Welcome to DU!
The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards.
Join the community:
Create a free account
Support DU (and get rid of ads!):
Become a Star Member
Latest Breaking News
General Discussion
The DU Lounge
All Forums
Issue Forums
Culture Forums
Alliance Forums
Region Forums
Support Forums
Help & Search
This Is A Must Read For Every DUER-Fascinating Stuff
tbennett76
Why Gallup is Wrong and an Election Projection-Hint: Cell Phones.
First of all as we know, Gallup for some reason thinks a highly motivated electorate is going to break a 50 year trend and be four points more white than past electorates. More disturbingly, they are aligning almost directly behind Rasmussen and projecting a more Republican than Democratic turnout. If this were true, RIGHT NOW, the state polls would pick that up, even in hotly contested battlegrounds. What we have is a 25 pt lead in New York, a 14 pt spread in NJ, a close race in FL. a small lead in NH, VA, a sizable lead in CO, a blowout pending in NV, and a 5-7 pt lead in IA. It is also believed the POTUS has a 5 pt lead in OH. That state numbers are pretty much consistent with a small but solid lead nationally for POTUS. Not some state numbers, pretty much all of them.
I also want to blow out of the water a long held myth, about state polls lagging nationals. I believe the exact opposite. If you look at debates 1 and 3, the state polls immediately reflected the shift in preferences. The nationals, even the short trackers, did not. My theory is that a state poll is more likely to be properly stratified in the first place, in terms of demographics, and cell usage, than a national, and thus more likely to pick up enthusiasm accurately. As national polls have varying methodologies, it is hard to pinpoint the error, but one place to look is cell phone stratification. More cell only users are in the south, but it is understood that a flat weight is given to all cell users nationwide. According to the CDC,
adults living in the Midwest
(33.5%), South (33.6%), and West
(30.3%) were more likely than
adults living in the Northeast
(18.8%) to be living in households
with only wireless telephones.,
Did you catch that? A southern resident is nearly 50 percent more likely to have a cell only household. Cell only households typically run 2 to 1 for POTUS, sometimes 60-40, but in a more heavily African American populated area, probably closer to the former. Let's say that the average poll gets 100 African Americans, like PPP, honorable pollster, but hopelessly outgunned by response challenges. Now a hundred people has something like a margin of error of +-11%. But that is all they can get, so they go with whatever goofy number they find. Now they weigh the monkey testicles out of that skewed result, because they just can't get enough on the phone. So they actually enhance the inaccuracy. But say a pollster did not have the response challenges, and could reach cell phone onlies, but like most national pollsters, did not weight for cell use by region. "We called 25 percent cell only, and averaged it out nationwide."
Wrong. If you do that in the south, you miss 8.5% of potential respondents. Lets say that the 8.5 percent missed was weighted equally 50-50 Romney to Obama. That is 4.25-4.25. Now let us plug in the 2-1 margin, you get 5.7-2.8. That is a three point difference right there. NOW consider, that the 3 pt swing, in the south, a region accountable for roughly one third of the electorate historically, is single handedly, through not understanding cell phones, responsible for a full pt. Now consider the similar skew for the west, and for hard to reach Latinos, and so forth.
In the past the electorate was much more monolithically white. If you could get a good chunk of white response, when they represented more like 85 percent of the vote, the results would be overwhelming to whatever you were missing with non whites. But the electorate has shifted. Now we know the electorate is just not going to be 78 percent white. Both sides agree on something between 72-74. Probably, closer to 72. Let us just call it 73 percent. POTUS figures to receive, and agreed upon by ABC, 82 percent of the non white vote, maybe more. That accounts for 22.96 percent of the two party vote. At 40 percent white, historically on the low side, he would pick up another 29.2 percent of the vote. This brings him to 52. Let us reskew those numbers. Say the electorate is 78 percent white. With these same breakdowns that knocks him to 49. Very similar to Kerry. We go from comfortable win to hoping for an EV PV split. Now to show you how hopeless PPP is, they are calling for an electorate that is 70 percent white and show a tie. If the electorate is 70 percent white, POTUS wins by 8-10. But how do they get there? Simple. Robocalling. The actual Latinos they get via landline speak English. Yet history shows the polls in CO and NV in 2010 were 5-8 pts off. We also know that several points are lost when a poll does not call in Spanish as well. It would be fair to say also that the AA's they get on the line, having that expensive landline, probably skew just slightly more right than cell onlies. So we get an unrepresentative sample extrapolated to mean something, and we get weird polling. The actual assumed stats, based on numerous sources, say 75-25 Latino, and 98-2 African American for Obama.
If you add that all up, along with roughly 2-1 other, you get about, 83 percent, in line with ABC. Right now PPP shows 73. Not buying it. As for Mr. Gallup, they are using a decades old model, aligned for an era when people rarely moved, and were easily reachable. In 2010 they missed it by 9. If you apply that same assumption in this case, that means Obama leads by about 4. My understanding is that both sides agree this is roughly a 3 pt race, 50-47, in Obama's favor. I believe the electorate shakes out like this, 73 percent white, 40 percent for Obama, 83 percent for Obama non white. Bottom line is he is projected to win 52-47, with 347 electoral votes. And the biggest shocks I call for? 48 in Texas, tossup in AZ, 48 in SC, GA, and and 54 in VA. Any questions just ask, thanks
-Todd
P.S. two more thoughts, online polls are the most accurate, as internet access in more demographically representative than phone usage, and the winner, of the most accurate poll, a prediction, and the new model of the future, will be, RAND.
http://www.democraticunderground.com/1251179461
Why Gallup is Wrong and an Election Projection-Hint: Cell Phones.
First of all as we know, Gallup for some reason thinks a highly motivated electorate is going to break a 50 year trend and be four points more white than past electorates. More disturbingly, they are aligning almost directly behind Rasmussen and projecting a more Republican than Democratic turnout. If this were true, RIGHT NOW, the state polls would pick that up, even in hotly contested battlegrounds. What we have is a 25 pt lead in New York, a 14 pt spread in NJ, a close race in FL. a small lead in NH, VA, a sizable lead in CO, a blowout pending in NV, and a 5-7 pt lead in IA. It is also believed the POTUS has a 5 pt lead in OH. That state numbers are pretty much consistent with a small but solid lead nationally for POTUS. Not some state numbers, pretty much all of them.
I also want to blow out of the water a long held myth, about state polls lagging nationals. I believe the exact opposite. If you look at debates 1 and 3, the state polls immediately reflected the shift in preferences. The nationals, even the short trackers, did not. My theory is that a state poll is more likely to be properly stratified in the first place, in terms of demographics, and cell usage, than a national, and thus more likely to pick up enthusiasm accurately. As national polls have varying methodologies, it is hard to pinpoint the error, but one place to look is cell phone stratification. More cell only users are in the south, but it is understood that a flat weight is given to all cell users nationwide. According to the CDC,
adults living in the Midwest
(33.5%), South (33.6%), and West
(30.3%) were more likely than
adults living in the Northeast
(18.8%) to be living in households
with only wireless telephones.,
Did you catch that? A southern resident is nearly 50 percent more likely to have a cell only household. Cell only households typically run 2 to 1 for POTUS, sometimes 60-40, but in a more heavily African American populated area, probably closer to the former. Let's say that the average poll gets 100 African Americans, like PPP, honorable pollster, but hopelessly outgunned by response challenges. Now a hundred people has something like a margin of error of +-11%. But that is all they can get, so they go with whatever goofy number they find. Now they weigh the monkey testicles out of that skewed result, because they just can't get enough on the phone. So they actually enhance the inaccuracy. But say a pollster did not have the response challenges, and could reach cell phone onlies, but like most national pollsters, did not weight for cell use by region. "We called 25 percent cell only, and averaged it out nationwide."
Wrong. If you do that in the south, you miss 8.5% of potential respondents. Lets say that the 8.5 percent missed was weighted equally 50-50 Romney to Obama. That is 4.25-4.25. Now let us plug in the 2-1 margin, you get 5.7-2.8. That is a three point difference right there. NOW consider, that the 3 pt swing, in the south, a region accountable for roughly one third of the electorate historically, is single handedly, through not understanding cell phones, responsible for a full pt. Now consider the similar skew for the west, and for hard to reach Latinos, and so forth.
In the past the electorate was much more monolithically white. If you could get a good chunk of white response, when they represented more like 85 percent of the vote, the results would be overwhelming to whatever you were missing with non whites. But the electorate has shifted. Now we know the electorate is just not going to be 78 percent white. Both sides agree on something between 72-74. Probably, closer to 72. Let us just call it 73 percent. POTUS figures to receive, and agreed upon by ABC, 82 percent of the non white vote, maybe more. That accounts for 22.96 percent of the two party vote. At 40 percent white, historically on the low side, he would pick up another 29.2 percent of the vote. This brings him to 52. Let us reskew those numbers. Say the electorate is 78 percent white. With these same breakdowns that knocks him to 49. Very similar to Kerry. We go from comfortable win to hoping for an EV PV split. Now to show you how hopeless PPP is, they are calling for an electorate that is 70 percent white and show a tie. If the electorate is 70 percent white, POTUS wins by 8-10. But how do they get there? Simple. Robocalling. The actual Latinos they get via landline speak English. Yet history shows the polls in CO and NV in 2010 were 5-8 pts off. We also know that several points are lost when a poll does not call in Spanish as well. It would be fair to say also that the AA's they get on the line, having that expensive landline, probably skew just slightly more right than cell onlies. So we get an unrepresentative sample extrapolated to mean something, and we get weird polling. The actual assumed stats, based on numerous sources, say 75-25 Latino, and 98-2 African American for Obama.
If you add that all up, along with roughly 2-1 other, you get about, 83 percent, in line with ABC. Right now PPP shows 73. Not buying it. As for Mr. Gallup, they are using a decades old model, aligned for an era when people rarely moved, and were easily reachable. In 2010 they missed it by 9. If you apply that same assumption in this case, that means Obama leads by about 4. My understanding is that both sides agree this is roughly a 3 pt race, 50-47, in Obama's favor. I believe the electorate shakes out like this, 73 percent white, 40 percent for Obama, 83 percent for Obama non white. Bottom line is he is projected to win 52-47, with 347 electoral votes. And the biggest shocks I call for? 48 in Texas, tossup in AZ, 48 in SC, GA, and and 54 in VA. Any questions just ask, thanks
-Todd
P.S. two more thoughts, online polls are the most accurate, as internet access in more demographically representative than phone usage, and the winner, of the most accurate poll, a prediction, and the new model of the future, will be, RAND.
http://www.democraticunderground.com/1251179461
I have one thought. What does it do to the numbers if O is around 38% of the white vote? The campaign in Chicago says that's good enough
3 replies
= new reply since forum marked as read
= new reply since forum marked as read
is the one I took in my house. We are white, and Obama is winning 100% of the vote!
