http://www.democratandchronicle.com/article/20080925/NEWS01/80925009/1002/NEWS [FONT=arial, helvetica][FONT=Times New Roman, serif]Pollster says election could end in landslide[/FONT] [FONT=Times New Roman, serif] Jill Terreri Staff writer [/FONT] [/FONT] [FONT=arial, helvetica]The presidential election might be a tight race now, but one of the country’s top pollsters thinks the race will end in an electoral landslide. John Zogby, president of Zogby International, told a group of businesspeople today that it’s up to Democratic Sen. Barack Obama to convince voters to go with him. If he’s not successful, the country will likely vote for “a comfortable old shoe”, that being Republican Sen. John McCain. Despite the books Obama has written, Americans are still asking, “Who are you, where are you from?,” Zogby said. Zogby spoke at the College at Brockport’s Business Briefings breakfast series at the college’s MetroCenter campus on St. Paul Street. He was promoting his new book, The Way We’ll Be: The Zogby Report of the Transformation of the American Dream. [/FONT]
http://www.zogby.com/news/ReadNews.dbm?ID=1562 Released: September 25, 2008 Zogby Poll: McCain Recovers as Contest Takes Dramatic Turns; McCain 46% - Obama 44% First presidential debate still up in the air as campaigns shuffle schedules UTICA, New York - Republican John McCain's poll numbers improved slightly as he suspended his campaign Wednesday to head back to Washington to focus on the looming national financial crisis, moving from more than three points behind Barack Obama last weekend to two points ahead in a Zogby Interactive survey just out of the field this morning. <table border="1" cellpadding="0" cellspacing="0" width="375"> <tbody><tr> <td> [SIZE=+1]Data from this poll is available here[SIZE=+1] [/SIZE]</td> </tr> </tbody></table> [/SIZE] What is still unknown is what will happen to the first presidential debate in Mississippi, which had been set for Friday evening but which has been put on hold by the McCain campaign. The sponsor of the debate and the Obama campaign insist the debate will go on. Both surveys of likely voters nationwide were conducted using Zogby's online polling technology, which has proven accurate in national and statewide races dating back to 2004. McCain now leads Obama, 45.8% to 43.8%, the survey shows. <table border="1" cellpadding="0" cellspacing="0"> <tbody><tr> <td width="139"> The Horserace </td> <td valign="top" width="96"> Sept. 23/25 </td> <td width="108"> Sept. 19/20 </td> </tr> <tr> <td valign="top" width="139"> Obama </td> <td valign="top" width="96"> 43.8% </td> <td valign="top" width="108"> 46.8% </td> </tr> <tr> <td valign="top" width="139"> McCain </td> <td valign="top" width="96"> 45.8% </td> <td valign="top" width="108"> 43.4% </td> </tr> <tr> <td valign="top" width="139"> Not sure/Other </td> <td valign="top" width="96"> 10.4% </td> <td valign="top" width="108"> 8.8% </td> </tr> </tbody></table> The big shift in the race appears to have come among independent voters, where McCain now leads by nine points, 43% to 34%. In the survey conducted over the weekend, Obama led by one point among independents. Both candidates have a sturdy grip on their political bases, the survey shows. McCain and Obama each win 88% support from voters in their respective political parties. Among men, McCain leads by a 53% to 35% margin, up 15 points from the weekend survey. Among women, Obama leads by a 52% to 39% margin, up 5 points from the weekend survey. The survey, half conducted before McCain's announcement Wednesday that he would suspend his campaign to concentrate on the financial crisis and half conducted after the announcement, shows movement in McCain's favor after his announcement. Before the announcement - which included about half of the total polling sample - Obama led by one point. But McCain led by 5 points in polling completed after his statement about the suspension of his campaign. Overall, the interactive survey, conducted Sept. 23-25, 2008, included 4,752 likely voters nationwide and carries a margin of error of +/- 1.5 percentage points. Pollster John Zogby: "The financial crisis appears to have trumped the campaign at least for now, but what remains true is that this race is really very close. Obama was clearly leading before - we had him up by three points over the weekend - but I never thought his lead was as high as nine points, as at least one poll had indicated. We are careful to weight our poll samples to reflect the proper proportion of Democrats, Republicans, and independents. And of course, we always sample likely voters, not registered voters, to most closely reflect what would happen in an election." Zogby International was the most accurate pollster in every one of the last three presidential election cycles, and continues to perfect its telephone and interactive methodologies using its own live operator, in-house call center in Upstate New York, and its own secure servers for its online polling projects. In the 2004 presidential election, not only was Zogby's telephone polling right on the money, its interactive polling also nailed the election as well. In 2006, the Zogby Interactive online polling was on the money in 17 of 18 U.S. Senate races (the 18<sup>th</sup> was within the margin of error) a record of accuracy that is unmatched in the industry - as no other leading firm even attempts to poll statewide political races using an interactive methodology for public consumption. For a complete methodological statement on this survey, please visit: http://www.zogby.com/methodology/readmeth.dbm?ID=1336
I hope I don't have to post this too often. RCP is bogus. It includes outlier polls that are ridiculous.
Didn't know Rasmussen had fallen that far. :O :[ http://www.rasmussenreports.com/pub...ial_election/daily_presidential_tracking_poll
This zogby poll is from today, and includes 1/2 of the calls after McCain's suspension of his campaign. But I think all the daily poll watching is missing the point. McCain now leads Obama, 45.8% to 43.8%, the survey shows. <table border="1" cellpadding="0" cellspacing="0"> <tbody><tr> <td width="139"> The Horserace </td> <td valign="top" width="96"> Sept. 23/25 </td> <td width="108"> Sept. 19/20 </td> </tr> <tr> <td valign="top" width="139"> Obama </td> <td valign="top" width="96"> 43.8% </td> <td valign="top" width="108"> 46.8% </td> </tr> <tr> <td valign="top" width="139"> McCain </td> <td valign="top" width="96"> 45.8% </td> <td valign="top" width="108"> 43.4% </td> </tr> <tr> <td valign="top" width="139"> Not sure/Other </td> <td valign="top" width="96"> 10.4% </td> <td valign="top" width="108"> 8.8% </td> </tr> </tbody></table> Obama lost 3%, and Not sure gained 2.6%. What Zogby was talking about in the first article is that Obama is unable to seal the deal and put the race away. Instead, he's leaving people truly undecided and at the last minute (in the booth), they'll vote for McCain. The race factor and the Hillary factor are at play, too. Some % are lying to the pollsters and won't vote Obama because of his race. Sucks, but it's true. And there's 20% of Hillary's voters, or as many as 6M votes headed to the McCain column in Nov.
Outlier polls are part of the data, assuming the method is at all rigorous. There's no statistical validity to simply throwing out certain polls that you don't like or don't think show reality. Polls don't show "reality," they show a snapshot of moods, and that's best gauged through the averaging of all the data. Fivethirtyeight.com and electoral-vote.com both have Obama with significant leads in the electoral college based on the latest state polls. Two weeks ago, both had McCain in the lead, but polls swung big for Obama last week (which pretty much everyone agrees on, not just a few "liberal" polls).
Look at Zogby's map though. He has 234 in the Obama category, with Pennsylvania and New Mexico as tossups, both of which looks pretty safe Obama. So on his own map, he has Obama basically with that 260 EV's, with Colorado, Nevada, New Hampshire, Indiana, Ohio, and Virginia as toss up. Obama just needs NV+NH, or any of the other states to get to 269 (at which point it would be decided in his favor in the house).
I would wake up and pay attention when a highly regarded and accurate pollster who's a Democrat (Zogby) comes out and says what he did. He's not looking at the trivial swing back/forth - as you said, McCain up last week. When you have 10 polls and 9 are within +/- 2 and the other is +15, then the +15 really looks out of whack, don't you agree? When you average the 9 and get EVEN, but average in that +15 and get one up by +2.5, then it makes RCP bogus.
You think a professional pollster would throw that word around lightly? Nobody would take him seriously if he misused the basic words of electioneering.
He's not unique in that. Almost no major statistical models are affected by the trivial back and forth. Fivethirtyeight.com lagged behind the mounting state polls coming out big in favour of Obama because the model wants to see steady change to change its own forecast...exactly to avoid "trivial swings" back and forth. Not at all. Even single polls that are surely wrong contain useful information for modeling. If one candidate is actually ahead, one would expect some outliers that show him WAY ahead. Think of a locus of points around the "real position"...if someone is up 5, the high point of that spread could be +15. Whereas if the candidate is down 5, his high point might be +5. You want as much data in the "data cloud"...the bigger the data cloud (as long as it is real information), the more confident that the center of the cloud is "true."
I still don't see why you regard Zogby so highly. During the primaries, you have: SurveyUSA's pollster report cards: 1. Market Shares 2. ABC News 2. Scroth and Associates 2. Behavioral Research CTR 5. Selzer and Co 6. Fox News 6. Field Poll 6. Greenberg Quinlan 9. SurveyUSA 10. Franklin Pierce 10. Monmouth 12. Gallup 13. Research 2000 14. Quniipiac .... 17. Strategic Vision .... 21. Rasmussen 22. Zogby 23. ARG 24. Insider Advantage 25. Public Pollicy Polling Then using Five Thirty Eight's report card (which weights states in difficulty to direct, so states who avoided tough contests to poll don't get extra credit). 1. Selzer and Co 2. SurveyUSA 3. Rasmussen 4. U of NH 5. Market Shares 6. Field Poll 7. Mason-Dixon 8. Research 200 9. Quinnipiac ... 12. Public Policy Polling 13. Strategic Vision 14. Insider Advantage .... 16. ABC / Washington Post 17. Zogby .. 20. Suffolk 21. CNN / Opinion Research 22. LA Times / Bloomberg 23. ARG 24. Fox News 25. Gallup ... 31. Zogby Interactive
The outliers may be useful if you watch the trend in a single one of those polls over time. There's virtually no information in the outlier if 9 polls have McCain up 10 and one has Obama up 15. Or vice versa.
Well, Zogby has Pennsylvania as leaning toward McCain by 6%. Still, though, he credits Obama with 234 electoral votes, which means Obama needs just 35 more votes. In all the "purple" states, right now Zogby says Obama loses ever one but North Carolina (15 votes) and New Mexico (5), and Indiana "too close to call" (11). Pretend that there's absolutely no margin of error, by Zogby's numbers it seems like it comes down to Indiana, or 11 electoral votes. Hardly a "blowout." But then a lot of people are saying 3% of people will change their vote in the privacy of the booth just because Obama is black. So he needs a big margin if he has any hope to win. But people also said the "cell phone vote" would be huge in 2004 for Kerry and it didn't emerge. *Shrug* This is the very first time I've really looked at any poll in any detail, and now I remember why. So much can change day-to-day, week-to-week. Who knows what to really think. If I were a Republican and I was excited to see Sarah Palin with her finger so close to the nuclear button, I guess I'd be happy to read the article. I just don't know that I'd really believe it.
Mook, I think Zogby is looking at everything with the lens he has as an experienced pollster. I'm not going to say it's his gut that causes him to talk about a landslide, but it's not likely today's poll, or last week's. Everything taken as a whole, including trends and expectations.
The outliers are useful only in combination with multiple polls. A single outlier poll, taken alone, tells you nothing at all. A bunch of polls, the outliers are useful information. Someone who "really is" way down, won't get an outlier showing him way up. So an outlier showing someone way up gives you an indication that the candidate isn't way down. As an example. But this is only true if taken in combination with other data. Statistical method never throws out data, unless it's simply fraudulent or the method of acquiring it was invalid.
Pattern matching by an expert is an interesting human capability, but I certainly wouldn't trust it when it comes to enormous amount of data. Humans aren't wired to process huge amounts of data well. Zogby's opinion is interesting, but not much moreso than any pundit who has a lot of experience. If he's not using his own statistical results, then he's talking as a pundit, not as a statistical analyst.
