FL: Datamar: McCain (R) 48%, Obama (D) 47%

[J. J.]

I'm not crazy about this poll, but looking at the FL polls might help determine what, if any Bradley effect there is.  What's the MOE?

[J. J.]

How could looking at a poll help you determine an inherently unpollable variable?  Do explain.

You look at the results and compare them with the polling.  I want to see if the poll is off and by how much.

For example, if Obama wins by within two points, we could conclude there wasn't a positive B.E.

If McCain wins by three points, we might be able to conclude there was a slight B.E.

[J. J.]

How could looking at a poll help you determine an inherently unpollable variable?  Do explain.

You look at the results and compare them with the polling.  I want to see if the poll is off and by how much.

For example, if Obama wins by within two points, we could conclude there wasn't a positive B.E.

If McCain wins by three points, we might be able to conclude there was a slight B.E.

What

Are you for real?

Yes, I'm interested in the data that this election might provide.  I'm speaking to FL in particular in this case.  My guess is a very weak B.E., but I expect to be present.

[J. J.]

So, you choose one of the worst polls available without a proven track record.

Then, if it's relatively pro-Obama, then it's simply statistical variation.

But if it's relatively pro-McCain, but within the same MoE then "omg it's because people won't admit they won't vote for the black man?"

MoE only matters when it doesn't reinforce your pre-existing theories?

I want to say a lot of cuss words right now but I'm restraining myself.

Lunar, it one of the polls in FL, and most of the polls are bunched together.  It is a state I'm looking at, not a poll, which is why I listed some states.  MT and UT might be two as well, to see if Obama is underpolling, though there are not a lot of UT polls.

I'm looking for states where there has been a lot of polling and it's fairly closely grouped.

[J. J.]

The polling database currently has Obama +1 in FL.

McCain +3 is still within MoE.

But you still shouldn't have all the polls showing it undercounting by that much (if that happens).  What I'm trying to see is if the polls consistently over counts the black candidate or under counts the white one.  I want to do that in states where there are a lot of polls, relatively stable.

And that 4 point swing might be out of the MOE.

I'm interested in getting the data, almost as much as I'm interested in the result.

[J. J.]

and the only reason the polls could be off is because of race.

nothing else matters


GOD

No, but if there is a pattern across polls and states, it probably is.  There was in 3 out 4 races in 2006.

[J. J.]

But Democrats largely overpolled in many states that year.  You need to subtract other error due to other variables before you isolate your own, you should know that.  Feinstein overpolled by just as much as anyone else.  If Feingold did too (I don't know what his year was) we couldn't conclude an anti-Jewish bias, would we?

Well, we did have a Jewish candidate running on the ticket nationally.  There was no over polling.

[J. J.]

OK, dude, it's over. No Bradley Effect. You were wrong, you got owned by reality yet again, suck up to it.

BRTD, it looks like a 1-3 BE on the nationals, and it did occur in some states.  Ironically, Iowa was one.