General Discussion
Related: Editorials & Other Articles, Issue Forums, Alliance Forums, Region ForumsAbout that New AP Poll
by Alan Abramowitz
Obama leads Romney by 1 point among likely voters. But wait . . . Obama leads Romney by 15 points among all adults and by 10 points among registered voters. What is going on here? The answer is, they're using a ridiculously tight likely voter screen that counts only 63% of registered voters as likely voters. That might make sense for a midterm election but the normal turnout of registered voters in a presidential election is around 80-85%.
http://www.dailykos.com/story/2012/09/19/1133835/-About-that-New-AP-Poll
The "ridiculously tight likely voter screen" is actually about 67 percent. First, here is the press release.
By NANCY BENAC and JENNIFER AGIESTA, Associated Press
WASHINGTON (AP) Americans are feeling markedly better about the countrys future and about Barack Obamas job performance, but the presidents re-election race against Republican Mitt Romney remains a neck-and-neck proposition as Election Day creeps ever closer, according to a new Associated Press-GfK poll.
Buoyed by good mojo coming out of last months national political conventions, Obamas approval rating is back above 50 percent for the first time since May, and the share of Americans who think the country is moving in the right direction is at its highest level since just after the death of Osama bin Laden in May 2011.
Romney, his campaign knocked off-stride in recent weeks, has lost his pre-convention edge on the top issue of the campaign the economy.
The poll results vividly underscore the importance that turnout will play in determining the victor in Campaign 2012: Among all adults, Obama has a commanding lead, favored by 52 percent of Americans to just 37 percent for Romney. Yet among those most likely to vote, the race is drum tight.
Obama is supported by 47 percent of likely voters and Romney by 46 percent, promising an all-out fight to the finish by the two campaigns to gin up enthusiasm among core supporters and dominate get-out-the-vote operations. Thats an area where Obama claimed a strong advantage in 2008 and Republicans reigned four years earlier.
http://ap-gfkpoll.com/uncategorized/our-latest-poll-findings-11
More: http://surveys.ap.org/data/GfK/AP-GfK%20Poll%20September%202012%20Topline_1st%20release.pdf
Interview dates: September 13 17, 2012
Number of interviews, adults: 1,512
Number of interviews, likely voters: 807
Are you currently registered to vote at your address, or not?
Yes 79
No 21
So registered to vote is roughly 1,195 respondents, or 67 percent of the total sample. That is a ridiculously tight screen.
Still, here is how the current poll compares to the previous poll...
Among all adults:
Obama 52, Romney 37 (previous poll - Obama 48, Romney 44, an 11-point swing.)
Among registered voters:
Obama 50, Romney 40 (previous poll - Obama 47, Romney 46, an 9-point swing.)
justiceischeap
(14,040 posts)to make it a race? No one on DU has suspected this at all.
DemocratSinceBirth
(99,710 posts)The Dems had +4 in 00 and +8 in 08 according to exit polls. In 04 it was 00... I think it's more likely the advantage will be somewhere between 00 and 08.
Response to ProSense (Original post)
cthulu2016 This message was self-deleted by its author.
DemocratSinceBirth
(99,710 posts)Alan I. Abramowitz is the Alben W. Barkley Professor of Political Science at Emory University in Atlanta, Georgia.
I question the finding the Dems will only comprise only one percentage more of the electorate than the Republicans and that African Americans will comprise only ten percent of the electorate when they were thirteen percent of the electorate in 2008 and eleven percent in 2004.
cthulu2016
(10,960 posts)I, of course, haven't counted all the registered voters myself so everybody is relying on statistical data.
And if I happened never be exposed to the "right" answer then I could be ignorant on the matter.
But I don't see any statistical anything anywhere on the google machine to suggest that we have seen 80-85% of registered voters actually voting since the 1960s.
As with all things, if I am wrong I am wrong, and happy to say so.
ProSense
(116,464 posts)I believe the other polls use a screen above 80 percent.
ProSense
(116,464 posts)The overall voting age (18 and older) citizen population in the United States in 2008 was 206 million compared with 197 million in 2004. Of that total, 146 million, or 71 percent, reported being registered to vote. That's slightly lower than the 72 percent who reported being registered to vote in the 2004 presidential election, but does represent an increase of approximately 4 million registered voters. The percentage of those registered to vote that actually did so was slightly higher in the 2008 election (90 percent) than in 2004 (89 percent).
http://www.census.gov/newsroom/releases/archives/voting/cb09-110.html
So the percentage of registered voters among the adult population was 71 percent, but the percent of those who voted was 90 percent, up from 89 percent in 2004.
Alan Abramowitz
http://en.wikipedia.org/wiki/Alan_Abramowitz
cthulu2016
(10,960 posts)1) it appears that no two sources agree on how many registered voters there were in 2008, and
2) without knowing the AP number of registered voters their estimate of registered voter participation cannot be assessed.
Abramowitz's critique may be right or it may be erroneous, depending on whether he and AP are using the same sets of numbers, which appears unlikely, but is possible.
The census number is not reliable because it is from the census surveyself-reported for both registration and voting. It is not based on any actual records of voter registration or votes cast. That survey said 146 million registered.
The National Association of Secretaries of State reported 184,213,797 registered voters as of late October, 2008. (Plus whatever North Dakota had, since they didn't do pre-registration and couldn't be counted.) Like the census data, the NASS count is also unreliable. It is somewhat inflated by double-counted registrations and invalid registrations.
Other counts of registration say 169 million, which is between the two. So it's all over the place.
28 million is a huge discrepancy. And depending on which count of registered voters one picks, it will return very different results for %s.
Now, in 2008 AP reported 187,332,357 registered voters. 31 million more regstered voters than the census number. So we know that AP probably has a very, very high number of registered in their thinking. But if AP thinks there are that many registered voters and has a screen resulting in 63% then it is apples and apples... 63% of a much larger number.
So the question is, Is AP is using the census number of registered voters or the state-by-state count? If AP is not using the same census figure Abramowitz uses then the whole argument is comparing apples and oranges, and falls apart.
Sayng 63% of 146 will vote versus saying 63% of 187 will vote is a ginormous difference.
(A factor I would be curious about in all these likely voter screens is how they weight 2010 behavior. Since that was a high-turnout republican year and depressed for Dems it would make a difference.)
ProSense
(116,464 posts)...they were reporting that 90 percent of the adult population is registered to vote.
cthulu2016
(10,960 posts)Though if they believe that 90% of the eligible population is registered to vote then they would also believe that not many registered voters turn out.
It looks like the George Mason election project (the first place I went to look) won't even get into the registered question, despite doing great work estimating prison populations, resident aliens, etc..
ananda
(28,860 posts).. this election.
That SO sucks!
sadbear
(4,340 posts)A lot of them will be staying home November 6.