Posted on 10/31/2004 5:15:03 PM PST by Cableguy
Besides the chart, there is also a table on that page. Both, regardless, clearly say October 29-31.
Hmmm FUNNY how the CNN/Gallup poll is reported differently on CNN then what this article says. Here is the crucial point from the CNN Website.
"Using voting behavior data from pervious elections, the Gallup organization attempted to estimate how the undecideds would vote Tuesday.
The result was a tie of 49 percent each for Bush and Kerry, with 1 percent for Nader and 1 percent for other candidates."
They GUESSED to get the 49-49 tie you guys are hyping. Note also. ALL the "hisotrical data" is pre-9/11.
So instead of 49-49 it is actually B-49. Kerry 47 2% other or undecided.
No. If Bush was polling +15, (65%) the contract would be at 90+.
What happened to your argument saying that the poll data I posted is a week old? Please research before making false accusations next time. Thanks Wildcat!
Now you are trying to change the argument.
It isn't really worth a drawn out argument, but the fact you won't admit you made an error in attacking the poster is really childish. Come on, dude, don't be so bitter.
| Kerry/ Edwards | Bush/ Cheney | Nader/ Camejo | Other (vol.) | None (vol.) | No opinion | |
| Likely Voters | ||||||
| 2004 Oct 29-31 ^† | 47 | 49 | * | 1 | * | 3 |
| Final Allocated Estimate | 49 | 49 | 1 | 1 | -- | -- |
I reported what was on Gallup and USA Today websites. You first said it was old. Now you say that because Gallup/USA Today didn't initially clarify what they meant by LV before I posted the results, you me. Never mind that Gallup/USA Today before today this year never allocated undecided to LV.
You are a *!@#$ and I am going to stop responding to your idiotic posts.
PS: The link takes you to the 10/29-31 sample, if you can figure out how to work your mouse. Everyone has figured it out on this thread except you.
Gee wonder what I am suppose to make of this.
Pew has it 51%-48% for Bush
http://people-press.org/reports/display.php3?ReportID=232
"IM REALLY starting to think EVeryone wants to be tied so no one is wrong!"
I'd say that is a very astute observation.
I found this interesting at the horserace blog site:
http://jaycost.blogspot.com/2004/10/scoop-on-polls.html
Thursday, October 28, 2004
The Scoop on Polls
This evening, I found myself becoming genuinely frustrated with all the polls. Part of this is my fault. I decided to iron out the statistical problems inherent to predicting a presidential election -- and just not worry about the methodological problems.
I regret having done that. I am inclined to discontinue my statistical analyses. The reason for that is, upon doing some research, I am shocked at the number of polling outfits that are generally unreliable. I plan to make my living as a political methodologist, and it was foolish of me to take my methodological cap off. It is what I am best at.
In particular, I have become vexed by the issue of partisan weighting, though other issues have come to the forefront of which I was unaware. I think it important to pass this information along.
Of all the characteristics which go into public decision-making over presidential elections, none is more important than partisan identification. Since the publication of Cambell, Converse, Miller and Stokes's The American Voter in 1960, academics have understood that there is no single pre-existing factor that predicts a person's vote as well. Democrats vote for Democratic candidates, even if they consider themselves to be conservatives. Ditto for Republicans.
Thus, partisan identification matters for presidential polls. There are two schools of thought on this issue. Some polls weight their respondents so that their ultimate sample composes a pre-determined set of Republicans and Democrats. Some polls do not. Both schools of thought are all right in my book. The question to ask is that of the pre-determined crowd -- whether their weights will conform with the ultimate result on election day.
This is where presidential polling intersects with the dynamics of the horserace. While an individual's partisan inclinations are relatively stable over time, the two parties can manipulate the ultimate turnout ratio on election day through GOTV efforts. This is why GOTV is so incredibly important.
Judging from everything I have seen this election year -- from GOP organizational excellence to Democratic and 527 organizational incompetence -- I believe that this year's partisan ID breakdowns on election will be more Republican than we have seen in a presidential election. Thus, polling firms that weight their samples by partisan identification should account for the known events on the ground. Unfortunately, precious few of them do. In an election such as 1996, when the presidential election was not close, deviations from the ultimate partisan breakdown of the electorate is not problematic. However, when the difference between the candidates on 11-02 could be as little as 2%, improper partisan identification weights can result in an inaccurate prediction.
Another issue that has come to my attention is the problem of methodological publicity. As a person involved in academia, I cannot tell you how important it is for academics to make their methods available to the public at large. Most academic articles go to great lengths to explain their method before they actually provide you their results. The reason for this is that method matters. Unfortunately, a surprising number of polling firms do not make their methods available.
In addition to my own snooping on the internet, this evening I found an article written by incredible DJ Drummond that gives a superb summary of the methods and openness of each of the major polling firms in this election. I have relied on him for much of this piece. I believe many of these results will surprise you. Allow me to take a few moments to summarize the major results from these two sources:
ABC News/Washington Post: They weight their sample for partisan identification based upon Voter News Service (VNS) exit polling data from 1992, 1996, 2000. They allow a +/- 3% variation for every poll on the statistic produced for Democrats, Republicans, Independents. Thus, on average they are working from a 35% GOP, 39% Democratic, 26% Independent sample. This means that their average sample will more often than not be skewed toward Democrats, if my guess about the ultimate partisan breakdown is correct. To their credit, they have a detailed explanation of their methodology. They also provide crosstabulations for voter choices across racial, income and education lines (just to name a few).
AP/Ipsos: Many of you political junkies will recall that this is the poll last week that called the race Kerry 49-46. It was the only one of its kind. Drummond notes: "Ipsos weights its poll, but does not detail the breakdown in its Press Release or Questionnaire, though some demographic information was released in their latest poll." This makes the Ipsos poll impossible to evaluate as a predictor of the ultimate vote result. How many Democrats, how many Republicans, how many Independents do they use per poll? It is impossible to say.
The Battleground Poll: This is the only poll which weights by party identification to correspond with what we know about this election's GOTV stories: 42.3% Democratic, 42.3% Republican, 15.4% Indepndent. While I think they undersample Independents, the idea of Republican/Democratic parity strikes me as extremely realistic (it would correspond to a net 4% increase in GOP turnout this year). Unfortunately, The Battleground Poll does not publish demographic breakdowns for its polls. However, The Tarrance Group (the Republican polling firm involved in this bipartisan firm) provided Drummond with detailed information on his request.
CBS News/ NY Times: To the extreme credit of this poll, detailed demographic breakdowns are readily available. However, and most important, is that Drummond notes that they "tend" to overweight their sample toward Democrats.
Gallup: This poll is truly the gold standard in presidential polling. They leave their partisan identification unweighted. This lends itself to swings in partisan identification, but it can enable it to pick up the shifting partisanship of the electorate. Gallup also has gobs and gobs of demographic statistics available to its subscribers. They have a demographic analysis that is perfectly in line with industry standards. They are the best poll in the business. Bar none. In particular, they have a good reputation for filtering unlikely voters out of their samples.
Harris: This poll is generally a blank. They weight their respondents to industry demographic standards, but they won't say anything else about their methods or their internal data. Drummond exhorts, "Ignore these guys, unless they start putting some hard data behind the headlines in their releases."
TIPP: This is another poll that remains mum. Go check out their website, tipponline.com, and try to find any methodological detail. I dare ya! No information about how they weight (if they weight) their respondents by demography or by partisan affiliation. No information about their method whatsoever. They just publish a poll every afternoon and expect you to trust them. I henceforth choose not to!
LA Times: The LA Times is equally mum about their internals. They claim to weight according to demographic characteristics, but leave it unclear the extent to which these weights match industry standards. They do not have a hard weight for partisan identification, but their internals almost always indicate that they give preference toward Democrats (as they did in their last 48-48 poll). They also, by the by, heavily (like 12%) skewed today's OH poll toward the Democrats.
Pew: Pew is a polling outfit that does an excellent job of providing you with its internals. If you go to the Pew site, your socks will be knocked off at their crosstabs (of course, given how deep the crosstabs go, they'll have a high MOE). However, Drummond notes that their partisan breakdowns skew toward Democrats.
Time: Time does not provide a great deal of internal data. However, Drummond reports that they weigh their polls precisely according to industry demographic standards. Their polling slightly skews toward the Democrats -- 35% to 34% (Independents 31%).
Zogby: Drummond seems to think that Zogby simply fell ass-backward into his present notoriety. Zogby predicted 1996 accurately and came close in 2000. However, his track record is generally lousy. Zogby has a devastating methodological problem -- his polls are not conducted randomly. He uses regional quotas. He is the only one of all these polls (that make any methodology known) that employs this technique. Furthermore, Zogby does not conform his data to industry standards. Rather, he uses -- among other sources -- his previous data. Neither of the standard-setters in the polling industry accept this practice. Zogby also weights his polls to strictly conform with the 2000 partisan turnout results -- a result that BC04 has spent tens of millions of dollars to change. Zogby, a supposedly independent pollster, also called the race for Kerry...IN THE SPRING! Zogby is also one of the chief practitioners of the internet poll, which is really quite unacceptable. Because I have (unfortunately) used Zogby here at this site, I will quote the very estimable Drummand at length (consider it part of my pennance): "All in all, Zogby’s habit of confusing his personal opinion with data-driven conclusions, his admitted practice of manipulating the respondent pool and his demographic weights, by standards not accepted anywhere else, along with mixing Internet polls with telephone interview results, forces me to reject his polls as unacceptable; they simply cannot be verified, and I strongly warn the reader that there is no established benchmark for the Zogby reports, even using previous Zogby polls, because he has changed his practices from his own history."
ARG: ARG generally does not provide you with methodological details, but their September poll was heavily skewed toward Democrats. 41.4% to 35.5%. Given their general Democratic tilt of their polls, it is likely that this practice is pervasive. It is also questionable whether their polls are actually random.
Rasmussen: You all know what I think about Rasmussen. I bring him up to note that Drummond's request for information about his methods went unfulfilled, and Scott Rasmussen does not provide detailed breakdowns of his internals or his methods on his website. To quote Drummond: "Not the size of the respondent poll, nor whether the sample is randomly developed, there is no weighting method cited, and there is no breakdown of respondents’ responses, which might allow analysts to compare Rasmussen’s results with anyone else."
So...what does this mean? Essentially, it means that the reliable polls are, roughly speaking, Time, Battleground, Gallup. All of the rest skew toward Democrats, and should be viewed with caution. A surprising number of the rest are either using questionable methods or are using unadvertised methods. This is absolutely taboo among social scientists. Methodology is the only element that the researcher can control, and thus it is critically important -- indeed it is an ethical responsibility -- for the researcher to make his/her methods avaiable.
The problem intensifies when we look at state polls. When we consider the number of different polling outfits, we are eventually forced to ask ourselves (doing our best impersonation of Jerry Seinfeld): "Who are these people?" Given how many national polling firms have methodological problems of large and small sizes, we should probably suspect that all those smaller firms have just as many problems as well.
I hope all of you will keep this information in mind as you see the polls spill out over the next five days. The best way to gauge this election is to look at the ground. The parties have good polling information (they are the ones solely concerned with getting an accurate portrait, as opposed to making news). This blog shall redouble its efforts to this purpose in the next five days.
How is that a defense of your false accusation against Cableguy?
Gallup Ping
FReepmail me if you want to be on or off the list.
Tonight's Gallup poll question with respect to congressional party preference shows a 4% swing from last week: From 50-47 in favor of Republicans to 48-47 Democrats (see table at USA Today website: http://www.usatoday.com). This indicates that Gallup almost certainly had at least 4% more Democrats in this week's sample. If you adjust the Presidential results for this difference, Bush has a 6 point lead, very similar to last week's results. The higher sampling of Dems would also explain the 5% decline in Pres. Bush's approval rating from last week.
The current Gallup poll throws out early voters. Once someone votes, they are no longer a likely voter.
Yikes....this isn't as good as I was expecting.
Turnout, turnout, turnout.
It is in line with all the other polls. It is 49/47 LV Bush without allocation. Besides, this is a weekend poll, and probably has a lot more Dems in the sample than not.
I take it as good news.
Okay look. What happened to everybody's theory of wait until the last weekend...they will have to start showing a bigger Bush lead for the sake of their integrity.
I understand the weekend phenom, but how far off can it be this close to the election. I would guess 2 more points in Dubya's favor, but yesterday I would ahve said Bush was 4-6 points ahead.
Today I think it is just 2 points.
That is not a very good lead for an incumbent.
Not true re exclusion of likely voters from the Gallup poll. See their footnote at the USA Today poll results page.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.