Politico: AP Poll: Obama by one (+6 for McCain)

[Heisenberg]

AP poll: Obama by one
By: Alexander Burns
October 22, 2008 07:11 PM EST
It’s not that the tide of the race is turning. But for Republicans anxiously scanning the headlines for any sign of promising news, a few encouraging tidbits turned up Wednesday — signs that recent rumors of the McCain campaign’s demise have been premature.

Most heartening for conservatives will likely be a new Associated Press poll, conducted by GfK Roper Public Affairs & Media, which showed the presidential race as a one-point contest. According to this survey, Obama leads McCain among likely voters nationwide by just 44 percent to 43 percent.

That showing represents a six-point improvement for McCain since AP-GfK’s last poll, which had Obama ahead of McCain by seven points, 48 percent to 41 percent.

Perhaps even more encouraging to Republicans are two new surveys conducted by the highly respected Mason-Dixon polling firm, in the key states of Florida and Virginia, which show McCain well within striking distance in both battlegrounds.

In Florida, Mason-Dixon shows Obama’s numbers slipping ever so slightly while McCain holds steady — and now leads the Democrat, 46 percent to 45 percent. Last month Obama held a two-point lead over McCain in Florida, 48 percent to 46 percent.

In Virginia, a state that political analysts including former White House deputy chief of staff Karl Rove have tagged as a probable win for Obama, the Democrat is on top by two percentage points, 47 percent to 45 percent.

This is an improvement for Obama since Mason-Dixon’s last round of polling in the Old Dominion, which showed McCain ahead of Obama by three points. Other polls have shown similar movement toward Obama, with the RealClearPolitics polling average handing the Illinois senator a nearly seven-point lead in the state.

Mason-Dixon is among the most experienced polling firms when it comes to surveying Virginia, so even a slight Obama lead in the newest numbers offers hope to Republicans that the state’s 13 electoral votes may still be within reach.

In Pennsylvania, where McCain seems increasingly to be pinning his hopes and where polls have shown Obama with a comfortable lead in the high single digits or low double digits, Republicans can take heart in Democratic Gov. Ed Rendell’s conservative assessment of the state of the race.

“I’m still a little nervous, so I have asked Obama to come back,” Rendell told CNN Monday evening. “We understand he’s got demands from 20 different states, but we’d like to see him here.”
Rendell’s remarks came after the governor contacted the Obama campaign to ask that the Democratic presidential nominee and his former rival, Sen. Hillary Rodham Clinton and former President Bill Clinton, return to the state for another round of campaigning.

Coming from Rendell, a savvy political operator whose turnout machine helped propel Hillary Clinton to victory in the Democratic primary, the comments suggest that McCain is running more competitively in Pennsylvania than recent polls indicate.

The overall outlook of the campaign is still dispiriting for the GOP, and a Fox/Opinion Dynamics poll released Wednesday — which showed Obama ahead nationally, 49 percent to 40 percent — reinforced the impression that Obama maintains a solid lead.

A new round of state-level polling commissioned by CNN and TIME magazine also showed Obama with leads in four important swing states including Virginia, where the Democrat led McCain, 54 percent to 44 percent.

But for Republicans determined to hang on to the White House, the glimmer of positive news comes at a useful moment, as some commentators are starting to write off McCain’s candidacy as a lost cause.

On “Meet the Press” Sunday, NBC political analyst Chuck Todd suggested that McCain’s campaign might be “conceding the popular vote.”

“We’re seeing a lot of tightening in places that while Obama probably won’t carry them, he’s not going to lose by large margins,” Todd said. “That means the McCain path is solely now an Electoral College path.”


http://www.politico.com/news/stories/1008/14838.html


For all you fellow Obama supporters declaring the race over . . . think again.

[Toe Jam]

Bahaha.

The media really wants to set election night up as close and interesting.

[Heisenberg]

Bahaha.

The media really wants to set election night up as close and interesting.




I guess I have to be consistent -


please TJ, don't play the media card.

[Burgold]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?

[Heisenberg]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?


I was wondering the same thing - It's weird but it seems like it may be closer than what you would think.

[brandymac27]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?


I read somewhere he was ahead by 9. I dunno.

[Mickalino]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?

It proves my point that polls are far from being "scientific"

[Burgold]

In a way, this is good news for Obama. Obama's biggest worry is voter apathy. He needs new voters to go out and vote in droves. They'll be less likely to do so if they think the election is a done deal.

Still, the polls do seem to be all over the place and I do think there will be a tightening over these last two weeks.

[BRAVEONAWARPATH]

http://www.huffingtonpost.com/2008/10/22/ap-poll-that-shows-tighte_n_137002.html


A new poll was released earlier today by the Associated Press that seemed to make imply that the race was tightening on a national scale. The Associated Press said:

The poll, which found Obama at 44 percent and McCain at 43 percent, supports what some Republicans and Democrats privately have said in recent days: that the race narrowed after the third debate as GOP-leaning voters drifted home to their party and McCain's "Joe the plumber" analogy struck a chord.
AmericaBlog looked over the methodology, however, and a found that the poll skewed far too heavily in favor of Evangelical Christians. Click here for the complete breakdown:

The problem? In 2004, evangelicals/born-again Christians made up 23% of voters. But that same group makes up 44% of likely voters in AP's poll released today. That's almost double the number - it's totally implausible.

[Predicto]

The pollsters are doing their honest best, but they are having a heck of a time pegging this race property.

Traditionally, you weigh your results based on the turnout of the last election - how many Repubs voted, how many Democrats, how many young voters, how many older voters, etc. You adjust your survey results to reflect this expectation. Historically, the "likely voter" is generally older and whiter than the average voter, so a tie in your raw polling data really means that the Republican is ahead by a few points.

The problem is, this year the turnout in the primaries was not the same as the turnout in the 04 election. Many more young people and minorities came out to vote. No one knows if they will do it again in November. Plus, many more people are identifying themselves as Democrats than did in 04, and the turnout in the Democratic primaries was much higher.

The pollsters weigh all those things differently - ergo, the different poll results.

[Heisenberg]

The pollsters are doing their honest best, but they are having a heck of a time pegging this race property.

Traditionally, you weigh your results based on the turnout of the last election - how many Repubs voted, how many Democrats, how many young voters, how many older voters, etc. You adjust your survey results to reflect this expectation. Historically, the "likely voter" is generally older and whiter than the average voter, so a tie in your raw polling data really means that the Republican is ahead by a few points.

The problem is, this year the turnout in the primaries was not the same as the turnout in the 04 election. Many more young people and minorities came out to vote. No one knows if they will do it again in November. Plus, many more people are identifying themselves as Democrats than did in 04, and the turnout in the Democratic primaries was much higher.

The pollsters weigh all those things differently - ergo, the different poll results.


Interesting post - probably spot on.


I guess we'll find out in less than 2 weeks.

[PeterMP]

The pollsters are doing their honest best, but they are having a heck of a time pegging this race property.

Maybe this is enough to put me in the tin foil hat group, but this screams of me as they've changed their definition of a "likely voter" midstream or something else in their methodology. I guess, they might have some reason for doing it, but it is also a good way to make headlines. If you look at total respondants, you see a drop from 12 to 10 (which is in-line with other polls), but likely voters they see a drop of 7 to 1.

Unless they explain the disparity between likely and total between the two polls, I don't buy it. It isn't normal and none of the other big polls show this kind of movement.

http://www.ap-gfkpoll.com/pdf/AP-GfK_Poll_2_Topline_10108_FINAL.pdf

9/27/08-9/30/08

Likely Voters; Total Respondents

TOTAL Barack Obama and Joe Biden, the Democrats; 48; 49

TOTAL John McCain and Sarah Palin, the Republicans; 41; 37

http://www.ap-gfkpoll.com/pdf/AP-GfK_Poll_3_Topline_FINAL.pdf
10/16/08-10/20/08

Total Respondents; Likely Voters

TOTAL Barack Obama and Joe Biden, the Democrats; 47; 44

TOTAL John McCain and Sarah Palin, the Republicans; 37; 43

[Hubbs]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?

While some very good points have been made by other posters in this thread, you'd be surprised what you can do with just the smallest tweaks in methodology. Flip which candidate's name you say first, for example, and your results will change by a point or two. Same thing happens if you say both the first and last names versus just the first names. And the phrasing of the question can do all sorts of crazy things with the numbers.

That's all if you push for answers or include "leaning" voters and such. If you only include those who are very sure about who they're voting for, those changes won't show up.

[TuckahoeSkin]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?


Looking at the internals of many of these polls show an over sampling of groups that tend to favor Obama. For instance, in the Suffolk Poll of Va, the percentage of D's polled was roughly 10% higher than the previous two elections. The number of R's was roughly 10% less than the previous polls. In other polls, blacks are oversampled compared to whites. In others, more women are sampled than are represented in the electorate.

You can make statistics do cartwheels when you try hard enough.

[PeterMP]

While some very good points have been made by other posters in this thread, you'd be surprised what you can do with just the smallest tweaks in methodology. Flip which candidate's name you say first, for example, and your results will change by a point or two. Same thing happens if you say both the first and last names versus just the first names. And the phrasing of the question can do all sorts of crazy things with the numbers.

That's all if you push for answers or include "leaning" voters and such. If you only include those who are very sure about who they're voting for, those changes won't show up.

Looking across polls (as Burgold is doing) like comparing the AP poll to the NBC poll, especially when you are talking about likely voters, is dangerous for those reasons.

What is normally safe is to look for general trends within a poll. Of course, if they change their methodology midstream, then that throws things off. Most don't report anything about their methodology or changes in it.

[brianm23]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?


Their sampling is probably not random and the number of samples collected are probably not enough. I would be more inclined to go with a CNN or NBC/CBS/ABC type of poll before I would "politico".

[Heisenberg]

I guess I'll stick with the Gallup and RCP polls, they seem to be pretty reliable from what I have read.

[brianm23]

http://www.cnn.com/ELECTION/2008/map/polling/

[Predicto]

Looking at the internals of many of these polls show an over sampling of groups that tend to favor Obama. For instance, in the Suffolk Poll of Va, the percentage of D's polled was roughly 10% higher than the previous two elections. The number of R's was roughly 10% less than the previous polls. In other polls, blacks are oversampled compared to whites. In others, more women are sampled than are represented in the electorate.

You can make statistics do cartwheels when you try hard enough.

Absolutely.

Of course, the Suffolk Poll would respond that 10 percent more people currently identify themselves as Democrats than did in the last two elections. Democratic registration is way, way up, as reflected by the respective primary turnouts.

So what do we do with that information? We can ignore it and stick with 2004 figures (Zogby does that). Or we can guess, based on what we THINK will happen this year (Gallup does that). Or we can do something else.

It's hard. It's not a conspiracy to get to a particular result to favor anyone. Every pollster wants to be the one who nails it exactly right.

[Predicto]

I guess I'll stick with the Gallup and RCP polls, they seem to be pretty reliable from what I have read.

Fivethirtyeight.com tinks that Rasmusson is the most reliable historically.

It is also one of the polls that has had the race closest so far.

[mjah]

It proves my point that polls are far from being "scientific"
Perhaps you meant to say, "far from being 'accurate.'"

If one poll shows a 1 point difference and the other shows 11 points, at least one of 'em is just wrong.

They all depend on some critical factors, such as getting a sufficiently random sample of people and accurately weighing the results to reflect the demographics of the country -- which, as some point out, is a bit unsettling given that a poll is supposed to give a reasonable representation of demographics in the first place. I haven't seen a thorough writeup on what sample size would be sufficient to capture demographics as well as overall population, but I assume it would be larger and therefore more expensive than roughly 1,000.

All of which are good reasons to average the results we get from polls, each of which is just a bit different from the others. You could argue that the polls are a small aggragated decision market of sorts: none knows everything, but they all know something. I wouldn't be surprised to see the average be pretty damned close to the result, saved for any Bradley effect and (countering that) underestimation of new voter demographics during this election.

Which brings me to another question: Why do the sites average the polls instead of taking a median? I haven't run the numbers, but I could see it occasionally making a notable difference in the stats they report. Particularly when one poll says +11, and another one that says +1 hasn't come out yet.

[Mickalino]

Puerto Rico's voice has not been heard.





















Nor will it be.

[81artmonk]

I guess this election isn't over before it began as the media would have us believe.

[Hubbs]

I guess this election isn't over before it began as the huge cackling liberal media conspiracy would have us believe.

Fixed that for you.

[PeterMP]

Fixed that for you.

Fox News has it as +9 for Obama up from +7. Zogby, who at least was a regular on Hannity's show, has it as +10 for Obama up from +5.

The liberal AP has it as +1 for Obama, but its the liberal media that's calling it over?

[twa]

The polls are so weird this cycle. NBC just reported their most recent poll shows Obama 11 points ahead. How can there be such descrepencies? Are they asking very different questions or looking at different groups?

Weighing is a major factor as well as if they /and how they determine likely voters.

nothing is as simple or complex as asking people how they will vote.;)

[mistertim]

I guess this election isn't over before it began as the media would have us believe.

I think it is actually more in the media's interest to make it sound close, otherwise people tune out and they lose money. Either way, one poll out of dozens over the last few days doesn't necessarily mean that much on it's own. The election is by no means over but when you're a Republican and you're losing in Virginia, North Carolina, have completely given up in Michigan, are behind in many other traditional swing states and only have 2 weeks left you better get in gear pretty fast.

[twa]

Fivethirtyeight.com tinks that Rasmusson is the most reliable historically.

It is also one of the polls that has had the race closest so far.

Thats because they are conservative in determining likely voters and the weighing.
It is the only national poll I pay much attention to at all...and even that is a educated guess;)

[PeterMP]

Thats because they are conservative in determining likely voters and the weighing.
It is the only national poll I pay much attention to at all...and even that is a educated guess;)

Do you work for Rasmussen?

How do you know anything about what they do w/ respect to likely voters?

[twa]

Do you work for Rasmussen?

How do you know anything about what they do w/ respect to likely voters?

I read...the info for most is out there if you look and I find the methodology interesting

and no,I do not do polling .....I'm a bodyman

[PeterMP]

I read...the info for most is out there if you look and I find the methodology interesting

and no,I do not do polling .....I'm a bodyman

Rasmussen essentially releases no information on how they go about deciding likely voters. They do release information on what their party weightings are, but even there it isn't clearly obvious to me how they apply them (I have some guesses, but you don't get enough data from them to actually prove it), and I think I've read every page at their site.

I don't know what you've read or where you've read it, but I'd love to see it.

[twa]

I'll see if I saved anything good for ya,Scott ain't that shy about it.

[Mickalino]

I guess there's no point in me voting, now that the polls have already decided the outcome

[Hubbs]

Fox News has it as +9 for Obama up from +7. Zogby, who at least was a regular on Hannity's show, has it as +10 for Obama up from +5.

The liberal AP has it as +1 for Obama, but its the liberal media that's calling it over?

Have you checked your sarcasm meter lately? ;)

[G.A.C.O.L.B.]

Outlier. Nothing more, nothing less.


But when in doubt go to fivethirtyeight:

http://www.fivethirtyeight.com/2008/10/todays-polls-1022.html


Today's Polls, 10/22

While there are plenty of individual results for Matt Drudge to get excited about (or for that matter the Huffington Post), the fact is that the overall trend in this election is roughly flat, and has been for about a week or so. That is bad news for the candidate trailing in the race, which in this case is John McCain.

http://farm4.static.flickr.com/3213/2964743915_8c84158441_o.png

Five of the eight tracking polls moved toward Barack Obama today; the other three moved toward McCain. It's possible that there was a very slight shift in momentum toward John McCain last week, and that it's been counteracted by a similar shift toward Barack Obama after the Colin Powell endorsement. It's also possible that we've simply been reading too much into what amounts to statistical noise. Either way, our Super Tracker thinks that Obama is now no more than half a point off his peak numbers, which were realized earlier this month. John McCain needs to begin closing these numbers, and soon.

There are lots of numbers to pick through that the state level today, but few of them are wholly surprising. CNN gives Barack Obama a sizable lead in Virginia, and Mason-Dixon -- one of the few pollsters to call the state for John McCain before -- now gives Obama a smaller one. On the other hand, Mason-Dixon is the latest pollster to have Florida tightening. We suddenly have a flood of West Virginia polling, and most of it suggests that the state isn't likely to flip. For that matter, however, one of McCain's reach states -- Maine -- does not appear to be tightening.

McCain does get some decent results in the set of Zogby Interactive polling -- which, unfortunately for McCain -- you should probably take no more seriously than this one (http://www.theonion.com/content/news_briefs/zogby_poll_john_zogby). We could stand to get some fresher numbers from Pennsylvania, but Morning Call has been running a tracking poll there for about a month now, and McCain hasn't been within single digits since roughly the 1st of the month.

[AsburySkinsFan]

Dose of reality
National polls mean nothing
State polls mean everything

Do away with the electoral college and maybe then I'll care about what the National polls say, until then show me the battle ground states.

[Henry]

Dose of reality
National polls mean nothing
State polls mean everything

Do away with the electoral college and maybe then I'll care about what the National polls say, until then show me the battle ground states.

Doesn't look great for McCain there either.

http://www.realclearpolitics.com/epolls/maps/obama_vs_mccain/

[AsburySkinsFan]

Doesn't look great for McCain there either.

http://www.realclearpolitics.com/epolls/maps/obama_vs_mccain/

And that's the good news. ;) But, turn out is key!

[techboy]

Dose of reality
National polls mean nothing
State polls mean everything

Do away with the electoral college and maybe then I'll care about what the National polls say, until then show me the battle ground states.

This is both true and false.

It is true because of what you say.

It is false because if McCain moves strongly enough in the national polls, the good PR is likely to filter into the state polls.

[PeterMP]

This is both true and false.

It is true because of what you say.

It is false because if McCain moves strongly enough in the national polls, the good PR is likely to filter into the state polls.

I don't believe this poll, but in reality it's essentially impossible to move a national poll at this point in time AND not move the important state polls. Historically, undecidedes are more concentrated in "battle ground" states. This becomes even more true as you get closer to the election. Most of the far-right wingers in Texas that are going to vote for McCain have already come over to McCain to the point that there aren't enough left to move the national polls.

Plus, a good bump in a national poll is good PR, and there is a real "I want to vote for a winner." and "I don't want to vote for a loser" effect.

[Lombardi's_kid_brother]

The polls ALWAYS supposedly tighten during the last week. And then they seem to end up where they had traditionally been.

And the battleground state polls have not moved in weeks.