Bush recent approval at 28 percent - Page 6

DuH2 · #51 (**permalink**) 05-09-2007, 01:56 AM

..and yet he is capable of saying what he wants to do in clear English..while your crowd is not.

He is able to veto and sustain a veto..your crowd can to beat it.

Your crowd cannot even maange to drop the idiotic word games and put forward a stand alone bill callign on the USA to run out on Iraq?

Why is that?...

Only two plausable possibilities.

1. Most Amreicans think yor solutions are INSANE so you r afraid to express or support them penly and without ambiguity and catchy wordage.

2. Your popularity as a group is so low that the Nazis out poll you amongst Jews.

Reid can't even say what he wants to say without having to rush out and muddle up what he meant with BS becasue of the outrage the American people hold towards his defeatist loser policies.

Moore is a running joke...etc.
..and past all that as Liberty pointed out this is pretty much the lowest # you guys can find....and not being supported by other polls.

...

Its no shock its highlighted by MSNBC. MSNBC is prone to highlighting only the most negative polls they can find abotu Bush or the War(s)..they always have with Bush especially.
Probably another reason why their ratings suck so bad...people don't trust them.

jsh1120 · #52 (**permalink**) 05-09-2007, 05:35 AM

Quote:

Originally Posted by Liberty";p=&quot

Quote:

Originally Posted by jsh1120";p=&quot

Rasmussen's sampling methodology and robo-calling appears to oversample people who are more likely to be reached on the first call, "stay at home" types who are somewhat more likely to vote Republican. This, combined with fewer "call backs" by Rasmussen compared to other organizations tends to bias the results slightly toward conservative and Republican voters.

Rasmussen has been weighting polls by party affiliation at least since the 2004 election, so any oversampling through the mechanism you've described above is not likely to result in significant bias. More relevant are the party ID assumptions -- and Rasmussen has differed somewhat from other polling agencies. In 2004, he assumed a greater proportion of Republican voters than other pollsters. (37R/37D/26I vs. 35/39/26)

His calling procedure also tends to indicate more partisanship (greater proportions of R/D voters, fewer independents). Some of that may have to do with the "robo-calling" -- maybe people more likely to admit partisanship to a robot, or perhaps partisan voters are more likely to respond to an automated system.

Rasmussen did about as well as expected in 2004, but was criticized for using the same (37R/37D/26I) assumption in the months after the election, therefore not accounting for shifts in party affiliation. I believe he changed that assumption last year, accounting for party ID by using sort of 3-month rolling average.

This is pretty "inside baseball" so perhaps it deserves a thread somewhere else, but for the record:

I'd be interested in knowing where your information about the "assumptions" of party identification comes from. Frankly, the notion that Rasmussen used one assumption and "other pollsters" used another (37R/37D/26I vs. 35/39/26) seems suspect to me. Are you contending that Rasmussen and "other pollsters" weight their results by assumptions about Party Identification? And if you are, are you saying that all "other pollsters" used the same weighting?

If, in fact, Rasmussen is using party identification as a weighting variable for sampling (at all) their methodology is especially suspect. Weighting is usually reserved to correcting sample bias in traits that do not change as a result of a change in a person's opinion (e.g. sex, age, geographic location) NOT in terms of a self-described opinion such as party identification. The latter is a DEPENDENT variable, not a legitimate weighting (i.e. independent) variable.

In other words if I "assume" the "true" distribution of party ID is (80D/10R/10I) and weight my results accordingly, I can guarantee a Democratic majority in terms of voter intentions. Such an "assumption" is not legitimate and the "true" distribution of party ID cannot even be evaluated to weight a sample. That is especially true on the eve of an election where party ID fluctuates somewhat to match voter intentions.

I think what you might have meant was that Rasmussen's RESULTS (not weights) were more Republican than the average of other polls. I strongly suspect that is true as I noted earlier.

As noted in an earlier post, Rasmussen's Republican tilt was especially evident in the 2006 elections. Shortly after that election I compared Rasmussen's results in individual Senate elections to actual election results. In almost every case Rasmussen overestimated the Republican share of the vote. If memory serves, in some cases this led to incorrect calls, (e.g. I believe Rasmussen called Missouri for Talent rather than McCaskill.) In other cases, when the Republican won, Rasmussen overestimated the margin. (e.g. Kyl in Arizona.)

All in all, if you're correct that Rasmussen overestimates partisans (both R's and D's) versus independents, it suggests a more deep seated problem with their methodology in predicting election results, especially as turnout rises. (i.e. Partisans are more likely to vote. Increased turnout always indicates a higher than average turnout of independents.)

The most problematic aspect of predicting election results is not getting Party ID right, it's predicting who will actually vote. Rasmussen's problem appears to be that they are prone to underestimating (a) the proportion of the electorate that is independent; (b) (at least in 2006) the likelihood that independents will actually vote; and (c) (again in 2006) the R vs. D split of independents who do vote.

Finally, as I noted before, comparisons of results from different polling organizations and predictions of "winning" are both problematic. More reliable conclusions come about from tracking results of the same organization over time, canceling the effect of methodological differences. Using that approach, Rasmussen's over time results match those of other organizations: a steep drop in support for President Bush over the last two plus years.

jsh1120 · #53 (**permalink**) 05-09-2007, 06:19 AM

Quote:

Originally Posted by DuH2";p=&quot

..and yet he is capable of saying what he wants to do in clear English..while your crowd is not.

He is able to veto and sustain a veto..your crowd can to beat it.

Your crowd cannot even maange to drop the idiotic word games and put forward a stand alone bill callign on the USA to run out on Iraq?

Why is that?...

Only two plausable possibilities.

1. Most Amreicans think yor solutions are INSANE so you r afraid to express or support them penly and without ambiguity and catchy wordage.

2. Your popularity as a group is so low that the Nazis out poll you amongst Jews.

Reid can't even say what he wants to say without having to rush out and muddle up what he meant with BS becasue of the outrage the American people hold towards his defeatist loser policies.

Moore is a running joke...etc.
..and past all that as Liberty pointed out this is pretty much the lowest # you guys can find....and not being supported by other polls.

...

Its no shock its highlighted by MSNBC. MSNBC is prone to highlighting only the most negative polls they can find abotu Bush or the War(s)..they always have with Bush especially.
Probably another reason why their ratings suck so bad...people don't trust them.

Godwin's Law at work again...Congratulations on being the first to bring up "Nazis" on this thread.

http://www.kuro5hin.org/story/2004/6/30/33339/3949

Somehow I doubt that when MSNBC was reporting Newsweek polls in 2003 that showed Bush's approval at 60% you were complaining.

I suspect that the 28% support for Bush is a bit low. So if you'd rather stick with a two to one majority against the President, that's fine.

Meanwhile, the latest CNN/Gallup results show:

() 54% disapprove of Bush's veto.

() 57% support sending Bush another funding bill conditioned with timelines.

() 61% support replacing the timelines with enforceable benchmarks.

() 47% favor Congress setting policy for Iraq. 36% favor President Bush.

Overall, support for Bush and his war is stuck at about one out of three Americans. Another quarter want the US out of Iraq immediately, a figure that rises to about 40% for a deadline of March, 2008 regardless of what happens in the meantime.

The remaining 25-30% qualify their support for continued efforts in Iraq on more complicated withdrawal timelines and evidence of the Iraqi government's progress in taking over the fighting and resolving various other political issues.

The snapshot doesn't look good. The trend is even worse for your (ever shrinking) "crowd." You have about four months to turn it around. Good luck.

Liberty · #54 (**permalink**) 05-09-2007, 06:27 AM

Quote:

Originally Posted by jsh1120";p=&quot

I'd be interested in knowing where your information about the "assumptions" of party identification comes from. Frankly, the notion that Rasmussen used one assumption and "other pollsters" used another (37R/37D/26I vs. 35/39/26) seems suspect to me.

Many pollsters keep such data private. Rasmussen used to post his on his website, but I haven't been able to locate it today. Maybe you have to be a registered user to view that now. I would guess that each may have their own Party ID assumptions due to slight/moderate differences in polling.

Quote:

Originally Posted by jsh1120";p=&quot

Are you contending that Rasmussen and "other pollsters" weight their results by assumptions about Party Identification? And if you are, are you saying that all "other pollsters" used the same weighting?

Yes and no. Rasmussen does (at least since 1994), others do as well (Zogby), many do not (Gallup comes to mind).

Zogby (2004): "My polls use a party weight of 39% Democrat, 35% Republican and 26% Independent."
http://zogby.com/news/ReadNews.dbm?ID=859

Quote:

Originally Posted by jsh1120";p=&quot

If, in fact, Rasmussen is using party identification as a weighting variable for sampling (at all) their methodology is especially suspect. Weighting is usually reserved to correcting sample bias in traits that do not change as a result of a change in a person's opinion (e.g. sex, age, geographic location) NOT in terms of a self-described opinion such as party identification. The latter is a DEPENDENT variable, not a legitimate weighting (i.e. independent) variable.

An argument can be made either way. You've describe some limitations for weighting above, and some potential limitations for not weighting in your previous post (systematic sampling bias). Rasmussen went to weighting because he clearly oversampled Republicans in the 2000 race. As I described earlier, the weighting seemed to work out for him in 2004, but then he failed to account for changes in party affiliation, and didn't try to correct that until last year.

His current weighting scheme is sort of a compromise -- it uses a three month weighted average which carries with it the assumption that party affiliation tends to change slowly. This may be accurate for some voters, but there are probably a good number of voters who change more erratically (single issue voters, less educated voters, etc.)

There seemed to be a good number of academic articles pointing out problems in weighting vs. not weighting, with at least a few suggesting various methodological fixes:
http://journals.cambridge.org/action...ine&aid=454439

see also:

http://synapse.princeton.edu/~sam/po..._1nov_8AM.html

Quote:

Therefore I currently think that both weighting by party ID (Rasmussen now) and not weighting at all (Gallup now, Rasmussen in 2000) have serious problems. A better way to weight would be to use a question or questions with fixed answers, such as "Who did you vote for in the last election, Bush, Gore or Nader?" Time magazine does this, but does not weight. Of course, the unreliability of memory is a problem. Zogby Interactive does a sensible version of this: party ID is asked at a different time than candidate preference, which might de-link the variables. If anyone knows of other organizations that go beyond simple party-ID-weighting, please let me know.

Quote:

Originally Posted by jsh1120";p=&quot

I think what you might have meant was that Rasmussen's RESULTS (not weights) were more Republican than the average of other polls. I strongly suspect that is true as I noted earlier.

No, it was definitely his weights (with respect to 2004) -- his new weighting scheme should bring him more in line with other pollsters.

Quote:

Originally Posted by jsh1120";p=&quot

As noted in an earlier post, Rasmussen's Republican tilt was especially evident in the 2006 elections. Shortly after that election I compared Rasmussen's results in individual Senate elections to actual election results. In almost every case Rasmussen overestimated the Republican share of the vote. If memory serves, in some cases this led to incorrect calls, (e.g. I believe Rasmussen called Missouri for Talent rather than McCaskill.) In other cases, when the Republican won, Rasmussen overestimated the margin. (e.g. Kyl in Arizona.)

That's interesting because I believe this would have occurred after he employed the new weighting scheme.

Quote:

The most problematic aspect of predicting election results is not getting Party ID right, it's predicting who will actually vote. Rasmussen's problem appears to be that they are prone to underestimating (a) the proportion of the electorate that is independent; (b) (at least in 2006) the likelihood that independents will actually vote; and (c) (again in 2006) the R vs. D split of independents who do vote.

Yeah, not exactly hard science, eh?

Quote:

More reliable conclusions come about from tracking results of the same organization over time, canceling the effect of methodological differences.

Or, simply averaging conclusions across different polls?