Newsweek Poll Undersamples Republicans by 10 Points: Only 27% Republican (Election 2004 - 37% GOP)

PR Newswire via Yahoo ^ | November 12, 2005

[new yorker 77]

LAST QUESTION in the polling data:

D2. Now I have just a few more questions so we can describe the people who took part in our survey ... Regardless of how you might have voted in recent elections, in politics TODAY, do you consider yourself a Republican, Democrat, or Independent?

27 Republican

37 Democrat

32 Independent

2 No party/Not interested in politics (VOL.)

1 Other party (VOL.)

1 Don't know

100

[DrDeb]

"But race, age and gender don't change overnight. Political affiliation can and does"


Based on my experience with polls/statistical analysis, it's my hypothesis that party affiliation correlates as significantly (if not more significantly) with a respondent's perception of this President's JA than any other demographic variable.

[I encourage you to study closely the demographic composition of most MSM polls. Even though pollsters claim they 'weight' for demographic consistency, their 'samples' typically feature a higher percentage of Democrat constituency groups than the general population, e.g., singles, women, minorities, adults under the age of 30, and the unemployed . . . Surprise, surprise, these polls also feature a MUCH higher percentage of Democrats/Democrat-leaning Independents than one would expect given the composition of the electorate in 2000, 2002, and 2004.

If pollsters refuse to weight for party affiliation, then they MUST weight ACCURATELY other demographic variables . . . Quite frankly, it would be easier (and more valid) to simply weight for party affiliation.]

[DrDeb]

"They control for variables that can compared against the census, this cannot be done in the case of party ID, all we have to go from is flawed exit polls."


We don't have to rely on exit polls; we could use as our benchmark the 'party' composition of the House of Representatives -- which, by the way, would be infinitely more valid that using our FLAWED (and dated) census data!!

BTW: Don't you think it odd that Republicans have won the last three elections and hold the House by a ratio of 53R:46D:1I* and yet it's the Democrats who are consistently oversampled in these polls by 8-15 points?!

*BTW: Without the gross voter fraud in EVERY major urban area, the President would have gleaned 53% of the popular vote as well in the last election!!

[Hattie]

Nowadays, people may be reluctant to admit that they're Republicans.

And I guess they'd be proud as punch to admit to being a Democrat?

[Alter Kaker]

We don't have to rely on exit polls; we could use as our benchmark the 'party' composition of the House of Representatives -- which, by the way, would be infinitely more valid that using our FLAWED (and dated) census data!!

That's a ridiculous idea. If pollsters had been using that method, then Adlai Stevenson, Hubert Humphrey, George McGovern, Walter Mondale, Michael Dukakis and Bob Dole would all have been elected President.

[Alter Kaker]

*At least according to the polls.

[DrDeb]

Excuse me AK, but using 'House' party ratios to weight current polling samples makes as much sense as the methodology currently employed.

[BTW: Are you arguing that polling samples skewed toward Democrats would INCORRECTLY predict the outcome of a presidential election (by making it more likely that such a sample would automatically predict the election of a Democrat)?! Hmmmmmmmmmm, as I recall, this exact scenario happened quite frequently last year, and interestingly enough, is still happening today with -- you guessed it -- the President's current JA rating!]

[jmaroneps37]

ratmedia droppings