Newsweek poll sample may be skewed to overstate Pacific coast voters

[StJacques]

Many of you will be hearing the spin today about the Newsweek poll that suggests Kerry has erased Bush's lead. You can "pop up" the results of that poll here:

http://www.pollingreport.com/wh04gen.htm

and the web article discussing the results here:

http://msnbc.msn.com/id/6159637/site/newsweek/

I am going to present evidence that it is likely that there are problems with the sampling of the poll. If you navigate to the PollingReport.com page link above you will see the following quote included:

"Sept. 30 interviewing limited to Pacific and Mountain time zones, after the presidential debate concluded."

This poll was conducted over three nights, Sept. 30 - Oct. 2, and it polled 1,013 registered voters. We are not given any indication that the poll sample was regulated to make certain that the respondents were balanced out to produce a sampling representative of the overall American voting population. If the debate ended at 8:30 p.m. Mountain and 7:30 p.m. Pacific, that means that there were about thirty minutes to poll voters in the Mountain time zone and one hour and a half to poll voters in the Pacific time zone. It is a fair assumption that Princeton Survey Research Associates did not poll after 9:00 p.m. since they excluded the Central time zone from their first night's sample.

So the question arises, "did Princeton Survey Research Associates unfairly 'weight' the polling sample to include a number of respondents from the Mountain, and especially, the Pacific time zones out of proportion to their overall weight within the voting population"? It may be that 2/3 of their first night's polling results came from Pacific coast voters given that they had probably ninety minutes to poll their opinions compared with only thirty minutes for the Mountain time zone. Remember, the Pacific coast states of California, Oregon, and Washington account for seventy-three out of five hundred thirty-eight electoral votes, or 13.57% of all electoral votes, so if the sample is "representative" of the whole, the number of respondents would be close to that range. Since we are not given numbers as to the distribution of respondents across time zones we cannot say with certainty, but there are some clues within the internal numbers of the poll that suggest the answer is "yes." Allow me to continue.

From the MSNBC Newsweek article at the link posted above you can read the following quote:

"Among the three-quarters (74 percent) of registered voters who say they watched at least some of Thursday’s debate, 61 percent see Kerry as the clear winner, 19 percent pick Bush as the victor and 16 percent call it a draw."

Those results are not in line with other polling done on the performance of the two candidates in the debate. Take a look (you can "pop up" the results for yourself by clicking on the links):

Post-Debate Polls on "Who Won" the Debate"

Poll	Kerry Won	Bush Won	Other Opinion
Gallup	53	37	10
ABC News	45	36	19*
L.A. Times	54	15	31
Newsweek above	61	19	20

*ABC News "tie" figure is 17%, 2% are unaccounted for.

The Gallup, ABC News, and L.A. Times results are more or less within range of each other in analyzing public opinion on who won the debate. The Newsweek results go even farther in giving Kerry the advantage. Given the well-known strength of support Kerry has on the west coast, it appears to me that the post-debate "who won" results contained as an internal number within the Newsweek poll, coupled with the publicly-released information that the first night's polling only covered the Mountain and Pacific time zones, suggest that it is likely that Princeton Survey Research Associates has unfairly "weighted" the number of respondents from the Pacific coast states within their polling sample.

But there is more within the Newsweek poll's internal numbers that should raise eyebrows as to the accuracy of the final percentages of public support for the candidates and which, I will argue, is further evidence that the sample is unfairly weighted to respondents from the Pacific Coast. Take a look at the following internal numbers that analyze support for the candidates among Democrats, Republicans, and Independents:

Internal Numbers from Newsweek poll on Party Affiliation and Candidate Support

Party Affiliation	For Kerry	For Bush	For Nader	Other/Undecided
Democrats	86	12	0	2
Republicans	6	89	2	3
Independents	37	42	4	17

Now take a look at what happens if we do what MoveOn.org (don't shoot me folks, we'll use their own rules to show a "worst-case scenario" for the good guys and to demonstrate further the problems with the Newsweek polling sample) suggests pollsters should do in "weighting" their results to reflect a 35% Democrat, 35% Republican, 30% Independent voter turnout on election day.

Weighting Newsweek poll results at 35% Democrat, 35% Republican, 30% Independent voter turnout**

Candidate	Support from Democrats	Support from Republicans	Support from Independents
Bush	35 * .12 = 4.2%	35 * .89 = 31.15%	30 * .37 = 11.1%
Kerry	35 * .86 = 30.1%	35 * .06 = 2.1%	30 * .42 = 12.6%
Nader, Other, Undecided	35 * .02 = .07%	35 * .05 = 1.75%	30 * .21 = 6.3%

** Multiplier in "Nader, Other, Undecided" is sum of "For Nader" and "Other/Undecided" in previous table.

From there it's simple addition:

Computation of Newsweek poll results as a weighted average***

Candidate	Addition of Weighted Totals see above	Totals
Bush	4.2 + 31.15 + 11.1	46.45%
Kerry	30.1 + 2.1 + 12.6	44.8%
Nader, Other, Undecided	0.7 + 1.75 + 6.3	8.75%

*** The Newsweek poll covers only registered voters, not likely voters.

Based upon the evidence I cite above, I believe we have a sound basis upon which to conclude that the Newsweek poll was rushed to make the Sunday news and that it is likely that its results unfairly weight respondents from the Pacific coast, given the information included with the poll that interviews conducted on September 30 only included the Mountain and Pacific time zones coupled with a common-sense estimation of the attitudes towards the candidates expressed in the internal numbers released with the poll. I could go on about those internal numbers at greater length, and anyone who responds within this thread is welcome to comment further, but this poll is simply flawed.

[Hildy]

Wow...they've only got a month to make sure Nader isn't in the race, if you know what I mean.

[MNJohnnie]

http://www.rasmussenreports.com/Presidential_Tracking_Poll.htm

What narrowing gap???? The Rassmussen Daily Tracking poll shows us exactly .1 point from where we were Thursday. But OH that's right don't confuse them with reality. The Media template demands a Horse race so just throw all logic to the winds and buy into the media lie.

[StJacques]

I also looked at Rasmussen Johnnie. Point well made!

[Brett66]

I was wondering what the weighted percentages would be with an equal number of R's & d's. Thanks for the info, this is more believable.

[seppel]

it's no use. instead of hoping for not-so-bad numbers, rep's should just fight harder.
unfortunately flakes like kerry have been winning elections from good people for decades, whether in the US or Europe or other nations. that's because many voters are shallow and impressionable. most act in herds, voting whatever the main baffoon on the group thinks.
here in germany schroeder and his fellow atheist socialists have been fooling the populace for 2 elections now. apart from gay marriage, non-security, economic decline and moral filth, we now face being sucked up by Turkey. only so he can lock the votes of the 5 million Turks living here.
he's prepared to give up the sovereignty of the country to have his power and glory. just like kerry.

[MNJohnnie]

http://www.rasmussenreports.com/Presidential_Tracking_Poll.htm

What narrowing gap???? The Rassmussen Daily Tracking poll shows us exactly .1 point from where we were Thursday. But OH that's right don't confuse them with reality. The Media template demands a Horse race so just throw all logic to the winds and buy into the media lie.

[MNJohnnie]

http://www.rasmussenreports.com/Presidential_Tracking_Poll.htm

What narrowing gap???? The Rassmussen Daily Tracking poll shows us exactly .1 point from where we were Thursday. But OH that's right don't confuse them with reality. The Media template demands a Horse race so just throw all logic to the winds and buy into the media lie.

[pcx99]

Absolutely ZERO surprise here. Newsweek is one of the most biased publications (and websites) ANYWHERE.

[Congressman Billybob]

Brilliant piece of work. I will pass your analysis along.

Billybob

[stockpirate]

bump for later

[NickFlooding]

Conventional MSM wisdom is that conservatives traditionally poll under the radar, ie; we lie to the pollsters and say we will vote for the liberals. I would like to suggest that the MSM skews the polls so as to appear that the liberals are doing better than they really are.

I further suggest "W" never really had a double digit lead and it was an intentional MSM fabrication for the singular purpose of creating a "Kerry Bounce" from the first debate.

When it is all said and done the only poll that will count is election day. On that day, "W" will get a narrow plurality vote and a substantial Electoral College Vote.

[StJacques]

Check your e-mail Billybob. I got your message - thanks - and I sent you some information.

[StJacques]

I have received a couple of e-mails from others who have read this thread who have told me to look for additional responses, after they run down some information. I'm going to have to log off for a while but I'll return after about 10:00 p.m. Central time.

[StJacques]

It appears I made a small mistake in one of the tables I displayed, on the "Internal Numbers from Newsweek poll on Party Affiliation and Candidate Support." I reversed the cell entries for Bush and Kerry's support among Independents.

This is the way the graphic should read:

Internal Numbers from Newsweek poll on Party Affiliation and Candidate Support

Party Affiliation	For Kerry	For Bush	For Nader	Other/Undecided
Democrats	86	12	0	2
Republicans	6	89	2	3
Independents	42	37	4	17

I would like to point out for everyone's benefit that all of my calculations are correct. If you will look at those calcuations you will see that I have the support figures in the right place.

Sorry for the glitch.