Rasmussen Tracking Poll [Obama vs McCain]

[Alcon]

Just more proof that Scott Rasmussen is a Republican hack.

Or, a more plausible explanation:  Rasmussen does a hard weight by party ID, and Gallup doesn't, so bounces which affect party ID show up much more in the latter than the former.

[Alcon]

Had the RNC already happened by this point in 2004?

September 2nd was the last day of the RNC in 2004.

[Alcon]

Compared with the Rasmussen Weekly Tracking Poll from exactly 4 years ago, the numbers are completely reversed:

Week ending Sept. 2, 2004:

Bush - 49%
Kerry - 46%

Today:

Obama - 49%
McCain - 46%

http://www.rasmussenreports.com/public_content/politics/election_2004/week_by_week_numbers

Republicans had, well, had mostly completed, their convention by that point.

Not that it really matters much.  2004 was an incredibly galvanized year--results ranged from Bush +2.8 to Kerry +2.8.  Bumps, all movements, were smaller.

[Alcon]

Infomania: the samples for the last few days have been about Obama +2 (last night w/ Palin), Obama +6 (night prior w/ Lieberman) and McCain +2 (beginning of convention).

[Alcon]

I don't think the last two are correct on that, IMHO, and I think last night was slightly more pro-McCain, but who knows.  Are they just guessing?

Mathematical formula.  It's right, save for unpredictability of rounding.

[Alcon]

Suspiciously similar algorithm, in fact.  I forgot to give them a hat-tip

[Alcon]

Even Scott Rasmussen himself thinks his party ID weighting may be underestimating McCains lead:

"For a variety of reasons, the Rasmussen Reports daily Presidential Tracking Poll is less volatile than some other polls and always shows a somewhat smaller convention bounce than reported by others. This is primarily because we weight our results by party identification (see methodology). Looking at the data before adjusting for partisan identification, the Republican convention appears to have created a larger surge in party identification than the Democratic convention the week before. If this lasts, it could have a significant impact on Election 2008."

That's, of course, assuming that the party ID change sticks, which is not necessarily a valid assumption at all.

It's hard to say, but convention boosts are generally not boosts, but bubbles.  A long-term change in party ID would be disastrous for Obama, though.  It's hard to say how likely it is.

[Alcon]

Yea, base consolidation is kind of inevitable, and with a close Presidential election, partisan affiliation stigma is removed a little.  I'm talking a change beyond that, closer to 2004 numbers, though.  That'd be pretty disasterous for the Dems.  I'm skeptical, but it's looking slightly more likely.

[Alcon]

If this Republican gain in enthusiasm continues to the election in some form or another, there will be a tightening in party ID.  I can't say how much it would be, and it won't probably be as extreme as it looks post-convention, but it would be there and it would likely be somewhat significant.

Agreed. If Rasmussen's sample has locked in pre-Palin shifts of Republicans to the Independent category, it's going to be slow to reflect any flow back, should it have happened. I wonder if he's scheduled to realign his partisan breakdowns before the election should this hold for more than the immediate post-convention bounce.

Doesn't he do it (at least) monthly?

[Alcon]

McCain 49% (-1)
Obama 47% (nc)

Rasmussen agrees with Gallup again, and all is right with the world. (Except for the McCain lead.)

Now, it this with the new weighting.

Yesterday was with the new weighting too.

To my understanding, and I could be wrong, it read to me that yesterday numbers, which is today poll, is still on the old weight. But I could be wrong.

There's no reason that they couldn't immediately use the new weighting.  It's not like it affects what sample they take.  That's no guarantee that they did, but I don't see any logical reason why they wouldn't.

[Alcon]

The average McCain net over-performance on weekends in Gallup was +1.2, last time I checked.  Take that as you like.

[Alcon]

The average McCain net over-performance on weekends in Gallup was +1.2, last time I checked.  Take that as you like.

Over-performance?

Well, "over" being relative to his week-day average, not "over" to his actual performance, which is obviously a totally un-knowable number.  But you'd think, week after week, the swings would average out enough to give us a general idea of the "weekend bounce."

[Alcon]

I didn't do it, but IIRC, it went back to the start of the poll.

[Alcon]

I didn't do it, but IIRC, it went back to the start of the poll.

That could be a problem.  It seemed to be there over the summer.

It seemed to be, or seemed not to be?  In any case, I think they did it monthly too, and the results appeared to be mostly noise, with a +1-2 McCain advantage.  Including rounding, that's not especially impressive either way.

[Alcon]

6, 7, 8, points?

According to my chart - Obama 6.19% to 9.18%

so, it went from McCain +5 to Obama >+6 in the span of 24 hours?...time to start ignoring daily numbers, guys

I think this is why they have a three day average.

Way out of MoE.

[Alcon]

Why'd the doubt go away since lunch time?

[Alcon]

Why'd the doubt go away since lunch time?

When I originally got the numbers, I misread them and thought that the result was impossible.

After talking to the person, they assured me of the results, and I agreed.

I trust my numbers.

Can you give a general idea of who this girl/guy is?

[Alcon]

That would be kind of neat.  It's not that I don't trust you, but I prefer to see things by myself.

I just want to verify that it works.  I can destroy the information afterward.  But I'm a believe-it-when-I-see-it kind of guy, y'know.

[Alcon]

No.  Is that a requirement for the hack?

You can always ask for a screenshot.  I would.  PrintScreen button, paste it into Paint, upload it to Imageshack (http://www.imageshack.us/) and then there's not much question.  Still possible to fake, but a lot more convincing for me.

Of course, if your source doesn't want to, that's def. his/her right.  I'll just keep taking it with a half-grain of salt.

[Alcon]

Unless the race has gravitated toward Obama in the last few days, I'm seeing a pretty solid Obama +3, maybe a bit under.

[Alcon]

I wish we just posted the raw margins:

Obama +6.70 to Obama +6.63

Change: McCain +0.07

[Alcon]

Are people seriously arguing McCain's 0.07% gain is statistically significant?

A change doesn't magically become meaningful at the 95% rate.  That's dumb.  Being "out of the MoE" isn't magic, it's just arbitrary.  This change is at the low 50%s confidence rate.  That doesn't make it meaningless.  Statistically, yeah.

But that's a generalized nitpick with "statistical tie"-type stuff.  A 0.07% change is almost meaningless, and is grasping at straws.

[Alcon]

I think we have to look to see if this was statistical noise or the beginnings of a trend.

It reminds me of the French officer who first saw von Kluck's turn at the First Battle of the Marne.  Is this it?  I don't know.

If this is the beginning of a trend, it better be exponential, because otherwise McCain is totally screwed.  

[Alcon]

Not voting for someone because of the color of their skin is not racist now? k.

The Bradley Effect is not voting against someone on the basis of race.  It is saying to a pollster, *I'm undecided* or *I'm voting for Obama* when the voter really believes *I like McCain better, but I don't want the pollster to think I'm racist.*

It could also be "I'm racist, and I don't want the pollster to know."  They're completely indistinguishable.

[Alcon]

Not voting for someone because of the color of their skin is not racist now? k.

The Bradley Effect is not voting against someone on the basis of race.  It is saying to a pollster, *I'm undecided* or *I'm voting for Obama* when the voter really believes *I like McCain better, but I don't want the pollster to think I'm racist.*

It could also be "I'm racist, and I don't want the pollster to know."  They're completely indistinguishable.

That would still come under "I like McCain better."

But it would also very potentially come under "not voting for someone on the basis of race," which you said that the Bradley Effect cannot be.  That's untrue.