win up through Nov. 5th.

He is reporting today that Romney's momentum has stopped.