Two new polls today ...

[anatta]

AP Poll:

In the survey of 976 likely voters, Democrats Kerry and Sen. John Edwards (news - web sites) had 49 percent, compared to 46 percent for Republicans Bush and Vice President Dick Cheney (news - web sites). That's within the margin of error for the poll conducted Oct. 18-20.


Zogby: A Reuters/Zogby three-day tracking poll released on Thursday showed Bush edging Kerry 46-45 percent.

College student support Kerry: 52-39. From some Yahoo story:


Just weeks before the Nov. 2 election, researchers at Harvard's Institute of Politics found that 52 percent of all students want the Massachusetts senator elected president, 39 percent support Bush, and 8 percent are undecided.


In 14 hotly contested swing states, the poll shows Kerry leading Bush by 17 points among students.


The data suggest more students are leaning toward Kerry than six months ago, when Harvard last surveyed them. That poll, released in April, found Kerry leading Bush by 48-38 percent with 11 percent undecided.

The article goes on to say turnout is expected to be higher than in 2000, as interest indicators are through the roof.



I am suprised the gambling line favors Bush so much. It really looks more like a coin toss each day to me.

[tanda]

The AP poll is not credible.

See polipundit.com

[ QUOTE ]
A Little Incest, or a LOT?


The latest poll sponsored by Ipsos made its’ splash two days before it hit the water, which got me curious. When the full poll finally showed up this morning, I started to look into it, and bells went off. Not the pretty ones, either.

This poll claims Bush gets only 28% of the Independent vote (versus 44% in the average from other polls), while Kerry claims 56% of Independents (versus 45% in the average from other polls). You can throw that part out, because it’s total crap. That’s because Ipsos only interviewed 49 people in the entire country, identified as Independents, or only 5.19% of their unweighted base. even with reweighting, they only work out to 5.70% of the pool. This is out of line with EVERY established weighting used by anyone, anywhere this year. Any agency seriously interested in representative results, would have continued polling to get a more representative sample.

I know why the Ipsos poll only got 49 Indies, though. In their published report, they mention that they were aiming to get a response from people who had seen the debate, and that played heavily to the Left and Right. In other words, their chosen method passed by undecideds as a rule. Sorry to be mean, but that is indescribably stupid. By deliberately leaning to those with an interest in the debates above any other factor, Ipsos invalidated their findings, ab initio. I will demonstrate this when I review the numbers, in just a moment.

As I reviewed the numbers, I was struck by how peculiar they seemed, and I decided to go back and double-check the results from other polls. In doing so, I came across the October 2 Newsweek poll. You know, the one which heavily over-sampled the Mountain and West regions by sampling only there on September 30? As I read through the article again, I came across a graphic showing the Bush/Kerry matchup since July, and in the detail under the graph, what did I see but this comment:

“Source: Newsweek/Ipsos poll”

Oh really? The only two major polls which say Kerry took the lead, and the only polls to have done some very unscientific things in their sampling, happen to be connected to the same sponsor?

Yes. Exactly that.

I went to the Ipsos site to be sure, and there it is. Ipsos is a regular sponsor for Newsweek’s ‘GENext’ poll, which tracks young voter sentiment, and is directly tied to the AP/Ipsos poll.

The same. They don’t advertise it, but they have been there together all year long.

Well, now, is that important? To find out, I continued double-checking my numbers, and looked up the numbers for poll results, gender preference, race preference, party support, Job Approval, candidate personal favorability, and whether the respondents watched the debates, for the following polls: CBS News, Fox News, CNN/USA Today/Gallup, Pew, ABC News, the LA Times, Newsweek, and AP/Ipsos. The first six polls make up the ‘non-Ipsos’ polls, the last two are the ‘Ipsos’ polls.

The non-Ipsos polls says Bush leads 49.2-45.5. The Ipsos polls say Kerry leads 48.5-45.5.

The non-Ipsos polls say Men prefer Bush, 53.7-41.7. The Ipsos polls say it’s only 47.5-46.5 for Bush.

The non-Ipsos polls say Women prefer Kerry by only 47.0-46.3. The Ipsos polls says women prefer Kerry, 51.0-43.0.

The non-Ipsos polls say Whites prefer Bush 55.0-39.7. The Ipsos polls say it’s only 49.0-47.0.

The non-Ipsos polls and Ipsos polls agree that non-Whites prefer Kerry, by 66 or 67, to 26.5 or 27.

The non-Ipsos polls say Bush gets 93% of the Rep vote, 10% of the Dem vote, and 45% of the Ind vote. The Ipsos polls say Bush only gets 90% of the Rep vote, 8% of the Dem vote, and only 33% of the Ind vote.

The non-Ipsos polls say Kerry gets 4% of the Rep vote, 87% of the Dem vote, and 45% of the Ind vote. The Ipsos polls say Kerry gets 6% of the Rep vote, 89% of the Dem vote, and 49% of the Ind vote.

Non-Ipsos polls say Bush’s Job Approval is at 51%. Ipsos polls say Bush’s Job Approval is only 46%.

Non-Ipsos polls put Bush’s personal favorability at 51%, and Kerry’s at 47%. The Ipsos polls put Bush’s personal favorability at 49%, and Kerry’s at 52%.

In every category, the Ipsos polls are stronger for Kerry, even though six major polls say otherwise. Given the known fact that the Ipsos/Newsweek poll ignored the East, Midwest, and South on the first day of polling, and the inexcusable inbalance in party sampling on the Ipsos/AP poll, the Ipsos results may reasonably be regarded as INVALID.

There is a final point to observe in these details. Non-Ipsos polls reported that about 26% of their respondents watched some or all of the first debate, and had no report for specifically how many of their respondents watched all of the first debate. The Ipsos polls reported that 78% of their respondents watched at least part of the first debate, and 56.5% of their respondents watched all of the first debate. This is such a small demographic, relative to the country as a whole, that the Ipsos numbers must be considered skewed by definition. Networks reported that approximately 55 million people watched some of the debate (approximately 18% of the viewing public), and approximately 19 million watched all of it, for 6.3% of the viewing public. In other words, Ipsos aimed to miss over 90% of the viewing public, and more than two-thirds of registered voters, by the pre-conditions imposed on the Newsweek and AP polls.

The remaining question to be answered, then, is whether this was a little incest, or a lot.



[/ QUOTE ]

This was their last poll. Less than 6% IND. Deliberately preferred first debate watchers, thus benefiting Kerry. Never disclosed these sample parameters prominently. Buried them in the internals which they refused to release for 1-2 days.

L.A. Times did the same thing but instantly and prominently disclosed the sample was primarily debate watchers and for this reason disclaimed it as a representative poll. AP should have done the same.

As of now, AP refuses to release internals on the new poll.

Bush ticked upwards in all tracking polls today: Rasmussen, Zogby, both TIPPS, WaPo and ABC.

Bush is now 22-1-3 in the last 26 polls. Kerry has led 4 of last 52. In the last 72 polls, the two AP polls are 2 of only 3 that have had Kerry up by more than 2 points.

The Mason-Dixon polls have Bush 7-0-0 in red states and 3-3-1 in blue states. Gallup WI has Bush +6, one new poll and a rumored new poll in MI have Bush up.

Mason-Dixon was 21 for 21 in their 2000 polls and 22 for 23 in their 2002 polls.

Coin flip?

[anatta]

I wasn't aware of that info. Coin flip? Sure, Heads Kerry wins, Tails Bush Loses. (It was heads, sorry, we can do 2 out of 3 if you want).

[Dynasty]

For those following the polls, www.realclearpolitics.com always has the most recent polls listed at the top of their page, including new state polls.

Everyday, there are four tracking polls updated. These are today's numbers.

Rasmussen: Bush 48, Kerry 45, Nader 1
TIPP: Bush 48, Kerry 45, Nader 1
Washington Post/ABC: Bush 51, Kerry 45, Nader 1
Zogby: Bush 46, Kerry 45, Nader 1

Usually, there are other non-tracking polls too. These are new today.

AP-Ipsos: Kerry 49, Bush 46, Nader 2
Marist (w/leaners): Bush 49, Kerry 48, Nader 1

I tend to just ignore outlier polls which in this case would be the Washington Post and AP-Ipsos polls. The average of the other four new polls is:

Bush 47.75, Kerry 45.75, Nader 1

I think it's close. [img]/images/graemlins/smile.gif[/img]

[anatta]

Did you see Ohio Gallop with Kerry up five among registered voters? LV much closer. I don't think Gallop is as accurate as they could be with LV's this year, but its possible they are correct, that OH dems have registered lots of suspect voters, but I tend to think they will come out. It really looks like its coming down to the ground game.

I don't trust Mason-Dixon that much, some maybe. Their numbers appear slightly scewed towards Bush in General. ie. Oregon 1 point. PA 1 point. I think Kerry is up by more in these States.

Do you know anything about their poll?

[tanda]

[ QUOTE ]
Do you know anything about their poll?


[/ QUOTE ]

Mason-Dixon went 21-0 in states in 2000 and 22-1 in 2002 from the claim of a poster I trust. Also, from another source, in an evalution of public firms, they finished 1st 8 times and 2nd 5 times in the last 13 elections. These claims are consistent with the high reputation the firm has. Nonetheless, I have not independantly verified the claims.

For state polling, I trust Mason-Dixon, Gallup to some extent, Opinion Dynamics and Quinnipac the most. Certain state specialsis such as Des Moines Register and Columbus Dispatch are good as well.

ARG seems good as well, a little biased to the left, but consistent so that trends can be detected using their polls. Even a biased poll can be useful, if consistently biased, to spot trends. SurveyUSA and Rasmussen are useful as secondary polls but not primary polls. Zogby Interactive is awful.

[Cyrus]

Every statistician worth his Ambivalent Hairline will tell you that such flip-flopping poll results only mean one thing : statistical error.

Yes. It's a coin flip.

(Maybe Kerry has a slight edge in this case? Being a flipper and all that?.. [img]/images/graemlins/cool.gif[/img] )

[SinCityGuy]

He will get less than 1% of the vote, and most of them wouldn't even bother to vote if he wasn't on the ballot. It's not like 2000.

[sam h]

[ QUOTE ]
Bush ticked upwards in all tracking polls today: Rasmussen, Zogby, both TIPPS, WaPo and ABC.

[/ QUOTE ]

True enough. But if John Zogby sees a 47 - 45 spread on his last poll, I think there is a good chance he will predict Kerry to win.

Moreover, many people believe that if you are below 48 you cannot possibly win the EC, if you are above 52 you cannot possibly lose it, but that if you are in the middle, especially between 49-51, then there is a high enough chance of different candidates winning the popular and the EC that the state polls really are the ones to be watching.

[ QUOTE ]

Bush is now 22-1-3 in the last 26 polls. Kerry has led 4 of last 52. In the last 72 polls, the two AP polls are 2 of only 3 that have had Kerry up by more than 2 points.

[/ QUOTE ]

Which 26 polls? Link?

[ QUOTE ]


Mason-Dixon was 21 for 21 in their 2000 polls and 22 for 23 in their 2002 polls.

[/ QUOTE ]

Link?

[nyc999]

From a researcher -- take these polls with a grain of salt. Zogby has probably been the most accurate in the past (especially in their sampling), but there has been such aggressive methods to register new voters by both parties that most organizations have admitted that they are having problems accounting for all these new voters.