Posted on 07/31/2004 11:52:27 AM PDT by demlosers
No it wouldn't. My problem is with polling itself, and its pretensions to some sort of science, i.e. prediction. Pols aren't buying predictions from the polling industry. They are boosting the industry in order to buy influence, or the attempt to inflence the election. The only polls that might conceivably work are push-polls in their effect on certain impressionable sorts. At least they seem to believe these work. If they didn't think so, I don't think there'd be much of polling industry.
In order to guess by sample of a population you need a consistent behavior among each 'sample point'. Balls in a cylinder may be. Molecultes may be. People aren't.
In order to determine what a 'snapshot' of the vote might be at any point in time, you'd have to know the vote, exactly. Trying to predict that ahead of time, is fruitless. There are too many variables. People are not molecules in a cylinder.
But you might argue, why are only certain states toos-ups, 'battlegrounds' so-called? I would suggest it's because the other states are believed to be predominantly for one campaign, or the other, so much so that you don't even need a poll to tell you that. Frankly, myself, I'm not so convinced even of that. I suspect CA is still 'in play', as they phrase it. I think Bush could win California, with or without Arnold's help. The 'power' in the CA has always been LA county. But the election of Arnold showed a lot of turn-out in the 'green' counties, as they colored them.
Maybe. We will see .Ok.
Pray for W and Our Awesome Troops
Shoot, the Rasmussen poll had Bush beating Gore by 6 or 7.
Zogby had it 49-49. He was closest in 2000.
I waant a complete landslide! I want Dubya to take all of New England and New York. I think he has a chance.
"Ordinarily, you would say that, even though Newsweek polls are garbage, you could compare one to the other. But Newsweek polls are such junk, that you can't even do that. Still, they must have wanted more. Much much more."
Newsweek is usually skewed, so let's compare the latest two Newsweek polls, poster AmishDude. Finally enough Newsweek articles and polls have been published to analyze the latest Newsweek July 29-30, 2004 post-convention (Democrat) three-way poll in a similar fashion as was done previously to the Newsweek July 8-9, 2004 presidential three-way poll. Pardon the length, but the excessive documentation is necessary in case the primary data disappears or changes on the web. Bear with me...
"July 31 - Coming out of the Democratic National Convention in Boston, Sen. John Kerry now holds a seven-point lead over President George W. Bush in a three-way race with independent Ralph Nader, according to the latest NEWSWEEK poll. Three weeks ago, Kerry’s lead was three points."
"Half of the poll was conducted on Thursday night, before Kerry had delivered his acceptance speech in Boston. On Thursday, Kerry had a two-point lead over Bush (47 percent to 45 percent in a three-way race with Nader). In Friday night polling, his lead over Bush grew to 10 points (50 percent to 40 percent)."
"In an election expected to be decided by a small number of unaffiliated voters, independents now lean toward Kerry by a margin of 45 percent to 39 percent, with Nader pulling 7 percent."
"For the NEWSWEEK poll, Princeton Survey Research Associates interviewed 1,010 adults aged 18 and older July 29 and July 30 by telephone. The margin of error is plus or minus 3 percentage points."
Source: MSNBC/Newsweek - "A Baby Bounce"
Hard to accurately measure the complete convention 'bounce' factor when the half the Newsweek poll is taken during the day just before Kerry's prime-time acceptance speech Thursday night, and the other half of their polling was performed on Friday, the day after...
From another source - the PR Newswire, there is available even more of the Newsweek article with more polling data that was not available in the above quoted MSNBC/Newsweek web article above.
"In a three-way race with the Ralph Nader/Peter Camejo ticket added, Kerry/Edwards receives 49 percent of the vote; Bush/Cheney, 42 percent and Nader/Camejo, 3 percent, the poll shows."
"Looking at crossover voters from the 2000 election, 92 percent of Gore voters in 2000 support Kerry (5 percent say they will vote for Bush and 3 percent is undecided); 84 percent of Bush voters say they plan to vote for the president again (four percent of Bush 2000 voters are undecided, 10 percent say they will vote for Kerry and 2 percent for Nader)."
"For this Newsweek Poll, Princeton Survey Research Associates International interviewed 1,190 adults aged 18 and older on July 29-30, 2004. The margin of error is plus or minus 3 percentage points. This poll is part of the August 9 issue of Newsweek (on newsstands Monday, August 2)."
Source: PR Newswire - Newsweek Poll: Democratic Convention 2004
Even later, we now have available from yet another source, 'PollingReport.com' what appears to be all of the internals to the Newsweek/PSRAI 3-way presidential poll question of July 29-30, 2004. The other referenced MSNBC/Newsweek web articles exposed only portions of all the internals released by PollingReport.com today. What is a bit curious is that the poll methodology published in the two Newsweek articles say that the just released July 29-30, 2004 poll was based upon 1,010 'Adults' with an MoE of ±3%, yet the more complete version of the poll published by 'PollingReport.com' says 1,010 'Registered Voters' with a MoE of ±4%.
Here is the Newsweek/PSRAI poll data of July 29-30, 2004:
|
Newsweek Poll conducted by Princeton Survey Research Associates International. July 29-30, 2004. N=1,010 registered voters nationwide. MoE ± 4 (for all registered voters). |
||||||
|
. |
||||||
"Suppose the election were being held TODAY and you had the following three choices for president and vice president: George W. Bush and Dick Cheney, the Republicans; John Kerry and John Edwards, the Democrats; and Ralph Nader and Peter Camejo, the independent or third party candidates. [Choices rotated.] Who would you vote for?" If none of these: "As of TODAY, do you LEAN more toward Bush and Cheney, the Republicans; Kerry and Edwards, the Democrats; or Nader and Camejo, the third party candidates?" |
||||||
|
. |
||||||
| Bush/ Cheney |
Kerry/ Edwards |
Nader/ Camejo |
Other (vol.)/ Undecided |
|||
| % | % | % | % | |||
| ALL | 42 | 49 | 3 | 6 | ||
| Republicans | 90 | 7 | 1 | 2 | ||
| Democrats | 8 | 86 | 1 | 5 | ||
| Independents | 39 | 45 | 7 | 9 | ||
| Men | 48 | 43 | 3 | 6 | ||
| Women | 37 | 53 | 3 | 7 | ||
| 18-29 | 32 | 51 | 9 | 8 | ||
| 30-49 | 47 | 46 | 2 | 5 | ||
| 50 & older | 44 | 49 | 1 | 6 | ||
| Southern white | 54 | 38 | 3 | 5 | ||
| Non-South, white | 47 | 46 | 2 | 5 | ||
| Non-white | 20 | 65 | 5 | 10 | ||
| Urban | 30 | 63 | 2 | 5 | ||
| Suburban | 47 | 45 | 2 | 6 | ||
| Rural | 49 | 37 | 4 | 10 | ||
| 7/30 only | 40 | 50 | 3 | 7 | ||
| 7/29 only | 45 | 47 | 2 | 6 | ||
|
. |
||||||
| Trend: | ||||||
| 7/8-9/04 | 44 | 47 | 3 | 6 | ||
|
. |
||||||
Source: PollingReport.com, White House 2004: General Election Newsweek/PSRAI - July 29-30, 2004. N=1,010 Registered voters.
Based on a potential 2004 US voting age population of approximately 216,071,664 (18 years or older), of which around 92% are citizens or about 198,785,930 potential voters, which when calculated with a poll sample size of 1,010 Registered voters equals a Margin of Error of ±3.08% (at 95% confidence level), and is essentially what was listed by the two Newsweek web-published articles. However, if the Margin of Error of ±4% listed by 'PollingReport.com' is correct, then the presidential preference "Registered Voter" sample might possibly be just be a 'Registered Voter' subset of the Newsweek 1,010 poll respondents, which then would only have to be 600 'Registered Voters' who were polled on the presidential preference question. Most curious... Anyway, going with the 1,010 sample as 'Registered Voters' makes the data analysis easier and consistent with the previous Newsweek/PSRAI poll of July 8-9, 2004 with a stated sample size of 1,001 Registered voters.
Source: US Census QuickFacts
Source: US Census, PDF file, February 2002 Voting and Registration in the Election of November 2000, PDF page 5/16.
Source: American Research Group MOE calculator
The next table displayed directly below is the above Newsweek/PSRAI referenced internal poll data for the three-way presidential preference question displayed in spreadsheet format for the party affiliation breakdown analysis of the July 29-30, 2004 poll:
The table shown directly below is my distribution of the raw responses for the 1,010 Registered voters of the Newsweek/PSRAI three-way presidential poll question necessary to re-create the published poll results from the July 29-30, 2004 poll shown in the table directly above. The 1,010 Registered voter party affiliation breakdowns represent the weighted poll responses that were used by Newsweek/PSRAI and is detailed below. Notice that the analysis table shown directly below includes both a "raw vote" distribution with the mathematically derived final poll results and the same data presented in a tabular percentage format directly below the "raw vote" distribution. If the percentages in the table show below are rounded off so as not to have a fractional component, the results exactly match the Newsweek published poll results for the July 29-30, 2004 poll. This means that the Newsweek/PSRAI weighted party affiliation sample distribution must be around the 28.91% (R), 37.13% (D), and 33.96% (I) derived and shown in this table.
Newsweek/PSRAI Poll
July 29-30, 2004
Distributed 1,010 Responses
Registered Voters
Now as previously referenced in my message #198, another thread, which itself referenced the Newsweek/PSRAI July 8-9, 2004 polling data included in FR poster VRWC-minion's message #49, another thread of this thread, the following table denotes the Newsweek/PSRAI final poll results and weightings for the previous July 8-9, 2004 poll and the just released July 29-30, 2004 poll detailed above.
Please note that the two Newsweek/PSRAI presidential three-way polls were separated in time by ONLY twenty days. In the table below, note that the post-convention 'bounce' was only around 1.56% (rounded to 2% by Newsweek for breathlessly urgent headline purposes) from the Newsweek poll of twenty days earlier. What was the major factor accounting for the 1.56% convention 'bounce'? If you look carefully in the data table below, one will see that the Newsweek/PSRAI statisticians drastically changed the July 29-30 poll sample weightings from those of the July 8-9 poll by lowering the percentage of 'Independents' by a little over 10%, and increasing the percentage of Democrats by almost the same amount. Now your average poll consumer is supposed to believe Newsweek when their sample universe changes so rapidly in only twenty days. Just for fun, I used the poll sample weightings from the July 8-9, 2004 Newsweek/PSRAI poll to re-weight the poll data from the July 29-30, 2004 Newsweek/PSRAI poll to see how the revised poll results would differ.
Gee, what a surprise! My re-weighting results in a negative 2.26% 'bounce' (crater?) for Kerry instead of the Newsweek published positive 2% 'bounce' over Bush. In addition, the revised poll data shows a dead-even race using the overly Independent-skewed July 8-9 weightings. Oh well, live by the sword, die by the sword...
Finally, reproduced below are the year 2000 Presidential election popular vote data combined with the November 7th, 2000 election day Voter News Service (VNS) exit poll data (for party affiliation) to get the partisan breakdown of the US electorate. Of course this is almost four year old pre-9/11 data, so use in future electoral predictions/calculations at your own risk...
Source: New York Times, November 12, 2000, section 4, p. 4. "Who Voted", Voter News Service Data, 2000 presidential election.
Secondary Source: Kellogg Community College, Jo Williams, Lecture Notes Chapter 11 Voter News Service Data, 2000 presidential election.
Source: Final Election 2000 data - US Election Atlas, Dave Leip, 2000 Presidential Election Results
Hope this helps...
dvwjr
Thanks for the excellent post.
I saw Kerry on FNS this morning, and he was being a bit defensive about the Newsweek results, saying it doesn't show more of a bounce because it wasn't taken after his speech.
Helluva lot of work you put into that post. It is appreciated. Thanks.
I have to take issue with your derived voter distribution from the July 8-9 poll. I find it inconceivable that Newsweek would undersample Democrats to such an extent (10-12% below their population share) in their poll. I know you set up 3 independent equations from the poll numbers to solve for R,D and I, but these raw numbers from July 8-9 don't make sense to me.
WOW, thank you so much for all that work, and yes it does help a LOT! Boy are you going to be valuable to have here over the coming months :o)
Pray for W and Our Troops
But again, my problem is still with the polls, themselves. People are not molecules in a jar. They change their minds. They can, many of them, behave in unpredictable ways. Some may not even show up to vote. You can't apply the assumptions of sampling when the assumptions are so violated.
It might be tempting, for example, to look at the Dem/Rep breakdown and say, for all the nonsense otherwise reported, that might actually be useful, at least as a sampled 'snapshot' of the population. But the same problem is found. You still can't really know how each would vote, that day, or who would show up.
Again, the science isn't in polling, but in certain historical trends and knowledge of people at the precinct level, even by neighborhood or block. If one wants prediction, that's where the campaigns go, as they pay lip service to polling.
And from another message, I'd add that the national polls themselves, even if they could be judged as predictive (say by comparing national results on election day to polls taken a few weeks or so before, for ex), which they aren't, still misses the point. The election isn't decided by a nationwide popular vote. Our Founders created the electoral college. And the states can award those as winner take all. As phony as state polls, themselves, would be, particularly if there's the sense that Dem and Rep are fairly evenly divided based on neighborhood understanding as even reflected in districting, if one believed in polling, at all, it's the state polls that one would examine.
As much as I love statistics, it suffices to say that "Newsweek polls are garbage. Ignore them."
I think the Kerry campaign lost about 1-2% as a result of his speech, and Edwards sort of fumbling, as well. But the Dem didn't tune in to watch Republican-lite. They wanted Sharpton as their nominee. The Dem don't wrap themselves in the flag. They bitch about everything and burn the flag. And I think they are confused by what they saw. And that support is waning, very slightly. I think with the GOP solid support for Bush, that too could be harmed by a Rep convention that tries to pretend they are a bunch of Democrats. Karl Rove is no rocket-scientist. This is the Bush campaign's to lose. And Rove is just the guy who could lose it.
As for polls themselves: "One thousand people can represent the responses you would get if you interviewed 100 million. Elections show that." [LINK ] But I'd sure like to see his data on that. I surely would.
"I have to take issue with your derived voter distribution from the July 8-9 poll. I find it inconceivable that Newsweek would undersample Democrats to such an extent (10-12% below their population share) in their poll. I know you set up 3 independent equations from the poll numbers to solve for R,D and I, but these raw numbers from July 8-9 don't make sense to me."
As hard as it is to believe, that unusual distribution is what the Newsweek July 8-9, 2004 poll data turned out to support. I cannot find another solution set, via linear equations or by modifying a spreadsheet by hand that will match all the published internals and overall poll results. I performed the same analysis on the past four Newsweek polls which were available with internal party affiliation breakdowns. The four polls are listed directly below with their imputed (Rep/Dem/Ind) breakdowns, and overall poll results. The actual Newsweek/PSARI polls were rounded off with no fractional component, my final poll percentages are listed with two decimal places, a side effect of the calculations from the hand-distribution of discrete 'votes' in an analysis spreadsheet. When these final poll results are rounded so as to have no fractional component they match the published Newsweek numbers.
Source: PollingReport.com - White House 2004: General Election Newsweek/PSRAI polls.
Poster Dales thinks that Newsweek/PSRAI does no stratification on the political affiliation component of the poll, hence the variations due to different circumstances when each poll is conducted. The raw data is all there, if you find another solution set please let me know so that I might add whatever I overlooked to my "bag of tricks".
Best of luck...
dvwjr
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.