Why did Newsweek's poll interview 374 Republicans & only 303 Democrats?

[Karmadillo]

Aren't there more voters who are registered Democrats? This link quotes USA Today as indicating there are 72 million registered Democrats and 55 million registered Republicans. I know nothing about polling and assume there's some juggling to approximate likely voter turnouts, but it would still seem any poll that polled more Republicans than Democrats would be likely to end up with a favorable result for the Republican. If I'm being overly stupid, I wish someone would enlighten me. Thanks.

http://biz.yahoo.com/prnews/040904/nysa058_1.html

SAMPLE SIZE/MARGIN OF ERROR FOR REGISTERED VOTERS SUBGROUPS:
1,008 Registered voters (plus or minus 4)

505 Thursday interviews (plus or minus 5)
503 Friday interviews (plus or minus 5)

374 Republicans (plus or minus 6)
303 Democrats (plus or minus 6)
300 Independents (plus or minus 6)

[okieinpain]

1. you beat me to it, I was looking at the sample, why are dems

on the low end, especially after the 2000 election, when did repugs become the majority of the country. you would think they would try and get an even sample.

[Zynx]

2. In reality, Democrats outnumber Republicans by 4-5%

It should be about 39% Democrats to 34% Republicans, and 27% Indies. This is a bullshit poll.

[glitch]

3. why indeed.

I really want to know too. And I don't want some bogus because Rs are more inclined to vote than Ds response either, because that rationale is defunct this election cycle.

[LoZoccolo]

9. It might be random sampling, which has it's problems.

I know that Gallup is random. The problems are that the demographics of who answers the phone and who doesn't can skew the poll, and a random poll taken during the RNC could be expected somewhat to include more Republicans because they might be more likely to be at home to take the poll (watching the convention) than Democrats. They might insist on a random sample because they haven't come up with a methodology to overcome it's limitations yet, or might not want to - it would force them to decide just how to compensate for the skew of a random poll. Rassmussen and Zogby have some methods of overcoming the skew, but then again, that's a created process that might distort the findings as well. Polling is an art, really, and the methodology can affect things without there being any "conspiracy" to skew the findings.

[LoZoccolo]

4. First decent criticism of the /Newsweek/ poll that I've seen.

Much better to do some homework than just rejecting all data that doesn't fit with how you want things.

[Gman]

5. Good. Let 'em believe their skewed numbers

we'll laugh all the way to the White House.

[A_Possum]

6. The Time poll is the same

PLUS they asked FIRST for the "youngest male in the household" to talk to. And then if there was no male, they asked to talk to the "oldest female."

This is supposed to make it random, for some reason that remains unfathomable to me.

GIGO, sounds like to me.

[progressivebydesign]

15. That's really odd..

.. I'm not sure how they think that makes it random.

[patdem]

7. That is 23% more republicans than Democrats??? WHY????

and he only got a 10 point bounce? What was their methodology BEFORE the convention? has it changed???

[troublemaker]

27. Republicans are MORE LIKELY TO BE HOME DURING THE REPUBLICAN CONVENTION

And Democrats tend to be at home during the Dem convention. That's why it used to be understood that nobody polls during a convention... party loyalists are all at home to answer the phone. There were no major polls taken during the Dem convention. They were taken right after it.

[LibDemAlways]

8. How else could they show the chimp with

a significant lead? They have to pick and choose who they "poll" pretty carefully. This is all bullshit designed to demoralize Kerry supporters.

[MrBenchley]

10. Because they wanted to give Junior Birdman a bounce

whether or not he had one, obviously.

[WiseMen]

11. Friends, ALL modern polls use weighting to arrive at statistically

ballanced result. Precice initial sub-sample numbers are not relevant unless the sub-sample size is too small to support probabilities.

[TheStateChief]

12. Exactly...I agree...but I can't find the weighting...can you?

Not disagreeing with you at all but I haven't seen any source say that the weighting that Newsweek used was different from the raw numbers. If they didn't do any weighting to the raw numbers this poll is completely worthless.

[WiseMen]

16. I agree. But it is standard analytic proceedure to plug in the weights

[Zynx]

18. There is little evidence that they did. If they had, Bush would have only

had a six point lead.

[patdem]

30. Has gobbledegook served you all your life???

or is this a new profession for you????

[Zynx]

13. Give me proof that they re-weighted it.

I showed what the results would be if they did it properly in my thread and they show Bush with a relatively minor six-point lead.

[Frodo]

14. Newsweek doesn't balance by party affiliation. Nor does the LA Times.

When that resulted in a big Kerry lead it was not considered a problem here - it was evidence that the electorate was becoming more Democratic.

Now it's a problem.


In reality, it was a problem all along. But the news magazines (Time & Newsweek) want movement in the race. I think they have exagerated leads in both directions. We'll know better over the next week or so.

[DemMother]

19. It's probably also why they poll in the middle of a convention

instead of waiting a few days to let things settle--it probably contributes to wider swings. It's also interesting that it's the two big newsmagazines who did this. I wonder if pollsters have historically sampled voters on the final two days of the conventions.

[progressivebydesign]

17. Independents skew Republican.

Rarely will you find ex-Democrats as Independents. It happens. Usually, it is ex-republicans, OFTEN because they feel the republicans aren't conservative enough. It's a mindset common to middle aged white men. Perot was the darling of independents.. these are his leftovers. You tell me how any Perot guy would support Kerry over Bush.

[gohawks]

20. Newsweek: 42% polled were military households

... How can that be?

417 Military households (plus or minus 6)
578 Non-military households (plus or minus 5)

are 42% of families in this country military families? Think NOT.

and military families went big for Shrub in the poll 58%/36%.

[gatorboy]

22. That segment makes the polling very fishy to me....

Considering Only 30% of households have at least one member on active duty or a veteran, they were very "lucky" to hit as many military homes as they did....

[Karmadillo]

21. Does anyone know where one can get the internals for other Newsweek polls?

I checked their site under "pressroom" and found press releases like the one in the original post, but the polls for the Democratic Convention and the early July poll don't have any internals following the release. Thanks.

[gohawks]

24. Newsweek Internals.... at

http://biz.yahoo.com/prnews/040904/nysa058_1.html

[Karmadillo]

31. Thanks, but I can only find the internals for the most recent poll there.

I was just curious as to the D/R split in previous polls.

[betterdeadthanred]

23. Which is exactly the reason people need to chill

The polls, if everyone will recall, were off badly the last election. This election is gonna be based on turnout, period.

[WI_DEM]

25. this is another skewed poll--indeed why interview 71 more Reps

than dems?

[kentuck]

26. If you subtract 71 Repubs from the poll, it subtracts 7%...

Meaning the lead would be 4% rather than 11% - which would sound more logical...unless they have a special reason to put extra Repubs in their poll? But being so quickly done, I doubt they had the time to do that much analysis...

[troublemaker]

28. .

[troublemaker]

29. Their defense is that a purely random sampling is the only

Their defense is that a purely random sampling is the only way to catch swings in mood. For instance, If everyone watched the R convention and loved it so much they decided they want to be Republicans that fact would be lost if you normalize the sample based on previous party affiliation polls. So random isn't all bad.

But you don't do a random poll during a fucking convention. That's basic. You can poll during a convention if you use a method that adjusts the sample, but not if you poll randomly.

Both of these polls include Thursday polling. People were called at random on Thursday evening... the evening Bush was to speak. Who makes a point to be at home when Bush speaks? Republicans. That's why this random sample has a surplus of Republicans... it like conducting a poll on Sunday evening asking people if they like The Simpsons. Every person who loves the Simpsons so much they stay home to watch it is at home on Sunday evenings.

QE-motherfucking-D