More Polls Find Obama Leading In Multiple Bush States

[Richard Tafoya]

TPM:
http://tpmelectioncentral.talkingpointsmemo.com/2008/10/still_more_polls_find_obama_le.php

Still more polling, this time from the Associated Press (http://news.yahoo.com/s/ap/20081029/ap_on_el_pr/ap_poll_battlegrounds), paints a bleak picture for McCain in at least six Bush states:
Barack Obama now leads in four states won by President Bush in 2004 and is essentially tied with John McCain in two other Republican red states, according to new AP-GfK battleground polling... The polling shows Obama holding solid leads in Ohio (7 percentage points), Nevada (12 points), Colorado (9) and Virginia (7), all red states won by Bush that collectively offer 47 electoral votes....
In addition, Obama is tied with McCain in North Carolina and Florida, according to the AP-GfK polling.
To put this in perspective, consider that if Obama holds the Kerry states, and doesn't win either Ohio or Florida, he can still win by putting together a combination of two or three out of Colorado, Virginia, Nevada, and North Carolina, three of which he's leading by sizable margins.

And that's not even including the multiple other Bush states he's leading in, such as New Mexico, Iowa, and Missouri. In short: Even without Ohio or Florida, getting the Kerry states, plus two or three of these seven -- CO, VA, NC, NM, IA, MO, NV -- makes Obama the next President.

[Murrican]

Check the source, eh.

[Richard Tafoya]

The polling data is from Associated Press. The electoral analysis from TPM. The math, fairly basic approaches to a 270 sum.

[Murrican]

TPM is a distinctly biased source. That's clear from the content of their site.

[Richard Tafoya]

So the same data on Michelle Malkin's site would show McCain with an electoral advantage? Or you could make that case. How does McCain win via the lens of this latest AP polling?

[Murrican]

I'm not saying how who wins. I'm saying there is bias in the media, no matter the media, so knowing the source (and methodology of polling and poll analysis) is important.

That said, here's something from TPM on "push polling", though it inexplicably doesn't call it that...

And hence one of the problems with polls.

How Media Polls "Manufacture" Public Opinion

By David Moore - October 26, 2008, 7:15PM

While our focus on this special Talking Points Memo segment has naturally been on election polls, another area of concern is public policy polls. Typically, media polls are designed not to report public opinion but to manufacture a "public opinion" that is interesting and plausible for the news stories - often at the expense of anything that resembles what Americans are really thinking.

The major tactics media pollsters use are 1) asking forced choice questions, to get answers even from people who have no opinion; 2) providing information to respondents in the interview, in case they don't know enough already to offer an opinion - and in so doing taint the sample, so that it no longer represents the general public that has not been given that same information; and 3) failing to provide meaningful intensity measures that could help distinguish firmly held views from the "top-of-mind" views that respondents express in the press of a quick telephone interview.

Typically, the news media do not want to report polls that show a significant segment of the public unengaged and without an opinion on an issue. So pollsters "force" (pressure) respondents to come up with an answer, no matter how lightly held. And if respondents don't know enough, pollsters provide them some limited information.

The net result is that often media poll results conflict greatly with each other, because different polls give different information to respondents, and thus influence them to give different answers. That was the case recently with polls that asked people their reactions to the efforts by Congress and the President to address the economic crisis.

On Sept. 23, a Los Angeles Times/Bloomberg poll reported that Americans opposed the congressional bailout plan by a 24-point margin, 55 percent to 31 percent. That story was cited frequently on various television and radio shows as evidence that Americans were "angry" about the bailout. However, the very same day, a Pew Research poll, conducted over the same time period as the Times/Bloomberg poll, found that Americans supported the Wall Street bailout by a 27-point margin, 57 percent to 30 percent.

rest at: http://tpmcafe.talkingpointsmemo.com/2008/10/26/how_media_polls_manufacture_pu/

[LesterX]

So the same data on Michelle Malkin's site would show McCain with an electoral advantage? Or you could make that case. How does McCain win via the lens of this latest AP polling?

Exactly. It's an AP poll. The fact that it is posted on TPM does not change the numbers, so TPM's "bias" is irrelevant.

[Murrican]

So the same data on Michelle Malkin's site would show McCain with an electoral advantage? Or you could make that case. How does McCain win via the lens of this latest AP polling?

Don't know Michelle Malkin. Don't know the AP methodology, how they render their error margins, their historical accuracy, the geographic and precinct targeting, any bias, if any (that is, old, young, ethnic, previous voters, algorithm adjustment, polling times, etc.), and so forth. Any place you can point to which has info on the above would be useful. Also, what is AP trending? And how does it compare to other polls from same time and in same place? Pew, Rasmussen, Zogby, et al.

To be simple, not knowing anything more, McCain wins by the undecided breaking to him more than to Obama, especially in key states, and by poll results being skewed by everything from the Bradley effect to the bias of the poll employees on the phone and how they ask the questions... The bigger bias can come, of course, from collateral questions. If the following are asked you can imagine a result going in different directions -- who you fear?, importance of security?, importance of experience?, success of Iraq "push"?, 9/11 repeat possibility?, etc. Too many polls try to find out too much while they have the pollee on the line. I say all of this from a decades-long background in journalism, politics and being polled...

If there's an electoral college tie, remember, Dick Cheney casts the deciding vote... <egad>

[Richard Tafoya]

TPM's poll tracker links to the source of every poll. You can get your sampling and methodology info there. Fivethirtyeight.com and realclearpolitics.com will probably also link to or report those things.

[Murrican]

http://www.scoop.co.nz/stories/HL0810/S00371.htm

Polls, shifting, anomalies and surprises... And how numbers can mislead, lie and confound...

Gordon Campbell: Can McCain Make A Comeback

Monday, 27 October 2008, 4:10 pm
Article: Gordon Campbell

Only a week to go, and have John McCain’s attempts at a comeback peaked too soon? Eight days ago, the US opinion polls began to carry a few flickering signs of a McCain resurgence - after Barack Obama had pulled clearly ahead in the wake of the second presidential debate and the financial meltdown on Wall Street. Last Friday and Saturday for instance, both the Associated Press and the IBD/TIPP polls had McCain at a virtual tie with Obama nationally, or within a margin of error striking distance.

There’s more. In some battleground states that McCain needs to win to stay viable there have been inconclusive signs of a Republican revival. Check out for instance, the McCain rebound on the weekend in the two Strategic Vision polls ( see the Pollster.com page for full details) in Florida and in Ohio – though McCain still trails in other Ohio polls in what is a major ‘must win contest’ for him. More on the Ohio situation below.

True, in Virginia, Nevada and North Carolina, some of the McCain bounce-back is occurring in places previously seen as being relatively safe Republican states – and where the Obama legions have outperformed the Republicans on the ground - so any poll movement may simply be a wavering return to natural allegiances. While he has been flat-lining nationally, McCain has recovered a bit of ground in a few key states since the third and final televised debate : in that respect, Joe the Plumber seems to have done more for McCain than the Colin Powell endorsement has done for Obama.

Obama must still be considered the odds on favourite on November 4th, however. That McCain-favourable IBD/TIPP poll mentioned above for instance, contained a bizarre result among 18-24 year old young voters ( it put McCain ahead in that age group by a 78-22% margin ! ) that even the pollsters themselves were dubious about. Nate Silver, the baseball statistician turned political guru for the 2008 election, has analysed the IDB/TIPP poll on his Five Thirty Eight pollwatch website, and concluded that the odds for this 18-24 segment being accurate by chance, were something like 55 billion to one against.

Will race be decisive? In recent weeks, there has been an immense volume of media time and energy devoted to the so called Bradley Effect, which refers to the claimed distortion in the polls caused by white voters being unwilling to tell pollsters about their true intentions – namely, that they won’t be voting for that black guy for president. Note : the Bradley Effect is not about whether some people vote on the basis of race – clearly, some people do – but on whether they lie to pollsters about it beforehand, and are thus making Obama’s poll numbers look deceptively better now than they will turn out to be on election day.

The Charleston, Virginia Daily Mail has a pretty good general discussion here of the Bradley Effect, with one commentator estimating it to be causing about a 1-3 % inflation of Obama’s actual margin over McCain – which is roughly about the same as the “point or two in the polls” estimate made by the key Obama Harold Ickes, in a recent New York Times article.

Even that tiny fluctuation, given the wafer thin margins that exist in some states, could make some difference for McCain. On the other hand, there is a comprehensive debunking of the Bradley Effect in this week’s edition of Newsweek, by the ubiquitous Nate Silver, who has drawn on research into election contests held since 1980 ( the work was carried out by Harvard fellow Daniel Hopkins) to argue that the Bradley Effect finally petered out sometime in the mid 1990s.

rest at: http://www.scoop.co.nz/stories/HL0810/S00371.htm

[Murrican]

TPM's poll tracker links to the source of every poll. You can get your sampling and methodology info there.

With all due respect the methodology info is incredibly thin.

To wit:

The AP-GfK Battleground State Poll was conducted from Oct. 22-26 in eight states. It involved interviews by landline telephone with likely voters in each state, ranging from 600 in Florida and New Hampshire to 628 in Nevada. The margin of sampling error was plus or minus 3.9 percentage points in Colorado and Nevada, and 4 points in the other states.

from: http://news.yahoo.com/s/ap/20081029/ap_on_el_pr/ap_poll_battlegrounds

Nothing about:

how they render their error margins, their historical accuracy, the geographic and precinct targeting, any bias, if any (that is, old, young, ethnic, previous voters, algorithm adjustment, polling times, etc.), and so forth.

to quote myself from above...

But the error margins alone make a huge difference in the conclusions, if the error margins are accurate.

To quote: http://news.yahoo.com/s/ap/20081029/ap_on_el_pr/ap_poll_battlegrounds again:

Indeed, polls are mere snapshots of highly fluid campaigns, and this race has been unusually volatile. McCain was written off prematurely last year, and Obama seemed poised for victory in New Hampshire's Democratic primary just before Hillary Rodham Clinton thumped him.

The ground games are critical now, hence the Obama TV buy tonight, it's plea for supporters to take E-Day off from work or school to help, etc. And the Obama ground game is likely weakest in states Kerry lost in 2004 and Gore lost in 2000.

And a professor on TV just said the advance youth vote in Florida has been lower than expected, which supports what I just wrote about the importance of the upcoming ground game. If that's happening elsewhere, we may have a November surprise, especially for sofa-hugging supporters (i.e. ones who don't vote but support a candidate)...

M