Chris Bowers on the Pew Poll bias (turnout model)

[blitzen]

http://mydd.com/

There is no doubt that the protective bubble of massive, double-digit generic ballot polls has been broken. However, I would like to point something else out: in at least the latest Pew poll, the nation is postulated as vastly more conservative than in 2004, and Democrats have pulled ahead anyway.

The Pew poll turnout model suggests parity among self-identifying Democrats and Republicans, just like 2004. It also suggests that the electorate in 2006 will be overwhelmingly conservative compared to 2004. In 2004, according to exit polls, 21% of the electorate self-identified as liberal, and 34% self-identified as conservative. If those same turnout models holds in 2006, then Pew's own voting sample would produce the following result:

Democrats 55%--35% Republicans

Gee, that isn't very close anymore, is it? The problem is that Pew's current turnout model suggests a 10% advancement for Republicans on 2004, with 39% of the electorate as conservative (+5%), and 16% of the electorate as liberal (-5%). There also is no evidence whatsoever to indicate that Democrats and Republicans have remained at parity since 2004. All evidence indicates just the opposite. But anyway, let's assume that the Pew turnout model is correct, and that the nation has grown far, far more conservative in 2006 than it was in 2004. Even then, Democrats have still pulled ahead. Thus, in order to even come close in 2006, Republicans will need to have successfully made the electorate far, far more conservative in 2006 than it was in 2004. In order to win, they will need to go way beyond that.

Here is the deal. There are polls, such as Time and Newsweek, contradicting Pew, Gallup, and ABC. Those polls do not suggest that the nation has grown vastly more conservative than it was in 2004. However, even if all polls suggested that the electorate was far, far more conservative than it was in 2004, Democrats have still pulled ahead anyway. So, the country would have experienced a massive demographic shift that should have been favorable to Republicans, and Democrats have pulled ahead anyway. In a vastly more conservative nation, Democrats pull ahead anyway.

[Trajan]

1. Breathes deep ....

Exhalesssss ....

Thanks .....

[featherman]

2. Exactly...just a poll with it's own idiosyncratic set of internals

Through it into the pot with the others and mix thoroughly...then ignore the result.

[Zynx]

4. I knew something was off in that poll.

I think Gallup has it right in terms of what the final margin will be nationwide. Generic polls are not good predictors of the popular vote. They are merely interesting.

[still_one]

3. The poll also indicates 10% undecided

Those undecided ARE NOT GOING WITH Those in power, they already know what a mess has occurred with them

[Zynx]

5. I actually ended up with a 50-40 margin when I tried his analysis, but

still his point is solid. Also, there are WAYYYYY too many whites in that poll. The national electorate is only about 76% white, possibly less this year and their poll has it at 84%.

[DemocratSinceBirth]

10. At A Glance That Can't Be Remotely True...

Asians, African Americans, and Hispanics make up close to thirty percent of the electorate...

[ConsAreLiars]

6. Talk about "fixing" the numbers !!!

The point of polling surveys is to get a representative sample. There are legitimate reasons, IF you have the data, to adjust the numbers to reflect the fact that certain sector (like those without phones -- repeat, IF you have REAL data on that sector), but this particular "correction" is, well, strange. It assumes that the left-right proportions in the original data were somehow wrong, and needed "fixing." It basically assumes that somehow the sample magically obtained only 2/3 the number of conservatives that are "really" out there, and then boosts their numbers to fix that "error." There are statistics, and then there are outright lies.

[Bluestar]

7. AH-HA!! I knew it!

They have to cheat. They can never concede that we are ahead! "Pew" poll is correct--it stinks to high heaven.

[Zynx]

8. There were reports in 1994 of a Democratic surge late in the election

and it never took hold. I suspect this is pollsters hedging their bets.

[David__77]

9. Good to know.

I was also skeptical of some of the polls showing huge, huge margins for Democrats. Take these things skeptically in all cases.

[Flubadubya]

11. PERMALINK to the story...

http://mydd.com/story/2006/11/6/11649/0828