Does Damocles have a point? Maybe

[Guns Guns Guns]

Damocles said

Let's put it this way. The exit polling in Ohio in 2008 had a 1 to 3% oversampling for Democrats when it was all said and done, and that was an accurate poll once we had the full results (not the early results which had larger oversampling and were inaccurate). Many of these polls are oversampling Ds by 8% even up to 10% in Ohio. This is ignoring the reality that in D strongholds voter registration has dropped by a few hundred thousand...

If we correct for oversampling "errors" we find that the Ohio race is pretty much an even go back when they were trying to say he had "double digit" leads and fully within the margin of error for the poll...

If they are still using this same model, and Romney is actually in the lead it doesn't bode well for the Obama team.

Does he have a point?

And if so, is it possible the "oversampling" could go the other way?

The dramatic Romney surge in the latest Pew poll may have less to do with besting Obama in the first debate than with some dramatic shift in party identification in Pew polls.

According to numbers reported by Pew, Democrats fell from 37 percent in the September poll to 32 percent in October while Republicans rose from 28 to 33 percent among registered voters.

With such numbers and high support for the nominee among partisans, it's no surprise that Romney leads Obama.

With the same partisan breakdown in October as in September, Obama would still be leading, even among likely voters.

So what's behind the Republican surge in party identification?

As far as I can tell, the September-October shift is the largest recorded by the Pew trend line.

But is it a real surge?

If so, Republicans would have scored a partisan realignment of sorts, all in the wake of a terrific debate performance.

That seems far-fetched. More likely, the sample for the October poll just ended up with more Republicans and fewer Democrats than before.

Maybe because of random luck, or perhaps because Romney's strong and Obama's weak debate performance encouraged Republicans to respond to the poll and discouraged Democrats.

Let's see whether other polls also show a seismic shift of the partisan balance right now. If not, the switch in the horse race claimed by the latest Pew poll may not be for real.



http://www.huffingtonpost.com/helmut-norpoth/pew-poll_b_1949879.html?utm_hp_ref=elections-2012

 

[Damocles]

This ignores the use of models,the sampling is changed by multipliers according to the model. Let's say I'm running the poll, I get more Rs than Ds when I call and the model says that we need 8% more Ds than Rs we simply multiply the D numbers by a whole plus a fractional to make it match up as if our sample matched. This creates a portion of the margin of error that you read about that goes along with each poll...

Again, if they are still using the same models (I haven't read the full polls so I don't know hence the "if") then the current numbers do not bode well for Obama.

 

[Guns Guns Guns]

This ignores the use of models, the sampling is changed by multipliers according to the model. Again, if they are still using the same models (I haven't read the full polls so I don't know hence the "if") then the current numbers do not bode well for Obama.

According to numbers reported by Pew, Democrats fell from 37 percent in the September poll to 32 percent in October while Republicans rose from 28 to 33 percent among registered voters.

Let's see whether other polls also show a seismic shift of the partisan balance right now. If not, the switch in the horse race claimed by the latest Pew poll may not be for real.



http://www.huffingtonpost.com/helmut...elections-2012