The Phantom's Polling Analysis Thread

[BluePhantom]

More unbiased data from a Romney supporter?

Well considering I just told Ferret to avoid partisan comments and agreed that using only landlines will probably create a Republican advantage...I'd say I am being pretty objective. Maybe you ought to wait until I say something that is not objective before you get snippy about it

 

[PLYMCO_PILGRIM]

More unbiased data from a Romney supporter?

Data, in itself, does not contain bias. It only contains information.

Its how that data is used and presented that lets bias be inserted.

What specifically about how bluephantom presented the polling data led you to believe this was a partisan type of thread? (by saying what I quoted you, in effect, are accusing the op of having partisan motives)

 

[BluePhantom]

That 'likely voter' stat is telling....

Yeah I am looking closely at that. But I am also noticing that the only two tracking polls are both showing Romney leading. That can't be just RV vs. LV because Rasmussen is LV and Gallup is RV. I am wondering whether there is something in the nature of a tracking poll itself that would create advantage.

Initially I wouldn't think so. All they are doing is averaging them out over a time frame: three days for Rasmussen and five for Gallup. That means they are slower to react but also more consistent and less subject to statistical noise.

I can't really think of anything that would create that effect. Anyone else?

 

[RightWingFerret]

these next six months will be the indicator. dont expect any major increase in hiring. if anything, we will be stagnant being small/medium business are waiting for the election results,,,if Romney wins, they will gradually start hiring,,,if Obama wins,,,there goes the stock market and wealth will start leaving the country.

 

[BluePhantom]

Everyone, please......we have already gotten off to a rough start here. I am REALLY hoping this thread runs continuously through the elections and we can analyze in an unbiased fashion. We have already seen what happens, on the Polling 101 thread, when one person wants to make a political point instead of simply looking at the data. I really want to avoid that here...we can get nasty on other threads (and believe me I will be jumping in with guns blazing), but let's really try to keep things objective and civil here.

Please. I am asking nicely.

Thank you.

-BP

 

[BluePhantom]

these next six months will be the indicator. dont expect any major increase in hiring. if anything, we will be stagnant being small/medium business are waiting for the election results,,,if Romney wins, they will gradually start hiring,,,if Obama wins,,,there goes the stock market and wealth will start leaving the country.

Well what may or may not happen if Obama is elected vs. Romney being elected is best argued on a different thread. Now I will certainly agree that if the economy does not show continuous improvement, even if it's just slight improvement, it will favor Romney. If unemployment drops, and there is improvement it will certainly work to Obama's favor.

 

[PoliticalChic]

patisan hacks like you think other people think like you. lol..

the president outpolls romney in every aspect except the economy. and even on that, it's pretty close.

romney's likeability ratings are horrible (rightfully so) and the president has something like a 40% lead among hispanics and a double digit lead among women. he even has a 10 point lead among independents.

I am soooo...curious.

You're too smart to vote for the empty suit....
...is it going to be Romney, or skip-it?


C'mon....'fess up.....

i would never vote for anyone who thinks paul ryan has an acceptable plan. one, i might add, the catholic church thinks is "immoral".

i would never vote for anyone who thinks birth control raises issues of 'personhood'.

i would never vote for anyone who doesn't know what they think of the lily ledbetter law.

i would never vote for anyone who's thinks it would be appropriate at this juncture for us to start bomb bomb bomb-ing iran...

i would never vote for anyone who thinks shutting down planned parenthood, which provides huge numbers of women with basic care, is ok.

i think you're looking at romney's suit... that seems to be the one that's empty. there's nothing filling it except for what he thinks is going to help him with the fringers in the GOP 'base'.

and you can't tell me you think romney believes a word of what he says... the guy who started romneycare and pushed it as a federal program is suddenly morally offended by it?

heck, santorum is nuts, but at least he believes his own nonsense.

So....you're stayin' home.....?

Oprah and bon bons....sounds good.

 

[RightWingFerret]

I doubt the unemployment goes below 8.0% in the next six months. there are no indicators, and keep in mind all of those college graduates that we can add to the unemployment pool, along with the slow hiring period for the summer/fall. In the winter Florda hires in groves, then levels off in June, and In may we get the heavy northeast hiring, and by September it dies out everywhere. I just dont see any further pace of 230,000 jobs a month anytime soon.

 

[NYcarbineer]

That 'likely voter' stat is telling....

Yeah I am looking closely at that. But I am also noticing that the only two tracking polls are both showing Romney leading. That can't be just RV vs. LV because Rasmussen is LV and Gallup is RV. I am wondering whether there is something in the nature of a tracking poll itself that would create advantage.

Initially I wouldn't think so. All they are doing is averaging them out over a time frame: three days for Rasmussen and five for Gallup. That means they are slower to react but also more consistent and less subject to statistical noise.

I can't really think of anything that would create that effect. Anyone else?

This guy just threw a 3 day tantrum because I contested his assertion that RV's skew the numbers for the Democrats,

too funny.

 

[NYcarbineer]

message board Nate Silver wannbee

With phantom spreadsheets.

 

[Dr.House]

That 'likely voter' stat is telling....

Yeah I am looking closely at that. But I am also noticing that the only two tracking polls are both showing Romney leading. That can't be just RV vs. LV because Rasmussen is LV and Gallup is RV. I am wondering whether there is something in the nature of a tracking poll itself that would create advantage.

Initially I wouldn't think so. All they are doing is averaging them out over a time frame: three days for Rasmussen and five for Gallup. That means they are slower to react but also more consistent and less subject to statistical noise.

I can't really think of anything that would create that effect. Anyone else?

This guy just threw a 3 day tantrum because I contested his assertion that RV's skew the numbers for the Democrats,

too funny.

You failed epically in that thread...

Just sayin'...

 

[jillian]

I didn't say I wanted to do a scientific poll

My comment was based on my prefference for discussing poll and poll analysis with folks who don't just spew profanity and personal insults at people who disagree with them.

i thought the concept of this thread was a good one. *shrug*

You're right - I'm wrong. I'm holding a grudge against the way I saw this poster respond to someone else in a different thread.

And I'm doing exactly what he did.

I apologize to all - especially OP - for doing what I dislike most. I'll ban myself for a bit and try to come back a better poster.

that was cool.

 

[jillian]

I am soooo...curious.

You're too smart to vote for the empty suit....
...is it going to be Romney, or skip-it?


C'mon....'fess up.....

i would never vote for anyone who thinks paul ryan has an acceptable plan. one, i might add, the catholic church thinks is "immoral".

i would never vote for anyone who thinks birth control raises issues of 'personhood'.

i would never vote for anyone who doesn't know what they think of the lily ledbetter law.

i would never vote for anyone who's thinks it would be appropriate at this juncture for us to start bomb bomb bomb-ing iran...

i would never vote for anyone who thinks shutting down planned parenthood, which provides huge numbers of women with basic care, is ok.

i think you're looking at romney's suit... that seems to be the one that's empty. there's nothing filling it except for what he thinks is going to help him with the fringers in the GOP 'base'.

and you can't tell me you think romney believes a word of what he says... the guy who started romneycare and pushed it as a federal program is suddenly morally offended by it?

heck, santorum is nuts, but at least he believes his own nonsense.

So....you're stayin' home.....?

Oprah and bon bons....sounds good.

nope...

some of us work

 

[BluePhantom]

So here's what I am thinking here as I break down the demographics. I think I am going to do this two ways and publish data as such. The first will be data that has specific numbers attached to it that can be looked up by everyone for verification. I will post that data as "Verifiable" and cross reference them with averages for only the polls they are related to.

The problem is that a lot of polls (especially those pesky media polls) don't provide specific breakdowns of their demographics, so I will create another category that estimates those according to the demographics of registered voters in the United States. If it's a random sample of registered voters they should (at least over the long term) reach a sample that represents the demographics of registered voters in the United States as a whole. So for example if I can't figure out the exact number of men and women in a given poll sample I will use a set of 53% women to 47% men as that is the demographic breakdown of registered voters in the United States according to gender. I will post these figures as "Estimated". This would be somewhat parallel to a weighted LV model vs. an unweighted RV model for the purposes if analogical comparison. Over the course of several polls we should get things into the ballpark because randomness would suggest that we will reach a sample that is representative of the demographics of registered voters in the United States as a whole.

I think this will work just fine because the specific numbers are really not that important right now. What is more important is the trend and if we keep the method consistent we should be able to identify those trends over the long haul even if the accuracy is off by a few points within a specific narrow category (which is essentially margin of error anyhow).

I will use [URL="http://quickfacts.census.gov/qfd/states/00000.html"]this page from the Census Bureau[/URL] as the basis for the demographic breakdown for Estimated figures which is awesome because it breaks it down by state as well so we can use the same baseline data figures for the United States as a whole as well as swing state analysis.

Are there any questions or concerns? Are we in agreement?

 

[BluePhantom]

Let me follow up on post #34. I am going to track the following demographics: race, gender, political ideology/partisan identification, and age. On my personal spreadsheet I will track a lot more and from a lot more angles, but 1) I don't want to flood the thread with things that are only of interest to me, and 2) this is a LOT of work so the more I post the more time it's going to take and hence the longer the delay between the publishing of results on RCP and me posting the breakdown.

For race I will evaluate: white, Hispanic, black, and other

For gender I will evaluate both but will only post for women since they are the critical part of that equation.

For ideology I will evaluate Dem / Rep/ Ind but will only post for Independents since they are the critical group to watch.

For age I will track 18-29, 30-45, 46-65, and 65+ but will only post for the two middle age groups as they are the most critical.

I will only post other information per request or if it happens to suddenly become significant. Now once I get the spreadsheet fine tuned for this election it will be difficult to add other demographics, change the methodology, etc because the entire thing will be cross referenced among dozens of pages. I have had the main pages written since 2004 but every election has specific things to watch and measure and so it has to be customized for the current state of thing. I am very good at writing a spreadsheet to maximize flexibility and accommodate changing circumstances....for example I already have it set to reevaluate all data when agencies switch from RV to LV in September. It will recalculate trends according to its new matrix, extrapolate on previous data, and adjust for standard deviations in methodology. All I have to do is change two cells. The point is that we want to get this hammered down as tight as possible right from the start to avoid introducing an element later in the game that will be difficult to work into the spreadsheet.

Questions
1) Is there any other demographic that someone wants me to track?

2) Are the demographics I said I would post enough? Do you want me to post men as well as women for example? Dem and Rep ideology in addition to just Ind?

3) Is there any other information at all that people want me to track? Now is the time to speak up. I will work it into the spreadsheet if I can, but the longer we wait to add a new element to look at the harder it will be to incorporate and the less reliable the long term trend results will be.

 

[BluePhantom]

Demographics Breakdown for Week Ending 4/17/12

Ok so the data again is scarce because significant and applicable polls have just begun. There were only four polls that I could pull demographic data from and none of them applied across every category except for PPP.... congratulations to them for being thorough although they were not verifiable according to the criteria I listed before. The only poll that was totally verifiable was Quinnipiac who included not only crosstabs but some nice demographic breakdowns. I have calculated verifiable MOE of 1.9%, estimated MOE as 4.5%, and overall MOE as 3.29% across all categories.

For Estimated data there was some variation in regard to individual polls against the demographic data provided by the United States Census Bureau, but I am thrilled to report that when averaged out (when there were enough polls to average them out) the result was right on the nose with Census Bureau statistics. For this I am quite tickled as it suggests that the method I suggested earlier has hit it dead in the heart.

Here's the demographic breakdown (applicable polls in parentheses).


Gender
Verifiable (poll 1)
Men: Obama 43%, Romney 46%; Romney +3
Women: Obama 49%, Romney 39%; Obama +10

Overall Average of Verifiable Polls related to Gender: Obama 46, Romney 42; Obama +4

Estimated (polls 2,3,4)
Men: Obama 45%, Romney 49%; Romney +4
Women: Obama 52%, Romney 41%; Obama +11

Overall Average of Estimated Polls related to Gender: Obama 49, Romney 45; Obama +4

Overall (polls 1,2,3,4)
Men: Obama 44%, Romney 47%; Romney +3
Women: Obama 51%, Romney 40%; Obama +11

Overall Average of Polls Referenced in this Section: Obama 48.25, Romney 44.25; Obama +4


Race
Verifiable (poll 1)
White: Obama 36%, Romney 52%; Romney +12
Black: Obama 94%, Romney 3%; Obama +91
Hispanic: Obama 64%, Romney 24%; Obama +40
Other: Obama 63%, Romney 30%; Obama +33

Overall Average of Verifiable Polls Related to Race: Obama 46, Romney 42; Obama +4

Estimated (poll 2)
White: Obama 40%, Romney 54%; Romney +14
Black: Obama 89%, Romney 11%; Obama +78
Hispanic: Obama 68%, Romney 30%; Obama +38
Other: Obama 63%, Romney 30%; Obama +33

Overall Average of Estimated Polls related to Race: Obama 49, Romney 46; Obama +3


Overall (poll 1,2)
White: Obama 37%, Romney 52%; Romney +15
Black: Obama 92%, Romney 6%; Obama +86
Hispanic: Obama 65%, Romney 26%; Obama +39
Other: Obama 63%, Romney 30%; Obama +33

Overall Average of Polls Referenced in this Section: Obama 47.5, Romney 44; Obama +3.5



Age
Verifiable (none)

Estimated (poll 2)
18-29: Obama 56%, Romney 36%; Obama +20
30-45: Obama 47%, Romney 45%; Obama +2
46-65: Obama 51%, Romney 46%; Obama +5
66+: Obama 43%, Romney 53%; Romney +10

Overall Average of Estimated Polls related to Age: Obama 49, Romney 46; Obama +3


Overall (same as Estimated)



Ideology
Verifiable (poll 1)
Republican: Obama 8%, Romney 83%; Romney +75
Democrat: Obama 88%, Romney 6%; Obama +82
Independent: Obama 39%, Romney 46%; Romney +7


Overall Average of Verifiable Polls related to Ideology: Obama 46, Romney 42; Obama +4

Estimated (poll 2)
Republican: Obama 10%, Romney 82%; Romney +72
Democrat: Obama 84%, Romney 14%; Obama +70
Independent: Obama 52%, Romney 43%; Obama +9


Overall Average of Estimated Polls related to Ideology: Obama 49 Romney 46; Obama +3


Overall (poll 1,2)
Republican: Obama 9%, Romney 83%; Romney +74
Democrat: Obama 87%, Romney 8%; Obama +79
Independent: Obama 42%, Romney 45%; Romney +3


Overall Average of Polls Referenced in this Section: Obama 47.5, Romney 44; Obama +3.5






Included Polls (all polls RV)
1. Quinnipiac

2. PPP[D]

3. CBS/NYT

4. CNN/OR

 

[BluePhantom]

I just want to make one last point about methodology before i go to bed...'cause i am pretty exhausted. The reason why we must have a Verifiable category and a Estimated category is to properly calculate the data. We can't simply take a poll with a sample of 2,500 and average it equally against a poll with a sample of 800. The poll with the lower sample will be overemphasized and count for over three times the weight of the larger poll. This is totally bass-ackwards. So what I have to do is expand the raw data to break every category down into hard numbers against a percentage of the total sample across every single poll, sum up every subcategory, and average everything out against the mass totals.

This is easy enough to when it's just total number who supports Obama vs. Romney but the narrower we break it down the harder it is to get the exact numbers unless the polling agency provides them...which they have a bad habit of not doing. So we have two choices....prove it, or take our best guess based on reasonable supporting data.

The methodology I have proposed does both, with the Verifiable category we only focus on ratios that are known. With the Estimated category we are working with percentages against which we don't have enough information to produce raw numbers. So we use census statistics to fill in the gaps and trust that over the long haul any "noise" will filter out. In the first set of calculations it seemed to work perfectly as I said, but it's still an estimate and as such the margin of error has to increase to compensate.

The narrower we go in analysis the more we will be forced to estimate, and hence the larger the margin of error until the entire exercise becomes statistically irrelevant. It would eventually be like flipping a coin and saying "I call heads but I give myself a 50% margin of error". Well shit, in that case you win every time.

So while I can add other major demographics to track (such as education, or income) I am concerned that if I get any deeper into a single category (i.e. "married women" or "single women" vs. just "women" as a category for analysis) the margin of error will wipe out any benefit to the exercise.

ok....I am going to bed.

 

[Interpol]

As usual, we're looking at an election where both candidates walk into it with about 46-47% and then it becomes a scrap over the remaining 8-10%.

It feels like to me that the polls tend to ebb and flow all through the late spring and into the summer and that they even remain fairly tight after each party's convention because we live in an electoral pool that is sharply divided.

I wonder if the OP has anything relating to pre-convention or pre-summer polls that give us anything of an indication or not. I'm trying to remember.

I wasn't old enough to vote in the '92 election, but I remember things were all over the place.

I believe we're looking at a scenario much like the '04 election where one side is trying to make the case that the average person should stick with what they've got especially with two wars going on, while the other side tries to make the case that they should get to take over because they can do it better.

In that election it was all about foreign policy, while in this one it's all about the economy, except the administration is going to make the case why they need to continue on with overseeing this fragile recovery or the other guys will bring us more of what we got with George W. Bush, while the Republicans will make the case that this administration dropped the ball on the economy when the shoved Obamacare down our throats and that this new brand of Republican party won't drop the ball on the economy and will get us turned around again and even faster.

I think the glaring statistic is women and that we're going to see Mitt Romney over the next few weeks try to shape his campaign into centrist-friendly themes that appeal to them when it comes to the economy and health care, which are the two big ticket items among women.

 

[jillian]

Demographics Breakdown for Week Ending 4/17/12

Ok so the data again is scarce because significant and applicable polls have just begun. There were only four polls that I could pull demographic data from and none of them applied across every category except for PPP.... congratulations to them for being thorough although they were not verifiable according to the criteria I listed before. The only poll that was totally verifiable was Quinnipiac who included not only crosstabs but some nice demographic breakdowns. I have calculated verifiable MOE of 1.9%, estimated MOE as 4.5%, and overall MOE as 3.29% across all categories.

For Estimated data there was some variation in regard to individual polls against the demographic data provided by the United States Census Bureau, but I am thrilled to report that when averaged out (when there were enough polls to average them out) the result was right on the nose with Census Bureau statistics. For this I am quite tickled as it suggests that the method I suggested earlier has hit it dead in the heart.

Here's the demographic breakdown (applicable polls in parentheses).


Gender
Verifiable (poll 1)
Men: Obama 43%, Romney 46%; Romney +3
Women: Obama 49%, Romney 39%; Obama +10

Overall Average of Verifiable Polls related to Gender: Obama 46, Romney 42; Obama +4

Estimated (polls 2,3,4)
Men: Obama 45%, Romney 49%; Romney +4
Women: Obama 52%, Romney 41%; Obama +11

Overall Average of Estimated Polls related to Gender: Obama 49, Romney 45; Obama +4

Overall (polls 1,2,3,4)
Men: Obama 44%, Romney 47%; Romney +3
Women: Obama 51%, Romney 40%; Obama +11

Overall Average of Polls Referenced in this Section: Obama 48.25, Romney 44.25; Obama +4


Race
Verifiable (poll 1)
White: Obama 36%, Romney 52%; Romney +12
Black: Obama 94%, Romney 3%; Obama +91
Hispanic: Obama 64%, Romney 24%; Obama +40
Other: Obama 63%, Romney 30%; Obama +33

Overall Average of Verifiable Polls Related to Race: Obama 46, Romney 42; Obama +4

Estimated (poll 2)
White: Obama 40%, Romney 54%; Romney +14
Black: Obama 89%, Romney 11%; Obama +78
Hispanic: Obama 68%, Romney 30%; Obama +38
Other: Obama 63%, Romney 30%; Obama +33

Overall Average of Estimated Polls related to Race: Obama 49, Romney 46; Obama +3


Overall (poll 1,2)
White: Obama 37%, Romney 52%; Romney +15
Black: Obama 92%, Romney 6%; Obama +86
Hispanic: Obama 65%, Romney 26%; Obama +39
Other: Obama 63%, Romney 30%; Obama +33

Overall Average of Polls Referenced in this Section: Obama 47.5, Romney 44; Obama +3.5



Age
Verifiable (none)

Estimated (poll 2)
18-29: Obama 56%, Romney 36%; Obama +20
30-45: Obama 47%, Romney 45%; Obama +2
46-65: Obama 51%, Romney 46%; Obama +5
66+: Obama 43%, Romney 53%; Romney +10

Overall Average of Estimated Polls related to Age: Obama 49, Romney 46; Obama +3


Overall (same as Estimated)



Ideology
Verifiable (poll 1)
Republican: Obama 8%, Romney 83%; Romney +75
Democrat: Obama 88%, Romney 6%; Obama +82
Independent: Obama 39%, Romney 46%; Romney +7


Overall Average of Verifiable Polls related to Ideology: Obama 46, Romney 42; Obama +4

Estimated (poll 2)
Republican: Obama 10%, Romney 82%; Romney +72
Democrat: Obama 84%, Romney 14%; Obama +70
Independent: Obama 52%, Romney 43%; Obama +9


Overall Average of Estimated Polls related to Ideology: Obama 49 Romney 46; Obama +3


Overall (poll 1,2)
Republican: Obama 9%, Romney 83%; Romney +74
Democrat: Obama 84%, Romney 14%; Obama +70
Independent: Obama 52%, Romney 43%; Obama +9


Overall Average of Polls Referenced in this Section: Obama 47.5, Romney 44; Obama +3.5






Included Polls (all polls RV)
1. Quinnipiac

2. PPP[D]

3. CBS/NYT

4. CNN/OR

You know what I find interesting about those demographics, if the president has those massive demographic advantages among women, Hispanics and blacks (as well as among younger voters and independents), I'd think it would make him very difficult, if not impossible to beat. Yet the polls seem tighter than that. It's counter-intuitive.

 

[BluePhantom]

As usual, we're looking at an election where both candidates walk into it with about 46-47% and then it becomes a scrap over the remaining 8-10%.

It feels like to me that the polls tend to ebb and flow all through the late spring and into the summer and that they even remain fairly tight after each party's convention because we live in an electoral pool that is sharply divided.

I wonder if the OP has anything relating to pre-convention or pre-summer polls that give us anything of an indication or not. I'm trying to remember.

I wasn't old enough to vote in the '92 election, but I remember things were all over the place.

I believe we're looking at a scenario much like the '04 election where one side is trying to make the case that the average person should stick with what they've got especially with two wars going on, while the other side tries to make the case that they should get to take over because they can do it better.

In that election it was all about foreign policy, while in this one it's all about the economy, except the administration is going to make the case why they need to continue on with overseeing this fragile recovery or the other guys will bring us more of what we got with George W. Bush, while the Republicans will make the case that this administration dropped the ball on the economy when the shoved Obamacare down our throats and that this new brand of Republican party won't drop the ball on the economy and will get us turned around again and even faster.

I think the glaring statistic is women and that we're going to see Mitt Romney over the next few weeks try to shape his campaign into centrist-friendly themes that appeal to them when it comes to the economy and health care, which are the two big ticket items among women.

I think you have it right on the nose. And just to make a quick reference to Jillian's point the reason why the women vote is so much more important than the Hispanic or black vote is because of sheer numbers.

According to the page on registered voter demographics I posted earlier we have the following breakdowns.

Men: 47%
Women: 53%

White: 76%
Black: 12%
Hispanic: 8%
Other: 4%

So the fact that Obama is winning 89% of the black vote isn't nearly as important because it's just not a really huge voting block like women are. Women represent the majority of that demographic so holding a big lead there is just critical.